Journal Research - SIJRsijr.in/articles/june2018/Sahyadri-Journal-Research-Vol-4-Issue-1.pdf · Dr....

Sahyadri Journal of Research JUNE 2018VOL.4 ISSUE 1

SIJRJournal

Research

Research Papers

Review Papers

Scientific Articles

International

ISSN 2456-186X

Advisors

Editorial Board

Editors

Dr. D L Prabhakara (Director)

Dr. Manjappa Sarathi (Director-Consultancy)

Dr. Umesh M Bhushi (Director- Strategy & Planning)

Dr. R Srinivasa Rao Kunte (Principal)

Dr. Rajesha S (Dean- Academic)

Mr. Ravichandra K. Rangappa (Dean-Industry)

Mr. Ramesh K G

Dr. Shriganesh Prabhu, TIFR, Mumbai

Dr. Dinesh Kabra, IITB, Mumbai

Dr. Pushpalatha K - Editor-in-Chief

Dr. Ramakrishna Sharma

Dr. Navin N Bappalige

Dr. Rathishchandra Gatti

Dr. Ravindra Babu G

Dr. Niraj Joshi

Dr. M B Savitha

Mr. S N Bharath Bhushan

Mr. Balaji N

Mrs. Geetha S D

Mrs. Megha N

Mrs. Aysha Shabana

Mrs. Smitha

Disclaimer: The individual authors are solely responsible for

infringement, if any, of Intellectual Property Rights of third

parties. The views expressed are those of the authors. Facts

and opinions published in SIJR express solely the opinions of

the respective authors. Authors are responsible for citing of

sources and accuracy of references and bibliographies.

Although every effort will be made by the editorial board to

see that no inaccurate or misleading data, opinion or

statements appear in this journal, the data and opinions

appearing in the articles including editorials and

advertisements are the responsibility of the contributors

concerned. The editorial board accepts no liability

whatsoever for the consequences of any such inaccurate or

misleading data, information, opinion or statements.

Contents

SAHYADRI International Journal of Research | Vol 4 | Issue 1 | June 2018

Contents

Research / Review Articles

Editorial 1

Nano Ceramic Matrix Composite Development and Its

Applications 2-7

Isolation of Omega 3 Fatty acid from Fish oil 8-11

Useful Application of Plastic Waste in Composite Brick

Manufacturing 12-14

Aptitude Question Solver: AptitudeQS 15-21

An Integrated Approach for Personality Analysis using

OCR and Text Mining 22-26

Recognition of Overlapping Sound Events 27-31

IoT Based Energy & Waste Management for Smart Cities 32-35

ADS Recommendation Using Data Mining 36-39

Multilevel Encryption for Cloud Storage 40-42

War Field Spy Robot 43-45

Accessing Spatial Variability of SOC Content Using

GIS Based Interpolation Techniques 46-49

Study On Strength Of Hybrid Concrete Beam 50-52

Vol. 4, Issue 1ISSN 2456-186X (Online)ISSN Pending (Print)

Mailing Address:Editor Sahyadri International Journal of ResearchSahyadri campus, Adyar, Mangalore - 575 007, IndiaE-mail: [email protected]: www.sijr.in

SAHYADRIInternational Journal of Research

It gives us immense pleasure to bring out Volume 4, Issue 1 of Sahyadri International Journal of Research

(SIJR). The journal covers the wide disciplines in science, engineering and technology. SIJR is published

biannually and is an open access journal available online. The focus of the Journal is to motivate the

researchers of various disciplines to publish their quality research. The most important disciplines in

which we would focus are: Physics, Chemistry, Applied Mathematics, Electronics and Communication,

Mechanical Engineering, Civil Engineering and Computer Science and Engineering. This issue has the

research and review articles of current trends in various disciplines.

In this occasion I would like to express heartfelt appreciation to all authors and reviewers of the SIJR on

behalf of the entire editorial board and the publisher. It was with the mere co-operation, enthusiasm,

and spirit of the authors and reviewers we could make SIJR a grand success. I thank all the authors in

considering and trusting SIJR as the platform for publishing their valuable work. I also thank for their

kind co-operation extended during the various stages of processing of the manuscript in SIJR.

The reviewing of a manuscript is very essential to assure the quality of the manuscript published in any

journal. The inputs of reviewers are frequently used in improving the quality of a submitted manuscript.

I thank all reviewers for their excellent contributions and support for the journal.

I also wish to acknowledge the contributions made by the dedicated members of our Editorial Board,

the invaluable support given by advisors and the Sahyadri Management, and the hard working,

professional staff of the publishing office.

Finally, I would like to thank the readers of SIJR, for your interest in the journal. We welcome your

valuable feedback and ideas for further improvement of SIJR.

Editor-in-Chief

Editorial

1SAHYADRI International Journal of Research | Vol 4 | Issue 1 | June 2018

SAHYADRI INTERNATIONAL JOURNAL OF RESEARCH, VOL 4, ISSUE 1, 2018

ISSN: 2456-186X, Published Online June, 2018 (http://www.sijr.in/)

2

Nano Ceramic Matrix Composite Development and

Its Applications

Priyanka Singh1, Nathi Ram Chauhan2, Rajesha3*

1 JSS Academy of Technical Education, Noida, 201301, India 2 Indira Gandhi Delhi Technical University for Women (IGDTUW), New Delhi, 110006, India

3Sahyadri College of Engineering & Management, Mangaluru, 575007, India *Email: [email protected]

ABSTRACT

The nano ceramic matrix composite is used in variety of applications due to its unique physical properties and capability

to perform better at elevated temperature. Research is going on worldwide to understand the characteristics of such

composite material and better fabrication methods. Some of the fabrication methods, physical properties and

microstructure characteristic have been reported so far and process needs further investigation for better understanding

the characteristics of such matrix. In this paper, emphasis is given on are fabrication methods, physical behavior and

probable application of nano composites reported so far and efforts has been made to indicate the future scope of study

on nano ceramic composite.

Keywords: Ceramic composite, microstructure, Strengthening potential

1. INTRODUCTION

Nanocomposite are materials in which at least one of the

dimensions is in nano meter range. Reduction of reinforcement

to nano range makes interaction of particles with dislocation

more significant and leads to improvement in various

properties which have proven to be useful for a wide range of

critical applications. Extensive research is going on worldwide

to improve the desirable properties of ceramics by adding

reinforcements and limiting their inherent weaknesses. The

Ceramic composites based on SiC, Si3N4, TiN, TiB2,

Alumina, Zirconia, TiC and many other have been developed

and mechanical properties such as tensile, compressive, fatigue

resistance, fracture toughness, R-Curve behavior, creep have

been studied by various researcher for different CMCs. At

elevated temperature better physical properties of CMCs are

obtained as compared to monolithic composite due to

nanoscale reinforcement (increasing the surface to volume

ratio) which makes it suitable for ceramic cutting tools, wear

resistive components, radiation resistive ceramic component,

aerospace, & automobile components and other applications.

Fracture toughness decreases when components are subjected

to high temperature limits in some application, however, it is

also reported that toughness improves with reinforcement in

CMCs.

Main challenges are involved in synthesis of these materials

which require advanced processing techniques. These

challenges can be either due to the characteristic of reinforcing

phase or limited processing techniques.

This paper is aimed at reviewing the fabrication methods,

mechanical properties, strengthening mechanism and

application in field of ceramic matrix composite reported by

various investigators.

2. PREPARATION METHODS AND

PROPERTIES

Self-propagating high-temperature synthesis (SHS) method is

basically used for producing inorganic compounds by

exothermic reactions, usually involving salts. Synthesis of

nano-sized precursor powders is performed by special

techniques like high energy ball milling process, sol-gel

processing, gas condensation process, inert SHS reaction and

infiltration technique. Consolidation difficulties of nano-sized

powders, caused by their higher propensity to form strong

agglomerates because of very high ratio of surface area to

volume may be reduced by using various specialized

techniques while their fabrication.

Infiltration methods are used to fabricate ceramic matrix

composites reinforced with long fibers. This type of Ceramic

matrix is formed with a fluid (liquid or gases) which is

infiltrated into the fiber structure. The surfaces of the

reinforcing fibers are coated with a deboning interphase prior

to infiltration which weakens the bonds interface between

matrix materials and the fiber. Weak bonding allows these

long fibers to slide in the matrix and this results in the

prevention of brittle fracture [1].

mailto:rajeshashivanna@


3

In Sol gel processing, sol is formed by dispersing the matrix

and reinforcing particles in the liquid. The deposition of this

sol solution results in coating on substrates by spraying,

dipping or spinning process. Gel is formed from evaporation

of solvents and particles or ions join together to form network.

Thermal treatment is done to enhance mechanical properties. It

is low temperature process and generates highly pure and well

controlled ceramics [2].

Nano SiC/TiN composite was prepared by sol gel method

using, TiN nanopowder as toughening phase, b-SiC nano

powder as matrix phase and YAG (synthetic yttrium aluminum

garnet) as sintering additive. Combination of aqueous slurry

with spray-drying was used to prepare Nano SiC based

granules. It was then uniaxially pressed at 160 MPa for 10s

and pressed isostatically with a pressure of 250 MPa for 300

seconds [3]. TiC x/2009Al particles were ball milled at a

speed of 100 rpm for 50 hours. Combustion synthesis was

conducted in self-made vacuum vessel with a vacuum degree

about 0.5 tar. Composites were extruded at 773 K under an

extrusion ratio of 16:1 [4].

Al2O3/ AlN particle were ball milled in ethanol for 48 hours

and then dried [5]. In Al2O3/TiC nano composite, the nano

scale TiC powders were prepared into suspension using

dispersant as polyethylene glycol and alcohol as the dispersing

medium. After that, micro-scale TiC, Al2O3, and cobalt was

added into the suspensions. After ball milling for 48 hours, dry

type evaporator with vacuum was used for drying [6].

Alumina/zirconia/nano-TiO2 ceramic composites were ball

milled for 2 hours for mixing. The mixed powder was semi-

dried and pressed at 100 MPa [7].

In ZrC/SiC composite, ZrC powder was first heated in air

for about 10 hours at 250°C and mixed with SiC powder and

pyrolysed under argon at 600°C for 5 hours and further ball

milled [8]. In α- Al2O3/Si3N4 nano composite, α- Al2O3 and

Si3N4 were mixed and ball-milled with ethanol for 72 hours

and dried in a vacuum dry evaporator at 110°C [9].

Earlier most of the nano ceramic matrix composite were

developed using hot isostatic pressing (HIP), hot pressing and

sinter forging. Sintering kinetics is increased by the application

of working pressure above atmospheric pressure. Limitation of

conventional sintering techniques is the formation of strong

agglomerates of nano sized powder due to extremely high ratio

of surface area to volume which results in difficulty in

consolidation leading to poor mixing, inhomogeneous packing,

residual porosity and poor density.

With the advancement in fabrication technique today, most

of the nano ceramic composite are fabricated by Spark Plasma

Sintering (SPS). Heating rapidly to sintering temperature

(Temperature lower than conventional sintering) and less

holding time results in good control of the fine grain size,

retention of nano scaled microstructure and high relative

densities.

SPS of ZrC/SiC for developing ceramic composite was done

at 1950°C temperature for 15 minutes under vacuum with a

pressure of 50 MPa. The sample was then cooled to 1200°C

with 25°C/min in order to reduce any quenching stresses. It

was observed during sintering of the composite that overall

strain associated to the applied load was preferably

accommodated by plastic deformation of ZrC to a much lesser

degree by the formation of stacking faults through phase

transition operating within SiC. Sintering temperature in this

process was high for retaining better properties [8].

ZrC/SiC Composite has been developed by solution based

processing using divinylbenzene, polycarbosilane

polyzirconoxane, to obtain ZS precursor prepared at 200°C

and then heat-treated for 2 hours to a temperature of 1500°C

with 5°C/min heating rate in argon atmosphere. Highly

crystalline ZrC and SiC phases were observed in ceramic

powders with 100–400 nm particle size. Distribution of Zr, Si,

C was uniform at different sites in the powder. ZrC/SiC weight

ratio was varied to control different element in the sample and

precursors with good stability and processibility was used in

polymer infiltration pyrolysis process. The prepared composite

sample with less carbon content exhibited good oxidation

resistance at high temperature [10].

In another method of fabricating ZrC/SiC Composite, tape-

casting process and vacuum hot-pressing was used. In this

method, ZrC and SiC powder was used as raw material for

tape casting. First, sols of 5 wt% polyvinyl butyral resin (PVB)

and 5 wt% polyethylene glycol were dissolved in ethanol as

the adhesive and plasticizer, respectively. The mixture was

placed in a water bath heated at 60 °C to obtain a homogenous

material. Second, 20 wt% ZrC powder was introduced to the

above mixture and dispersed in ethanol by ultrasonic agitation

for 2 hours to form a homogenous mixture with a certain

viscosity. Third, tape casting mould was used for placement of

sol on it at room temperature for 8 hours and were cut into

slices in the form of sheets. The same steps were carried out

for SiC and then ZrC and SiC sheets were alternately stacked.

The stacked sheets were heated at 550 °C for 60 min with

heating rate of 10 °C/min to remove the binder. Finally,

vacuum hot pressing furnace at 1700 °C was used for sintering

the laminated sample for 90 min under an applied pressure of

20 MPa. Fracture behavior of laminated ZrC–SiC ceramics

was quite different from brittle fracture and showed a non-

catastrophic failure behavior. The crack deflection extended

the crack propagation path and increased the energy

consumption capacity of laminated ceramics and thus

increased the fracture toughness [11].

Sintering of Al2O3/TiC was done between 1600ºC - 1700ºC

in sintering furnace in vacuum for 10-30 minutes [6]. Wetting

between metal interfaces and ceramic was improved by the

addition of cobalt in the composite. Cobalt presence at grain

boundary not only prevented TiC and Al2O3 from growing but

also restricted the reaction between TiC and Al2O3 during the

process, vapour phase formed during processing caused pores

in Al2O3/TiC ceramics. At maximum sintering temperature

cobalt liquefied and filled the pores in grains exhibiting better

density. However, in Al2O3/TiC composite fabricated by

Spark Plasma Sintering at different temperatures 1100 ºC,

1200 ºC, 1400 ºC, 1500 ºC for 3 minutes with 50 ºC /minute

heating rate at a load of 60 MPa, complete densification was

observed at lower sintering temperature than conventional

sintering [12].

Another method of fabrication of Al2O3/TiC nano

composite has been reported, using hot pressing. The samples


4

were prepared using alpha alumina, micro TiC, nano TiC and

Cobalt. Firstly the TiC powder (nano scale) was prepared into

suspension using dispersant as polyethylene glycol and alcohol

as dispersing powder and subjected to ultrasonic dispersion for

20 minutes. PH value of 9 was obtained by addition of

NH3.H2O. It was then mixed and ball milled for 48hours and

dried in dry type evaporator in vacuum. The dried powder was

poured in graphite die and hot pressed with a pressure of 32

MPa at 1650°C in vacuum sintering furnace for 20 minutes.

Composite showed better wear resistance and fatigue behavior

to be used as tool material [13].

Alumina/zirconia/nano-TiO2 nano composite sintered at

temperature of 1600 ºC for 1 hour with 5 ºC /min heating rate

from room temperature to 1000 ºC, and 2.5 ºC /min heating

rate from 1000 ºC to 1600 ºC. Addition of TiO2 formed

Al2TiO5 ceramic which showed resistance to thermal shock

[7]. However, high sintering temperature decomposed Al2TiO5

into Al2O3 and TiO2 leading to increase in apparent porosity.

Two step pressureless sintering for fabrication of Nano-

SiC/TiN nano composites reported in [3].Composites were

first sintered at 1900ºC for 15 minutes in vacuum furnace and

second step sintering at 1700 ºC, 1750 ºC, 1800 ºC and

1850ºC for 45 min at each temperature. YAG was used as

sintering additive for densification. The densification in

second step sintering was observed by slower grain boundary

diffusion which restricted the grain growth. Y2O3 was used as

sintering additive for fabrication of α- Al2O3/Si3N4 by hot

pressing at 1450°C temperature in vacuum for 30 minutes

under a pressure of 32 MPa. Due to the covalent bonding

nature of Si3N4, it was sintered to a high density by addition of

Y2O3 as sintering additive [9]. Al2O3/AIN composite was

fabricated at 1600 ºC under an applied pressure of 30 MPa for

180 minutes [5].

Fabrication of nano ceramic composite by stir casting

method is reported by few investigators. A resistance furnace

equipped with inert gas injection instrument and a graphite

stirring system was used to fabricate TiB2/ A356 Al nano

composite at casting temperatures of 750 ºC, 800 ºC and 900

ºC. An increment of volume fraction of reinforcements and

decrement of the particle’s size lead to the increment of the

porosity content. Decreased density and dislocation pile up

phenomena was observed due to the presence of slip band in

the matrix. Stir casting method of fabrication is not suitable for

fabricating nano ceramic composite [14].

Zirconia toughened alumina nanocomposite was developed

by SPS at 1100°C with heating rate of 500°C/min via

combination of High Energy Ball Milling process followed by

SPS of γ-alumina powders added with zirconia and yttria. Full

densification was observed in nanocomposites which was

SPSed at 1100°C. The hardness and toughness was increased

which is almost three times than monolithic alumina due to

addition of zirconia [15].

2.1 Mechanical Properties

Development of new material implies value-added to the

physical properties. CMCs shows improvement in mechanical

properties due to superior strength, hardness, abrasion

resistance and chemical inertness in comparison to other

materials. Reduction of grain size to nano meter range leads to

hardness increment which improves mechanical properties as

reduced wear behaviour, higher fracture toughness and higher

resistance against abrasion. Physical properties of CMCs vary

with fabrication techniques indicating that better the

fabrication method better is the development of CMC material.

The reduction of grain size to nano metric range of SiC in the

composite ZrC/SiC fabricated by spark plasma sintering leads

to an improvement of fracture toughness and flexural strength

due to densification at high sintering temperature, the overall

strain associated with the applied load was preferentially

accommodated by the plastic deformation of ZrC [8]. The

creep behavior remained unaffected upto 1600 ºC in Al2O3–

TiC composites fabricated by the same method at lower

sintering temperature showed homogenous distribution of

titanium carbide in the alumina matrix, there was no new phase

formations during sintering. The fully dense Al2O3/TiC

composite showed higher Young’s modulus and hardness

values. Scratch test on the bulk sample showed that sample

sintered at higher temperature had better scratch resistance due

to strong bonding of the particle [12]. Addition of Cobalt in

Al2O3/TiC fabricated by chemical deposition method revealed

greater improvement in fracture strength of composite.

Fracture toughness was also increased reducing the crack

propagation [6]. Two step pressure less sintering of Nano-

SiC/TiN nano composites densified the composite with

improved properties as Vicker’s hardness, bending strength

and fracture toughness. Similarly addition of YAG in

composite enhanced toughness by crack deflection [3]. Hot

pressing of Al2O3/AIN composite at high temperature showed

significant improvement in the properties like flexural

strength, fracture toughness and relative density [5].

Hot pressed α- Al2O3/Si3N4 nano composite at lower

temperature revealed optimum mechanical properties as

flexural strength, hardness, fracture toughness and high

relative density [9]. Increased in Si3N4 content beyond

optimum value lead to crack formation on applied load which

weakened the grain boundary strength and caused stress

concentration. Flexural strength was decreased due to

propagation of cracks easily. Sintering of

alumina/zirconia/nano-TiO2 ceramic composites exhibited

higher density and less porosity. Hardness was significantly

improved along with resistance to thermal shock [7].

The strengthening mechanism of Nano ceramic matrix

composite is used to analyze the physical properties of the

material. Orowan Strengthening Mechanism [16-18] has been

used to analyze the effect of reinforcement and inter

particulate spacing of secondary phase disperiods. Orowan

strengthening is caused by resistance of particles to the passing

of dislocations. Creep resistance raises considerably even for a

small volume fraction due to dispersion of fine insoluble

particle in the ceramic matrix. Ceramic reinforcement particle

pin the crossing dislocation and assist dislocation in bowing

around the particles. For simulation of better mechanical


5

properties of nano ceramic composite Orowan loop

mechanism is used.

Yield strength σR of composite can be given by

(1)

Where b - Burgers vector

dp -particle diameter

G - Shear modulus

Vp - volume fraction of reinforcement.

The Hall Petch Mechanism relates grain size with strength.

Decreasing the grain size leads to increase in strength and

ductility. Fracture resistance also generally improves with

reductions in grain size. The yield strength of many metals and

their alloys has been found to vary with grain size according to

the Hall-Petch relationship:

= + (2)

Where ky - Hall-Petch coefficient (material constant)

D is the grain diameter

σy - is the yield strength of an imaginary polycrystalline metal

having an infinite grain size. Grain boundary play critical role in the yield stress of

material. There can be several different deformation modes

associated with different grain size, grain shape, temperature,

stress state and grain boundary structures [16][19].

Dislocations can generate in the alloy matrix during processing

due to coefficient of thermal expansion mismatch (CTE)

between the matrix and reinforcement phase and induce

residual stresses [16][20]

Mismatching of strain due to difference in CTE values of

matrix and particles leads to the generation of thermal stresses

at the interface which makes plastic deformation difficult

leading to enhancement in flow stress and hardness. The effect

of mismatch strain is given by

= β b (3)

where β - strengthening coefficient

α- difference between CTE of matrix element and

reinforcement element

ΔT-difference between the processing and the testing

temperatures

b-Burgers vector

Gm -shear modulus

dp -particle diameter

νp -poissons ratio. The Griffith’s energy gives the basic explanation for the

strengthening and toughening mechanisms of composite based

on equilibrium between the fracture energy and energy release

rate [21]. Rising R-curve behavior is observed in many

ceramic based composites.

Crack resistance of this class of material is expressed by

KR (Δ α)= Ki + Δ KR (Δ α) (4)

Where KR (Δ α)-fracture toughness of the material which

shows R curve behavior

Ki- intrinsic fracture toughness

ΔKR (Δα)- extrinsic increase of fracture toughness after a

definite extension beyond the initial crack tip Δα.

Griffith–Irwin formula for materials with an R-curve is given

by

(5)

γi and γR are the intrinsic and extrinsic fracture energy per unit

area of the cracked surface, respectively.

Frontal process zone (FPZ) ahead of crack tip is composed of

nano cracks rather than dislocations. Fracture toughness can be

increased by expanding the size of FPZ. The left side of

Griffith–Irwin equation indicates release rate of critical energy

beyond a definite crack extension in materials having R-curve

behavior.

3. APPLICATIONS

Production of ceramic matrix composite using nano

technology can be made more useful, cost effective and high

ending in the service conditions. Nano ceramic matrix

composite is used in variety of application based on its

structure, properties, strengthening and toughening

mechanism. Some of the applications of various ceramics are

discussed in this section.

The Al2O3-based composite is used as tool materials for

high speed machining compared to the traditional cemented

carbide cutting tools and high-speed steel because of the good

mechanical properties as high hardness, high corrosion and

wear resistance [22].

Nickel based alloys are typically being used in high

pressure turbines, mobile phones, medical equipment,

transport, buildings and aerospace application due to its better

corrosion resistance, toughness strength at variable

temperatures. Materials such as titanium and nickel based

alloys having properties of light weight and high strength to

weight ratios are highly demanded in aerospace, automotive

and power industries for their critical components [23].

Ceramics based on the carbides of the Group 4 transition

metals are of great interest for applications at high

temperatures in both aerospace and nuclear applications [8].

SiC nano composite are used in applications requiring high

endurance such as car brakes, car clutches and ceramic plates

in bulletproof vests [24].

Aluminum oxide referred to as alumina are used in

structural applications. Alpha phase alumina is stiffest and the

strongest among all oxide ceramics having high hardness,

high refractoriness, excellent dielectric and thermal properties


6

which makes it a suitable choice for different applications

such as in manufacturing of wear pads, grinding media, seal

rings, high temperature electrical insulator and aerospace

components [25].

Alumina based ceramic, (ZrO2) toughened improves the

toughness and resistance to fracture and are used as tool material

[26]. Zirconia (ZrO2) based ceramic material with adequate

mechanical properties are used for manufacturing of medical

devices [27].

Alumina Mullite ceramic is used as a traditional and

advanced ceramic material because of having favorable thermal

and mechanical properties for structural applications [28].

Silicon nitride (Si3N4) ceramic material having excellent

thermo mechanical property is most suitable for structural

applications, bearings, cutting tools and engine components. It

performs better at high temperature retaining high strength and

creep resistance and low thermal expansion coefficient giving

good thermal shock resistance [9]. Titanium based ceramic are

used in the manufacturing of wear-resistant tools, cutting tools

and coating for abrasive steel bearings [29].

4. CONCLUSION

In this paper effort has been made to understand the CMCs

process of development and its characteristics. The

application of various CMCs literature report has been

discussed and following conclusions have been made.

Spark Plasma sintering is found to be most advanced

technique for fabrication of CMC which restricts the grain

growth and densify the material at high temperature providing

better strength.

Effect of various phases play an important role in internal

stress distribution within the composite affecting the

properties. Mechanical behavior of nano composite is

presented which showed strength and hardness increased by

nano particle reinforcement. However, flexural strength is

decreased in some cases with the increase in reinforcement

percentage.

Exclusive study and development of better CMCs

fabrication techniques are needed for improving overall

properties to be used for wide range of application.

REFERENCES

[1] Rosso M, “Ceramic and Metal Matrix Composite Routes

& properties”, Journal of material Processing and

technology, 175 (2006) 36-375

[2] Schmidth H, “Chemistry of material preparation by sol-gel

process”, Journal of Non Crystalline solids Volume 100

(1988) 51-64

[3] Xingzhong Guo, Hui Yang, Xiaoyi Zhu and Lingjie

Zhang, “Preparation and properties of nano-SiC-based

ceramic composites containing nano-TiN”, Science

Direct, China, November 2012

[4] Lei Wang, Feng Qiu, Jingyuan Liu, Huiyuan Wang,

Jinguo Wang, Lin Zhu, Qichuan Jiang, “Microstructure

and tensile properties of in situ synthesized nano-sized

TiCx/2009Al composites” , Science Direct, China, April

2015

[5] Qinggang Li, Chao Wu, Zhi Wang, “Mechanical

properties and microstructures of Nano-Al2O3 particles

reinforced Al2O3/AlN composite”, Journal of Alloys and

Compounds 636(2015) 20-23, China , February 2015

[6] Huang L.P., J. Li, “Properties of cobalt-reinforced

Al2O3–TiC ceramic matrix composite made via a new

processing route” Science Direct , China, September 1998

[7] Wahsh M.M.S., R.M. Khattab, M.F. Zawrah “Sintering

and technological properties of Alumina/zirconia/nano-

TiO2 ceramic composites”, Science Direct, Egypt,

December 2012 .

[8] Antou G., M.D. Ohin , R. Lucas, G. Trolliard, W.J. Clegg,

S. Foucaud, A. Maître, “Thermomechanical properties of

a spark plasma sintered ZrC–SiC composite obtained by a

precursor derived ceramic route”, Science Direct,

Material Science and Engineering A 643 (2015) 1-11,

UK, July 2015

[9] Xiaolan Bai, Chuanzhen Huang , Jun Wang, Bin Zou,

Hanlian Liu, “Fabrication and characterization of Si3N4

reinforced Al2O3-based ceramic tool materials”, Science

Direct, China, June 2015

[10] Dan Liu, Wen-Feng Qiu, Tao Cai , Ya-nan Sun, Ai-Jun

Zhao, and Tong Zhao, “Synthesis, Characterization, and

Microstructure of ZrC/SiC Composite Ceramics via

Liquid Precursor Conversion Method”, China, 2014

[11] Yuanyuan Li , Qinggang Li , Zhi Wang , Shifeng Huang ,

Xin Cheng, “Fabrication and crack propagation behaviour

of ZrC/SiC laminated Composite”, Material Science and

Engineering A647 (2015) 1-6

[12] Rohit Kumar, A.K.Chaubey, Sivaiah Bathula, B. B. Jha

and Ajay Dhar, “Synthesis and characterization of Al2O3-

TiC nano-composite by spark plasma sintering”,

IJRMHM, India, August 2015

[13] Zengbin Yin, Chuanzheng Yuan, Bin Zou, Hanlian Liu,

Hongtao Zhu, “ Cutting performance and life prediction of

an Al2O3/TiC micro-nano-composite ceramic tool when

machining Austenitic steel”, China, 2014

[14] Karbalaei Akbari M., H.R. Baharvandi, K.

Shirvanimoghaddam, “Tensile and fracture behavior of

nano/micro TiB2 particle reinforced casting A356

aluminum alloy composites”, Materials and design 66

(2015)150-161, Iran, October 2014

[15] GD Zhan, J Kuntz, J Wan, J Garay and A K Mukherjee ,

American Ceramic Society 86 [1] (2003

[16] M. Habibnejad-Korayema, R. Mahmudi , W.J. Poole ,

“Enhanced properties of Mg-based nano-composites

reinforced with Al2O3 nano-particles”, Materials Science

and Engineering Iran, A 519 (2009) 198–203

[17] Zhang, Z, Chen.D.L, “Contribution of Orowan

strengthening effect in particulate-reinforced metal matrix


7

nanocomposites”, Material Science and Engineering A

(2008), pp.148-152, 483-484

[18] Zhang, Z, Chen. D.L, “Contribution of Orowan

strengthening effect in particulate-reinforced metal matrix

nanocomposites: A model for predicting their yield

strength”, A Material Science and Engineering A (2006)

54, pp.1321-1326

[19] John Bosco R, “Melting points, mechanical properties of

nanoparticles and Hall Petch relationship for

nanostructured materials”, Sastr University, Bangaluru

[20] M. Habibnejad-Korayem, R. Mahmudi , W.J. Poole,

“Enhanced properties of Mg-based nano-composites

reinforced with Al203 nano particled”, Material Science

and Engineering A 519 (2009) 198-203

[21] Hideo Awaji, Seong-Min Choi, Eisuke Yagi,

“Mechanisms of toughening and strengthening in ceramic-

based nanocomposites”, Science Direct, Japan, February

2002.

[22] Zengbin Yin, Chuanzhen Huang , Bin Zou , Hanlian Liu,

Hongtao Zhu, Jun Wang , “Study of the mechanical

properties, strengthening and toughening

mechanisms of Al2O3/TiC micro-nano-composite ceramic

tool material”, Elsevier, Material Science and

Engineering A 577 (2013) 9-15

[23] Tresa M. Pollock and Sammy Tin, "Nickel-Based

Superalloys for Advanced Turbine Engines :Chemistry,

Microstructure and Properties", Journal of Propulsion and

Power, Vol. 22, No. 2 (2006), 361-374.

[24] Kumar Sidheshwar , Singh Bhagirathi, “Study on SiC

produced from rice husk as reinforcing agent”

,Department of Metallurgical & Materials Engineering,

NIT Rourkela

[25] Cheng-Liang Huang , Jun-Jie Wang , Chi-Yuen Huang ,

“Sintering behavior and microwave dielectric properties

of nano alpha-alumina”, Materials Letter 59 (2005) 3746-

3749

[26] A. Senthil Kumar , A. Raja Durai , T. Sornakumar,

“Machinability of hardened steel using alumina based

ceramic cutting tools”, International Journal of Refractory

Metals and Hard Materials 21 (2003) 109-117

[27] J.Chevalier, L. Gremillard, “Ceramics for medical

applications: A picture for the next 20 years

”, Journal of European Ceramic Society 29 (2009) 1245-

1255

[28] Eugene Medvedovski, “Alumina–mullite ceramics for

structural applications”, Ceramic International 32 (2006)

369-375

[29] Bellosi, R. Calzavarini, M.G. Faga, F. Monteverde, C.

Zancolò, G.E. D’Errico, “Characterisation and application

of titanium carbonitride-based cutting tools”, Journal of

Materials Processing Technology 143-144 (2003) 527-

532



8

Isolation of Omega 3 Fatty Acid from Fish Oil

Sahana C Hiremath1, C T Puttaswamy1, Sarathi Manjappa2*, Ajith B S 2

1Department of Chemical Engineering, BMS College of Engineering, Basavangudi, Bengaluru, India 2Research Centre, Sahyadri College of Engineering and Management, Mangaluru, India - 575007

*Email: [email protected]

ABSTRACT

Omega 3 fatty acids containing Eicosapentaenoic acid (EPA) and Docosahexaenoic acid (DHA) play a beneficial role in

human health and as functional food. In this study fish oil was purchased from Fish oil and Fish meal extraction factory,

Mangalore. The standards were purchased from sigma Aldrich. The fish oil was further characterized by various

analytical techniques such as High performance Liquid chromatography (HPLC), thin layer chromatography (TLC),

Fourier infrared spectroscopy (FTIR) in order to isolate omega 3 fatty acids from fish oil. Quality of fish oil was analysed

in order to determine impurities, acid value and peroxide value of fish oil which was in recommended range. Traditional

extraction of omega 3 fatty acids are expensive and time consuming. Basically high yield and concentrated omega 3 fatty

acids from sardine fish oil were obtained by supercritical fluid extraction, urea complexation, low temperature

crystallization and molecular distillation, enzymatic hydrolysis. The main aim to enhance and improve the concentration

of polyunsaturated omega 3 fatty acids from saradine fish oil and production of biodiesel through Tansesterification after

recovery of omega 3 fatty acids.

Keywords: Omega 3 Fatty acid, Enzymatic Hydrolysis, Transesterification

1. INTRODUCTION

Fish is one of the perishable human food overall the world. Its

nutritional value includes phosphorous, magnesium, selenium.

Fish oil which is rich in EPA and DHA helps in the reduction

of heart attack, strokes, abnormal heart rhythms, death [1].

Omega 3 are the polyunsaturated fatty acids having carbon

carbon as double bond in its position [2]. Determination of

omega lipids in the fish oil comprises of several steps

including extraction, hydrolysis and derivatization and making

UV active compound for measured by HPLC [3].

HPLC with UV detector was a challenging task by making

sample derivatization helped to isolate an omega fatty acids in

fish oil [4]. EPA and DHA from fish oil content varies from

species to species and season. Conversion of fatty acids to

ethyl esters followed by molecular distillation, urea

crystallization damages the omega 3 fatty acids. Re-

esterification of fatty acids leads to a disturbance in backbone

of glycerol [5].

To overcome this destruction and challenges, lipase is used

for concentrating omega fatty acids due to its important

properties [6-7].

2. MATERIALS AND METHODOLOGY

2.1 Determination of acid value:

The acid value of the sardine fish oil sample was determined

according to the standard procedure described in the American

Oil Chemists Society (AOCS). 25 mL of 1:1 toluene-isopropyl

alcohol and 2 mL of phenolphthalein indicator were added into

250 mL conical flask and neutralized with 0.1N potassium

hydroxide to a faint but permanent pink color. In another

conical flask, 0.75 g of oil sample was weighed and 25 mL of

neutralized solvent mixture were added to the sample and

mixed thoroughly. Titration was then carried out with 0.1N

potassium hydroxide to permanent pink color. The acid value

(mg KOH/g of sample) was given by the following formula

[8].

Acid value = (1)

A= mL of standard alkali used in the titrating the sample

B= mL of standard alkali used in the titrating the blank

N= normality of standard alkali

W= grams of sample

56.1 is molecular weight of KOH in grams.


9

2.2 Determination of p-Anisidine value

The p-Anisidine value of sardine fish oil sample was

determined using the standard protocol described in the AOCS

official method (AOCS). 0.7 g of oil was added into a 25 mL

volumetric flask.Volume was made up with diluted iso-octane.

The absorbance was measured at 350 nm with

spectrophotometer. 5:1 ratio of iso-octane and of p-anisidine

reagent were used as blank. 5:1 ratio of sardine oil and p

anisidine reagent was added to test tube and taken for UV

analysis after 10 minutes; the absorbance was measured at 350

nm. The p-anisidine value was given by the following

formula[8-9].

p-anisidine value = (2)

Where:

AS = absorbance of the oil solution: p-anisidine reagent

AB= absorbance of the oil solution

W= weight of sample

25= size of volumetric flask used

1.2= correction factor

2.3 Chemical hydrolysis of fish oil

The fatty acid of sardine fish oil composition has been

previously stated9.Among the fatty acids such as EPA (15.6%)

and DHA (10.7%), the hydrolysis of sardine fish oil was

performed in an organic and aqueous system. The protocol

was as follows: 2mL fish oil was added into 20mL of distilled

water along with 2g of NaOH and refluxed for around 8h at

1000C.

2.4 Thin layer chromatography

The various fatty acids are separated and fractionated by thin

layer chromatography. The lipids were fractioned using

hexane/ethyl acetate (75:25). The conditions for analysis of

fish oil were established by experimental selection of

appropriate stationary and mobile phases. Experiments were

performed on TLC silica gel aluminum plate. Good

separations were obtained in short time using mobile phase

Ethyl acetate (25%): Hexane (75%). TLC plate was spotted

with crude sample and esterified sample. Plates were then air

dried and analysed in UV chamber [4].

2.5 Preparation of UV absorbing derivatives

0.1-0.2 g of fish oil was transferred into round bottom flask. It

was dissolved in 2mL of acetonitrile and around 0.4 g of 4-

Nitrobenzyl bromide is added and exactly 0.2 g of potassium

carbonate was added in RB flask. The reaction mixture was

kept stirring at room temperature for 18 h. After refluxing the

sample was allowed to settle down and around 50 microliter of

sample was taken for analysis of TLC to check the state of

reaction. Remaining refluxed mixture is subjected to phase

separation using separating funnel and extracted with ethyl

acetate. Aqueous phase was separated and discarded and

further organic phase was concentrated using rotary evaporator

and subjected to HPLC analysis.

2.6 Characterization of fish oil by FTIR Spectroscopy

FTIR spectra of sardine fish oil were analyzed on a Perkin

Elmer RXI. The FTIR spectra were acquired from 3850-400

cm-1 with 4 cm-1 resolution by 20 scans with, monochromatic

infrared radiation as source, with LiTa 03 detector.

2.7 Concentration of n-3 fatty acids by enzymatic

hydrolysis

4g sardine fish oil, 6 mL of 1M phosphate buffer to activate

lipase enzyme and maintained pH of 7.5 and 1,150 Units (600

U/g) of lipase were transferred into a 50 mL conical flask. The

flask was transferred to water bath by maintaining temperature

at 35°C. The hydrolysis was started with the constant stirring

for 18 hours. The hydrolysis process was stopped by addition

of 2 mL methanol to the mixture. An amount of base was

added to neutralize the fatty acids obtained during hydrolysis.

The mixture was taken into a separating funnel and mixed

thoroughly with 50 mL hexane and 25 mL distilled water. The

upper layer containing ethyl esters was separated and was

washed twice with 50 mL distilled water. In order to remove

moisture content it is subjected with anhydrous sodium

sulphate [10] .After hexane removal at 45°C, ethyl esters were

recovered in a rotary evaporator.

2.8 Transesterification reaction

Transesterification was carried out in 50mL conical flask with

14mL of pure methanol and kept flask on the stirring plate

with constant speed to stir vigorously, and then slowly 0.50g

of NaOH added. Once NaOH is dissolved completely it forms

methoxide of sodium which serves as very strong and

dangerous base. Now place 30mL of fish oil which is warmed

at 65°C by placing it on stir plate under medium agitation, then

later add slowly the sodium methoxide solution to it which

forms a cloudy appearance and reaction mixture was kept for

around 20 minutes. Then transfer the above mixture into

separating funnel were biodiesel and glycerol was separated

based on their density [11].

3. RESULTS AND DISCUSSIONS

3.1 Quality of fish oil

Quality of fish oil was determined by standard AOAC method.

The obtained anisidine and acid value were in recommended

range.

3.2 FT-IR Spectrum

FTIR spectra of fish oil is shown in the figure 1, group

frequency ranging from 1750-1725 cm-1 clearly indicates esters

functional group ,by seeing figure 6 which clearly indicates the

presence of esters in the range 1745.84 cm-1.

3.3 EPA Standard

Figure 2, shows a typical Chromatogram of Eicosapentaenoic

Acid (EPA), a single fatty acid was found to have 92.63 area

percent of fatty acid DHA Standard.


10

Figure 3 shows a typical Chromatogram of Docosahexaenoic Acid (DHA), a single fatty acid was found

to have 85.73 area percent of fatty acids.

3.4 Chemical method of separation

A chromatogram for chemically hydrolyzed fish oil is shown

in the figure 4. Here a single fatty acid was found to have 0.4

area percent of the omega 3 fatty acids.

Figure 1. FTIR spectra (4000-400cm-1) of fish oil

Figure 2. Chromatogram of EPA standard

Table 1: Recommended quality parameters of fish oil and

Experimental values

Quality

Parameter

Recommended

Value

Experimental

value References

Acid Value

7-8 mg of

KOH/g of

sample

7.48mg of

KOH/g of

sample

12

p-anisidine

value ≤20 19.575 13

3.5 Enzymatic method of separation

A chromatogram for enzymatic hydrolyzed fish oil is shown in

the figure 5. Here a single fatty acid was found to have 2.0826

area percent of the omega 3 fatty acids.

3.6 Biodiesel yield and its properties

Transesterification of fish oil resulted in the reduced kinematic

viscosity and density was within permissible limit which plays

an important role in fuel atomization. The flash point, fire

point were in limit of safe storage and handling conditions.

Figure 3. Chromatogram of DHA standard

Figure 4. Chromatogram of chemical method of separation by

HPLC


11

Figure 5. Chromatogram of enzymatic method of separation by

HPLC

Table 2: Fuel properties of sardine fish oil ethyl esters

Parameters Obtained

value

Expected

value

viscosity at 40°C 3.41 20

Flash point(°C) 170 200

Fire point(°C) 180 215

Density kg/m3 830 920

4. CONCLUSION

Quality of fish oil was carried out by standard AOAC method.

The recommended range for p anisdine value of crude fish oil

is 7-8 mg of KOH /g. The obtained p anisidine value were in

the range of 7.48 mg of KOH /g of the sample. Anisidine value

is an empirical test which determines the advanced oxidative

rancidity of oils and fats. The obtained acid values were in the

recommended range, the acid values quantifies the amount of

acid present, TLC analysis helped in quantitative identification

of fatty acids present in fish oil. FTIR spectroscopy of the

sample clearly indicated that frequency ranging from 1745.84

cm-1 clearly indicated esters functional group.

Using various Analytical techniques, it is possible to isolate

different omega free fatty acids. Samples were hydrolyzed to

separate the fatty acids from their glycerol backbone and

analyzed directly using HPLC with UV detector. Chemical

method of separation shown that 0.4 area percent of omega 3

fatty acids were as enzymatic method of separation 2.0826

area percent of omega 3 fatty acids. Transesterification of fish

oil resulted in reduced kinematic viscosity and density within

permissible limit which meets and well within the biodiesel

characteristics as compared.

ACKNOWLEDGMENT

Authors are thankful to the Managements of

BMS College of Engineering, Bengaluru and Sahyadri

College of Engineering and Management for providing

infrastructure facilities to carry out their research work.

All India Council of Technical Education for their

financial support under Research Promotion Scheme.

M/s Mangaluru Marine Products.

REFERENCES

[1] Akoh, C.C., S. Sellappan, L.B. Fomuso and V.V. Yankah.

2002. Enzymatic synthesis of structured lipids. Lipid

Biotechnology. New York, NY: Marcel Dekker, Inc., 433-

460.

[2] David Cowan Agro food industry hi tech Aug 2010

.Concentration of omega 3 fatty acids using enzyme. Vol

21 n4.

[3] Ian Acworth,Marc Plante et al,Quantitation of

underivatized omega3 and omega 6 fatty acids in foods by

HPLC and charged aerosol detection,Thermo fisher

Scientific ,Chelmsford,MA,USA.

[4] Durst HD et al Anal Chem 1975, Preparation of UV

absorbing derivatives, 47, 1797.

[5] Klinkeson U, Aran H,et al .Chemical transesterification of

tuna oil to enrich omega 3 polyunsaturated fatty

acids.2004:87(3):415-21.

[6] Kralovec JA ,Wang W, Barrow CJ,Production of omega 3

triacylglycerol concentrates using a new food grade

immobilized Candida lipase 2010;63(6):922-8.

[7] Kosungi Y,Azuma Synthesis of triglycerol from

polyunsaturated fatty acid by immobilized

lipase,AOCS1994; 1397-403.

[8] P.Arul franco et al 2014 .Performance and emission study

of sardine fish oil Biodiesel in a diesel engine: oxidation

communications 37, no.3, 802-816.

[9] O’Brien, R.D. 2009. Fats and oils. Formulating and

processing for applications, 3rd ed., CRC press, London,

213-300.

[10] Wanasundara, U.N. and F. Shahidi. 1999. Concentration

of omega 3-polyunsaturated fatty acids of seal blubber oil

by urea complexation: optimization of reaction conditions.

Food Chemistry, 65, 41-49.

[11] Meher, L.C., D. Vidya Sagar and S.N. Naik. 2006.

Technical aspects of biodiesel production by

transesterification: a review. Renewable and sustainable

energy reviews, 10(3), 248-268.



12

Rajmunni Hombal1, Shwetha L G1, Pooja K1, Rathishchandra.R.Gatti2*

1 Department of Civil Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007 2 Department of Mechanical Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007

*Email:[email protected]

ABSTRACT

Plastic can be said as a waste, when it is not properly managed and hence imposes the negative environmental effects. All

types of non-bio-degradable and unused plastic waste, when cannot be recycled is sent to landfills. Landfills are becoming

a big environmental issue and hence expensive with lot of restrictive procedures, forcing the companies to look for

alternatives to dispose or reuse plastics. At the same time, reduced availability of conventional binding materials such as

clay tested in terms quantity and quality pose a frantic threat for the builders. The proposed research is an experimental

development and validation of the use of PET plastics to act as filler materials for brick manufacturing. Samples of bricks

were manufactured for studying few design parameters and were tested for the common brick evaluation standards as

prescribed by BIS proving that PET plastics can be used as filler materials.

Keywords: Composite, Environment, Plastics, Recycle, conventional

1. INTRODUCTION

Plastic waste involves the accumulation of plastic products in

the environment that adversely affects wildlife, habitat or

humans [1]. But plastic is a relatively cheap, durable and

versatile material and its products have brought benefits to

society in terms of economics and quality of life [2]. Because

of its demand and use, plastic waste generation has also

continued to grow. However, at the global scenario, although

its production and utility is being met according to the demand,

the proper disposal of plastics is not addressed satisfactorily.

Most of the plastics that are carelessly disposed to the

environment get directly or indirectly consumed by the animals

and has thus entered the food chain [3]. Since these plastics are

non-biodegradable and hence not digestible, they can block the

intestines leading to health hazards to animal life. Not many

plastics can be recycled [4]. Also, recycling not only incurs

costs but also emits harmful toxins to the air [5]. The other

option is to refill the plastics. However, plastics have become a

menace in the landfill areas making it expensive even for

landfill [6].

Generally, bricks are made using top soil from agricultural

fields and quarries, approximately half an acre land about

2000m2X 0.05m top soil is required for the making of about

1,00,000 bricks Essentially, bricks are produced by mixing

ground clay with, forming the clay into the desired shape, and

drying and heating. The manufacturing process has seven

general phases which include mining and storage of raw

materials, preparing raw materials, preparing different grain

sized particle, forming the brick, drying, heating and cooling,

de-hacking and storing finished products.

The objective of this research was to develop a solution for

the use of plastic waste in composite brick manufacturing. This

is achieved by experimenting samples of bricks made of

different grain sizes of Polyethylene Terephthalate (PET)

plastic waste that replace the weight of natural soil in order to

achieve the strength of bricks. The optimum mixing proportion

is then determined for the maximum compressive strength of

the brick. The typical tests that are done for bricks either in the

lab or in the field are commonly compressive strength test,

water absorption test, efflorescence test, above one metre

impact drop test, ringing sound test and hardness test. These

tests were performed according to the BIS standards for the

proposed composite bricks and their properties were studied.

2. EXPERIMENTAL SAMPLE

PREPERATION

The sample composite bricks were studied for the varied sizes

of the PET as shown in the figure 1 below. After once finished

with preparing different grain sized plastic material we move

on to the proportionality of mixing this plastic grain in the

manufacturing of brick.

Useful Application of Plastic Waste in

Composite Brick Manufacturing


13

Figure1: Different grain sizes of PET considered in the composite brick

samples.

The materials used to manufacture bricks are clay, fly ash,

plastic grains, and water. This all material mixed in the

proportion and mould in the still mould. After that left to dry

for one day. The dried bricks are sent to Kundapur brick factory

for heating in the kiln.

Figure 2: Process of composite brick sample preparation

3. COMPRESSIVE STRENGTH TEST 1 - FOR

OPTIMAL GRAIN SIZE

The four samples that were prepared as discussed in section two

were tested for compressive strength in the Double column

Universal testing machine as shown in the figure 3. From the

test, it was found that the brick sample of 4 mm plastic grain

size had the highest compressive strength. Thus this grain size

was considered for the next compressive strength test.

Figure 3: Testing of the composite bricks in Universal Testing machine

4. COMPRESSIVE STRENGTH TEST 2 - FOR

OPTIMAL PLASTIC PROPORTION

Five new samples of plastic composite bricks were

manufactured similar to the procedure mentioned in the section

2, but for the same grain size of 4 mm. However, in this

experiment, the proportion of the PET plastic to the coarse

aggregate percentage weight was kept as 5%, 10%, 15%, 20%

and 25% for the five samples respectively. It was found that the

sample 4 consisting of 20% PET percentage by weight had the

maximum compressive strength of 7.2 N/mm2.

5. WATER ABSORPTION AND OTHER

BRICK TESTS

The water absorption test was conducted on a new sample

manufactured as per section 2 with the 4 mm PET plastic grain

size in the proportion of 15% by weight to the weight of the full

brick. The initial weight of the brick was 2.46 kg and the final

weight of the brick was 2.67 kg. The water absorption

percentage was calculated to be 8.53%.

The same brick was later dried and tested for efflorescence.

This was done by dipping the brick in water for 24 hours and

then removed to dry in the atmosphere. Approximately, around

2546 mm2 = 3% of the brick’s surface was covered by white

patches which are highly acceptable as per the standards.

Impact drop test was performed by dropping the sample brick

at a height slightly above 1.5 m high. As the plastic grains are

well bonded with soil, the sample brick considered was able to

clear the impact drop test.

6. RESULTS & DISCUSSIONS

The results of the compressive strength 1 test for varied grain

sizes are shown in the figure 4. From the compression test result

it is clearly shows that the value of compressive strength is

maximum for the plastic grain size with 0.5 to 0.75cm followed

by 20% replacement of clay by plastic and it showed in the

maximum load on compression is 97kN.

According to Bureau of Indian Standards 1077:1992

commonly burnt clay building bricks average compressive

strength for first class brick should not less than 10N/mm2 or

100kgf/cm2.

Our experiment results justifies that this bricks are reaching

the BIS limits. Therefore we can use this bricks and replace the

first class bricks since it is economical and ecofriendly product.

The results of the compressive strength test 2 for varied plastic

proportion by weight is as shown in the figure 5.

The compressive strength of the sample bricks of proportions

of 5%, 10%, 15%, 20% and 25% plastic waste were 5.15, 5.35,

5.68, 5.25 and 5.2 N/mm2 respectively. The maximum load was

recorded as 72 kN, 75 kN, 82 kN, 97 kN and 93 kN respectively.

Firing

of

bricks


14

Figure 4: Experimentally observed effect of grain size of the PET

in the compressive strength of the composite brick.

Figure 5: Experimentally observed effect of plastic proportion by weight

of the PET in the compressive strength of the composite brick.

The water absorption percentage was about 8.53% which is

significantly less than the 20% set by the Indian Standards

1077:1992. The efflorescence test confirmed 3% of the surface

to have alkaline deposition not affecting much of the brick

structure.

7. CONCLUSION

The percentage of different grain sizes of plastic waste is

replaced by the weight of natural soil in order to achieve the

strength of bricks. This results in reducing the harmful effects

of the waste plastics in the environment. We also developed

a solution for reduction of the disposal of plastic waste by

replacing 20% plastic waste in order have maximum load at

crushing of 97 kN.

ACKNOWLEDGMENT

We are thankful to Mr Vaishak N L, Assistant Professor, and

Mr. Sudeep Shetty for extending their support to do this project.

We would like to extend our gratitude to Sahyadri project

support scheme -SPSS, an Undergraduate project grant for

doing this project.

REFERENCES

[1] Jambeck, J.R., et al., Plastic waste inputs from land into the

ocean. Science, Vol. 347, No. 6223, pp. 768_771, 2015.

[2] Andrady, A.L. and M.A. Neal, Applications and societal

benefits of plastics. Philosophical Transactions of the

Royal Society of London B: Biological Sciences, Vol.364,

no.1526, p p. 1977_1984, 2009.

[3] Rochman, C.M., et al., Ingested plastic transfer’s

hazardous chemicals to fish and induces hepatic stress.

Scientific reports, Vol.3, pp. 3263, 2013.

[4] Hopewell, J., R. Dvorak, and E. Kosior, Plastics recycling:

challenges and opportunities. Philosophical Transactions

of the Royal Society of London B: Biological Sciences,

Vol.364, no.1526, pp. 2115_2126, 2009.

[5] Tsai, C.J., et al., The pollution characteristics of odor,

volatile organo chlorinated compounds and polycyclic

aromatic hydrocarbons emitted from plastic waste

recycling plants. Chemosphere, Vol.74, no.8, pp.

1104_1110, 2009.

[6] Ishigaki, T., et al., The degradability of biodegradable

plastics in aerobic and anaerobic waste landfill model

reactors, Chemosphere, Vol.54, no.3, pp.225_233, 2004.

Grain size (mm)

Co

mp

ress

ive

stre

ng

th (

N/m

m2)

Percentage by weight of PET

Co

mp

ress

ive

stre

ngth

, N

/mm

2



15

Mohammed Mustafa*, Ronald Anthony D’souza, Deepthi H., Jeevan T. B., and Dr. Pushpalatha K

Department of Computer Science and Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007


ABSTRACT

An aptitude test is a systematic means of testing a candidate's abilities to perform specific tasks and react to a range of

different situations. Quantitative aptitude problems are difficult to analyze and formulate without having a clear

understanding. Basic concepts must be known thoroughly to solve the problem efficiently. Many people lack the

knowledge of solving the problem using the basic concepts. When solving aptitude tests for companies, candidates should

have the ability to solve the problem within a limited time. Hence, we have introduced Aptitude Question Solver (AQS)

which provides a step-by-step procedure for each quantitative problem.

Keywords: Mathematical word problem, natural language processing, aptitude questions.

1. INTRODUCTION

With the advent of computers, all aspects of society have been

influenced by it, including education. Computers are used at

all levels of education. But with recent advancements in

technology, even children are learning with computers.

Wikipedia, Freebase, YAGO, Microsoft, Satori and Google

Knowledge Graph are some of the well-known knowledge

bases. Information present in them could be used to build

specific decision making advisory systems. Question

Answering systems, which are a part of advisory systems are

viewed as futuristic replacement of call centers and are called

as virtual assistants [9]. An aptitude test is a systematic means

of testing a candidate's abilities to perform specific tasks and

react to a range of different situations. Quantitative problems

are a part of Aptitude tests. It involves a wide range of

mathematical problems. A mathematical problem is a problem

which can be controlled, analyzed and solved using methods

that contain sequence of formulae, theorems, postulates, and

axioms. The duration for solving each quantitative problem is

limited. Various studies show that students often face

problems while solving mathematical word problems like

generating variables and forming equations without any basic

knowledge and techniques to solve [1]. Sometimes, the correct

calculations will result in incorrect answers due to incorrect

problem representation [10].

Existing application such as Wolfram Alpha [2] requires input

in terms of equations or simple math word problems. It fails to

provide solution as the complexity of the mathematical word

problems increase. So, if the users are not able the form the

equations from the given word problem, he/she will get

completely stuck. Other applications are MathWay [3] and

WebMath [4] that helps to solve mathematical problems, but it

cannot process these problems when given in English

language. And it also requires the users to choose what

operation to be performed.

Keeping in mind these issues in existing systems we have

proposed AptitudeQS for solving these problems. AptitudeQS

can be of great use to understand the solution and to know the

techniques to solve these problems in a quick and efficient

way. The main purpose of our system is to provide stepwise

approach to solve any given aptitude question. Our system can

be used by any person who intends to learn and solve and can

also be used by students or candidates to improve their

aptitude solving ability. Users, who intend to use this system,

will need to enter the word problem question. The system will

interpret the question entered by the user and generate an

appropriate solution. The solution will be represented in a step

by step manner, which will help the student to understand the

solution and concept behind it.

This paper is structured as follows. Section III describes the

system architecture and the various components involved in it.

Section IV deals with the generation of training dataset and the

development of neural network. The process of analyzing the

query from user is explained in section V. Section VI

illustrates the system performance and its evaluation. Finally,

we conclude about AptitudeQS in Section VI.

2. RELATED WORKS

In recent years, technologies such as Wolfram Alpha [2],

WebMath [3] and MathWay [4] were developed which can

solve verbal mathematical problem only if the question is

simple. Wolfram Alpha is able to solve simple verbal

mathematical problems but fails to do so when the complexity

Aptitude Question Solver: AptitudeQS


16

of it increases. WebMath places the overhead of extracting

numerical data form the verbal statement upon the user.

MathWay is an interactive chat bot application which tries to

solve mathematical word problems but the problem here is that

we must choose the operation to be perform. Therefore, to

overcome all these drawbacks, we propose a new system

called Aptitude Question Solver. AptitudeQS can solve

complex verbal mathematical problems. The main goal of our

system is to provide the detailed procedure to solve any given

aptitude question.

3. SYSTEM OVERVIEW

Figure 1: System architecture

The system architecture of the AptitudeQS is shown in

figure 1. The components present in this system are dialogs,

pre-processor, vectorizer, neural network, entity recognizer

and AptitudeQS.

Initially, the system is trained before it can be used to solve

the questions. The dialogs component defines a set of

expressions and the corresponding methods to find the

solution. An expression is a template or a pattern that

describes a question, which can also include one or more

entities. An entity is a parsed element found within the user’s

query. Each expression has an intent, which is a method, to

solve the question associated with it. Using these expressions,

the system generates a training dataset that consists of

questions and the name of the expression to which it belongs

to. The system uses a neural network to classify the questions

to the corresponding expression and uses the generated

training set for training. Before feeding questions to the neural

network, they preprocessed using Natural Language

Processing (NLP) to understand the meaning and to remove

unwanted information from it. Finally, the output from NLP is

converted into feature vectors. These feature vectors are given

as input to the neural network and a similarity score associated

to each expression is produced as output. The expression with

the maximum score is considered as the classification result.

The AptitudeQS component is the central unit of our

system. This component is responsible for starting the training

process. It accepts the question from the user and consults the

neural network to identify the expression to which it belongs

to. Once the expression is known, the entity recognizer

matches the question with the expression to extract the entities

from it. The intent associated with the expression is then

invoked to compute the solution.

4. TRAINING NEURAL NETWORK

An Artificial Neural Network (ANN) is a computing system

inspired by the biological neural network present in animal

brains [5]. The main idea here is to build an ANN model to

classify the user’s query to a matching expression. Once the

classification of the query is done and a matching expression is

identified, the corresponding intent is invoked to compute the

solution.

A prerequisite to understand the input given by the user is to

train the system with predefined datasets. The steps involved

in training phase are generating training dataset, pre-

processing, vectorizing and training of neural network. The

system begins by generating a set of sample questions from the

expressions in the Dialogs component to form a training

dataset. Since the neural network accepts only fixed sized

inputs containing numerical data, the sample questions, before

feeding for training, need to be converted into a form that the

neural network can understand. This task is done by

Preprocessor and Vectorizer components. After this, the neural

network is trained using the converted training dataset and can

be used for prediction, to find out to which expression in the

Dialogs a given query matches.

4.1 Generating training dataset

A dialog is a collection of expressions with corresponding

intents. An expression is a pattern that defines the user input,

while an intent is a void function with expression and entity

attributes. The proposed system has a set of dialogs like “profit

and loss”, “average”, “clocks” and “ages” which are used to

train the neural network. An example of an expression is as

follows:

“Find the average of prime numbers between

@number.integer:num1 and @number.integer:num2”

where, @number.integer is an entity that is used for parsing

integer values present in the query, and num1 and num2 are

the alias names given to the two entities. num1 refers to the

first integer number in the query, while num2 refers to the

second integer value.

To train the neural network, a set of sample data is

required. For this, we generate a set of sample queries for each

of the expressions present in the system. The sample queries

are generated by replacing any entities present in the

expression with the values that they define. For example, using

the above expression, a sample query can be generated by

replacing the entities @number.integer:num1 and

@number.integer:num2 by random integers such as 10 and 20.

The so formed sample query after replacing will be


17

“Find the average of prime numbers between 10 and 20”

In this way, at least five sample questions are generated for

each expression. Let T indicate the training dataset. For each

of these sample questions, a dictionary is created with a key

named expression to store the expression and another key

named query to store the sample query. The expression part

behaves as label while the query part behaves as data. The

dictionary is then appended to the training dataset T.

Table 1: Variable definition.

Symbols Description

T Labeled training dataset

W List of words in each data

E List of expressions

IW List of words or symbols that needs to be

ignored

D Document in our corpus

training Training dataset after vectorization

output List of output labels

output_empty Initial output label filled with zeros

X Input dataset to neural network

Y Output dataset matrix of neural network

hno Number of hidden neurons present in neural

network

s0 Synaptic weights of links between input layer

and hidden layer

s1 Synaptic weights of links between hidden

layer and the output layer

psu0 Updates of previous synaptic weights between

input layer and the hidden layer

psu1 Updates of previous synaptic weights between

hidden layer and output layer

lmerror Last mean error

sdcount0 Direction count of s0

sdcount1 Direction count of s1

E Epoch – number of iterations of training

l1error Hidden layer error

l2error Output layer error

l1delta Layer 1 error rate

l2delta Layer 2 error rate

s1wu Synaptic 1 weight update

s0wu Synaptic 0 weight update

QW List of words present in a query

Algorithm 1 describes the process of generating documents,

expressions and words from the training dataset. All variables

are shown in Table 1 along with their description.

The algorithm begins by looping through each query in the

training data T. For each query, the algorithm tokenizes it,

using the word_tokenize method of NLTK Python package,

into a list of words. These words are appended to the word list

W. A tuple consisting tokenized words and the expression to

which the query belongs is created and then appended to the

documents list D. The word list W is then refined by removing

the unwanted words or symbols from it which are present in

the ignored word list IW (from line 6 to line 13). In line 14,

each word in W is transformed to lowercase letters. Finally, the

algorithm converts the lists W and E, each consisting of unique

words and expressions respectively. The algorithm ends by

returning W, E, D.

4.2 Transforming training data into bag of words

The next step after generating training data and organizing

data structures for words, expressions and documents is to

transform the training data into bag of words. Text Analysis is

a significant field for machine learning algorithm. The raw

data, sequence of symbols cannot be directly fed to the neural

network because most of them expect numerical feature

vectors with fixed size. To address this problem, we convert

text into fixed-length vectors of numbers using Bag-of-Words

(BoW) model [6]. This model focuses on the occurrence of

words in the document and does not keep track of their order.

In this model, each word is assigned a unique number. The

document is encoded as a fixed-length vector with the length

of the vocabulary of known words. The value in each position

in the vector is filled with a count or frequency of each word in

the encoded document.


18

Algorithm 2 describes the process of preparing the training

data in terms of vectors and the output label in terms of

numbers. In this algorithm, each of the queries present in

document is converted into a fixed-length vector using the

BoW model. It first gets all the tokenized words of the query.

Then the algorithm iterates through the word list W. If a word

in W exists in the query, a 1 is appended for that particular

word in the bag list, otherwise 0 is appended. Finally, an

output label is created by adding 0 at all the indices except for

the index belonging to the expression that the query matches.

The algorithm returns the training and the output lists.

4.3 Training the neural network

The classification of a query entered by the user to identify

expression to which the query matches the most can be

attained using an Artificial Neural Network (ANN). Neural

Network is an information processing model that process

information. The main component is the novel structure of

information processing system. Information processing system

takes information in one form and processes it into another

form. Neural network is organized in layers. Layers are made

up of interconnected ‘nodes’ that includes activation function

which defines the output of the node for the given set of

inputs.

An activation function in neural networks is used to

determine its output. It maps the resulting values in between 0

to 1. AptitudeQS uses the Sigmoid function as activation

function.

Figure 2: Sigmoid function

The Sigmoid function curve looks like an S-shape, as

shown in figure 2. The main reason to use Sigmoid function is

because the curve always exists between 0 and 1. Therefore, it

is used by our neural network to predict the probability as an

output. This system uses Sigmoid function to normalize values

and its derivative to measure the error rate. The value Sigmoid

function can be calculated using the following equation:

(1)


19

For error propagation, it is necessary to find the derivative

of the output of the sigmoid function. This derivative can be

calculated using the following equation:

(2)

Now that the algorithms to calculate the sigmoid function and

its derivative are defined, it is time to develop a training

function to create synaptic weights of the neural network.

Algorithm 3 describes this process of training the neural

network.

Finally, the algorithm returns the synaptic weight between

layer 0 and layer 1, s0, and the synaptic weight between layer

1 and layer 2, s1. These weights are converted into list and

stored in the form of JSON in a file named “synapses.json”.

5. PROCESSING QUERY

This section describes about how the user’s query is processed

to compute the solution. The query asked by the user is first

accepted by the dialog controller. If this is the first time an

instance of dialog controller is created, it calls Algorithm 3 for

training.

5.1 Pre-Processing

Intention of the query needs to be understood to process it. [7].

This can be done by using Natural Language Processing. The

next step is to pre-process the query entered by the user. The

main task of pre-processing is tokenizing, stemming and

transforming all the letters into lower case. The NLP process is

done by using Natural Language Toolkit (NLTK) Python

package [8].

o Tokenization – The process of breaking up the given

text into units called tokens. The tokens may be in the form of words, numbers or punctuation symbols. The main aim of tokenization is to explore the words in the given query and to give an integer id to every token. The list of tokens becomes input for vectorization.

o Stemming – The process of reducing words to their root form. The main goal of stemming is to reduce inflectional forms and convert it to the base form. English Stemmer is used for reducing the words to their root form.

Algorithm 4 is used to pre-process a query given by the

user. It begins by breaking the query statement into a list of

words called tokens (line 2). Each of these words are

transformed into lower case and stemmed down into their root

form (line 3). The algorithm then returns a list of preprocessed

words.

5.2 Vectorizer

The vectorizer component takes in the user’s query as input

and converts it into a fixed size vector using BoW model. The

following algorithm explains the process of vectorizing.

In algorithm 5, a list named bag of length similar to that of

words list, W, is created and filled with 0s. It calls Algorithm 6

with the query to obtain a list of pre-processed words, QW.

The algorithm then iterates through W and checks if each word

is in QW. If that word is present in QW, a 1 is added to the bag

at the position corresponding to that word. In this way, the

algorithm returns a vector that represents the user’s query.

5.3 Neural Network for prediction

Algorithm 6 creates an initial neural network model. The

neural network used in this system has three layers: an input

layer that accepts vector inputs, a hidden layer that processes

the information and an output layer that produces the output.


20

Now that the neural network model is developed and trained

using the sample training data generated earlier, it is now

possible to classify the query entered by the user to which

expression it belongs to. Algorithm 7 describes this

classification process. In this algorithm, we maintain a

minimum error threshold of 0.7. After prediction, the output of

the neural network with the highest score is considered as the

prediction result only if the score is greater than the threshold.

This algorithm finally returns the classified expression.

5.4 Entity Recognizer

Before calling the Invoke method of AptitudeQS component, it

is required to extract the entities, if any, from the query. The

task of extracting entities is done by Entity recognizer. This

component first identifies the entity types present in the

expression. Based on these entity type, it compares the query

with the expression to locate the values of each of these entity

types. Entity recognizer then returns a list of tuples containing

the entity types along with their values.

6. RESULTS AND PERFORMANCE

EVALUATION

We have implemented our system using python 3.6.0

programming language on a system running Windows 10

Operating System. For evaluating the performance of the

system, testing, was carried out on AptitudeQS. Table 2 shows

the test cases for AptitudeQS and its results for the given

input.

Table 2: Test Cases.

Question

Find the average of first 10 prime

numbers.

Expression Find the average of first

@number.integer:num2 prime

numbers

Expected Output 4.25

Output 4.25

Result Success

Question

Find the average of all prime

numbers between 30 and 50.

Expression Find the average of all {prime}

numbers between

@number.integer:num1 and

@number.integer:num2


Output 39.8

Result Success

Question

Find the average of first 40 natural

numbers.


@number.integer:count natural

numbers.


Output 20.5

Result Success

Question

Find the average of first 20

multiples of 7


@number.integer:count multiples

of @number.integer:num


Output 73.5

Result Success

Test ID 5

Question

The average of four consecutive

even numbers is 27. Find the

largest of these numbers.

Expression The average of @number:count

consecutive @evenOrOdd

numbers is

@number.integer:average. Find the

@largestOrSmallest of these

numbers.

Expected Output 30

Output 30

Result Success

The goal of the evaluation is to find the accuracy of the

system. Based on the test cases executed using few sample


21

questions, it is estimated that the system provides a

classification accuracy of 98%.

7. CONCLUSION

In this paper, we have proposed Aptitude Question Solver that

solves mathematical word problems. AptitudeQS accepts

query from the user and provides detailed solution. This helps

the users to learn how to solve aptitude questions and

improves their skills in solving. Currently the AptitudeQS

system solves four classes of questions, which are average,

profit and loss, ages and clocks. In future, the system can be

upgraded to solve other categories of questions. The system

can also be enhanced for solving aptitude questions by

accepting speech as an input.

REFERENCES

[1] D. Cummins et al., “The role of understanding in solving

word problems”, Cognitive Psychology, vol. 20, pp. 405-

438, 1988.

[2] Wolfram Alpha (2009). [Online]. Available:

http://www.wolframalpha.com/

[3] MathWay, online available at http://www.mathway.com as

on 18-05-2018

[4] WebMath, online available at http://www.mathway.com as

on 18-05-2018

[5] "Artificial Neural Networks as Models of Neural

Information Processing | Frontiers Research Topic".

Retrieved 2018-02-20.

[6] McTear, Michael (et al) (2016). The Conversational

Interface. Springer International Publishing.

[7] Sarkar et al., “NLP Algorithm Based Question and

Answering System”,

[8] Natural Language Toolkit. (2001). [Online]. Available:

http://www.nltk.org/

[9] Dong. X. L. Murphy.K.Gabrilovich .E. Heitz.G.Horn.

W.Lao.N. & Zhang.W. (2014). Knowlegde Vault. A web-

scale approach to probabilistic knowledgr fusion.

[10] R. Schumacher and L. Fuchs, ―Does understanding

relational terminology mediate effects of intervention on

compare word problems?,‖ Journal of Experimental Child

Psychology, vol. 111, pp. 607-628, 2012.



22

An Integrated Approach for Personality Analysis

using OCR and Text Mining

Pavithree B. Shetty, Sanath R. Kashyap, Sneha V. Kamath, Supraja,

Bharath Bhushan S. N*.

Department of Computer Science and Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007 *Email: [email protected]

ABSTRACT

In this paper we aim at recognizing handwritten text from an image and convert that into an editable document and

perform personality analysis. It is a challenging task because each individual's handwriting is unique. As a person’s

mind-set plays a major role when he pens down the content, in our work we use the content written by the person to

classify contextually if the person is positive minded or negative minded. In order to convert handwritten characters into

an editable document we have used optical character recognition (OCR) and machine learning techniques. Personality

analysis is done based on the frequency of occurrences of words.

Keywords: Optical Character Recognition, Text Mining, Machine Learning

1. INTRODUCTION

From past few years there is a lot of data which is being

generated which demands automated management without

much human intervention. This paper mainly focuses on

offline handwritten character recognition of English words by

initially identifying individual characters. Today due to the

advent of internet and its wider reach to public there are

millions of text data which is being generated every day and

this requires data management.

Optical character recognition (OCR) is the electronic or

mechanical conversion of handwritten, typed or printed text

images into machine-encoded text. This involves scanning of

the text character by character and then the character image is

translated into character codes, which is frequently used in

data processing.

Given a scanned image of handwritten text, we aim to

extract the text in that image using OCR algorithm and display

it in a editable document along with identifying if the content

has been written by a positive minded or negative minded

person by identifying and understanding what has been

written. Using text mining techniques we achieve the later part

of the problem statement.

In the field of business, management, education, various

billing systems and various ticket reservation systems lot of

data is being generated but this does not provide us any useful

conclusions. When this data is analysed we get valuable

information from which we can get the current interests of

people and improve the present business trends.

Text mining techniques can be used to derive useful insights

from the wide range of data. Text mining is analysing the data

which is contained in natural language text. From editable

document using machine learning and text mining techniques

we infer if a person is contextually positive minded or negative

minded.

The aim of the text classification algorithm is to determine

if the person is positive minded or negative minded based on

the content given by the user. Suppose we have two classes of

documents i.e., class1->content given by positive thinkers and

class 2- content given by negative thinkers. We have to train

our model with these documents. The text classification

assigns a Boolean value to each pair Φ (Qd,S)=1 where, S is

a set of predefined categories and Qd is the domain of

documents. The task is to approximate the true function Φ=

(Qd,S)->{0,1} ( where its 1 if classifier classifies the

document properly as positive minded or negative minded else

it is 0) using a function^

^ Φ= (Qd,S)->{0,1} such that the values of Φ and Φ has

approximately similar values.

Following section of the paper is organized as follows. Section

2 gives a brief literature survey on the text classification and

representation. Section 3 represents the proposed model for

conversion of handwritten characters to editable format and in

the later part we use text classifier to classify if a person is

positive minded or negative minded using text mining

approaches. Experimentation and comparative analysis

performed on the proposed models will be discussed in the

section 4. Finally we conclude the paper in section 5.


23

2. LITERATURE SURVEY

In [1], authors highlight the main techniques and methods used

in text document classification. It emphasises the

representation of text and machine learning techniques. The

methods and theories of text mining and document

classification is analysed in the paper. [2] This research article

contains a B-Tree based classification methodology which is

adapted for classification. The proposed compressed

representation and B-Tree methodologies are verified on the

publicly available large corpus to validate the effectiveness of

the proposed models. [3] In this paper, a learning model of text

classification for support vector machine (SVM) is evolved. It

creates a bridge between the characteristics of text

classification task and the generalisation performance of a

SVM in a quantifiable manner. [4] Major problem such as

handling large number of attributes, dealing with the

unstructured text, and choosing a machine learning technique

applicable to the text-classification application. [5] To increase

the performance of the Centroid classifier, a novel batch-

updated method is proposed in this paper. The aim of this

approach is to successively update the classification model by

batch, by taking advantage of training errors. [6] This paper

explores a new technique of feature selection metrics using

less number of keywords which is highly successful. [7] To

deal with multi-label classification problems, this paper

proposes the Probabilistic Neural Network (PNN) algorithm,

and is compared with Ml-kNN algorithm. This application

divides the MI-kNN algorithm into four parts which is used for

multi-label categorization problems. [8] This paper describes a

natural language processing system reinforced by the use of

association of words and concepts, implemented as a neural

network. Combining an associative network with a

conventional system contributes to semantic disambiguation in

the process of interpretation.[9] In this paper, a new text

document classifier is implemented using the support vector

machine (SVM) training algorithm and the K-nearest

neighbor(KNN) classification approach combined together.

The Support Vector Machine - Nearest Neighbor classification

approach is named as SVM-NN. [10] In this paper, it takes the

advantage of both longest common subsequence (LCS) and

VSM algorithm and proposes integrated text retrieval (ITR)

mechanism. LCS is used to evaluate the weight of terms and is

the main idea of the ITR mechanism, so that the weight

relationships and the sequence between the texts and the query

can be examined concurrently. [11] This paper measures the

virtual generalizing random access memory weightless neural

networks (VG-RAM WNN), which is an efficient method for

machine learning technique which is very simple to implement

and faster in training and testing. To build automatic multi-

label text categorization systems, VG-RAM WNN is used as a

tool. The performance of the VG-RAM WNN is evaluated on

two issues: 1) classification of the text characterization of

commercial activities, 2) Web page classification. [12] This

paper proposes an algorithm which will learn from the data set

provided to perform speech recognition task and multiclass

text task. This method is based on fresh and enhanced family

for boosting the algorithms. Boos Texter, which is the new

algorithm for boosting the performance, is used for text

categorization task. [13] This paper presents a mathematical

model of classification schemes and the one scheme which can

be proved optimal among all those based on word frequencies.

[14] This paper represents a method DP4FC which is used to

choose appropriate feature to categorize and differentiate the

appropriate documents from the inappropriate documents.

DP4FC is combined with the other classifiers. After getting the

appropriate document, the classifier creates the effective

category groups and takes appropriate decisions in classifying

and filtering. [15] In this paper, for dimension reduction, the

phonological different words, grammatical words, and the

stopwords are recognized and eliminated. There are two

algorithms for dimension reduction. They are frequent term

generation and improved stemming algorithms. [16] This

paper explains the flow of the processing of the information

and for text categorization. There are two efficient learning

algorithms. They are Partial Least Squares (PLS) and Support

Vector Machines (SVM) and is applied in other domain as

well. [17] In this paper, the authors explain about the steps.

They are rule generation, calculation of probability and pre-

processing. The training set document is read in the rule

generation. Negative and positive weights are calculated in the

calculation of probability. The document which is given as

input is divided into statements and paragraphs in pre-

processing. [18] This paper makes use of statistical term

clustering and syntactic processing to represent a document

which is more accurate than obtained by using traditional

keyword methods. [19] This paper compares the success rate

of automatic learning algorithms by means of speed in

learning, accuracy and speed in real time calculation for

categorization of text. This paper also checks the size of

training set and other representations of document. [20] This

paper proposes transfer of knowledge method which is

mapped from source to target domain based on feature

representation. A new future space is created first, then feature

representation map is built, and the target and source domain

is reweighed. With the help of this, in the source domain,

classifier models are trained which is used by target domain.

[21] This paper tells the use of linear regression residual for

binary text categorization. The main idea is to predict the

given test vector using its k nearest neighbors in both positive

and negative classes. [22] In this paper, the problem of

classifying text by removing the information which is gained

from clustering both testing and training is addressed. The

knowledge which is gained from clustering is needed to

increase the performance of the text classifier. [23] This paper

explains the algorithm which combines the feature of k-

nearest neighbor (KNN) and support vector machine (SVM)

methods to improve the precision of classification of text

which is based on variable precision rough sets (VPRS). [24]

This paper explores on enhancing the kNN which is improved

by implementing alternate distance functions which has

weights to measure the data from various viewpoints. By using

genetic algorithm, the weights for optimization are computed

[25-29]. This paper gives the solution by Back propagation

network, the techniques used for feature identification. The


24

back propagation network algorithm is used for the text

classification.

3. PROPOSED METHOD

We have broadly classified our project into two stages where

the first one deals with conversion of handwritten characters

into editable format and another one deals with classifying

people as positive minded or negative minded based on the

content of what they write. We will see both the stages in

details in the following section.

STAGE 1:

Optical Character Recognition:

Handwritten character recognition is broadly classified as

online character recognition which is the real time acquisition

and recognition of characters and offline character recognition

which deals with recognition of characters which is written on

a sheet of paper. This can be achieved with three techniques

namely OCR (optical character recognition), MICR (magnetic

ink character recognition), OMR (optical mark recognition).

Pre-processing and segmentation:

The image is given as a input to the OCR template matching

algorithm where the characters are processed which involves

segmenting the characters by using horizontal and vertical

profiling using OCR techniques.

Feature Extraction:

For the given input image we get the corresponding vertical

profile of complemented image from where we extract the

required features of the segmented characters.

Text Classification:

We obtain matrix representation of recognized characters and

thereby the text is classified. And hence we get the editable

format which is the output of our first proposed model.

Algorithm 1: Character Recognition:

Input: A text image

Output: An editable document

Method:

for i=1 to length(Training_Samples)

img=imread(dataset(i));

No_Lines=HorizontalProfile(img);

No_Char=VerticalProfile(img);

Identified_Text=OCR(No_Char);

Save(“Identified_Text.txt”);

End

Stage 2:

As a person’s mind contextually set plays a major role in the

what a person writes, we are classifying them as positive

minded or negative minded based on the content of the written

matter.

Text mining and machine learning techniques:

The output of the previous model i.e., editable document is

further processed in this model to get the final output. We use

text mining techniques and machine learning techniques along

with stop word elimination algorithm to classify if the person

is positive minded or negative minded which our end result is.

A survey was conducted to understand the regional

vocabulary of people and we collected English words from

them which is classified as positive word and negative word by

them according to their thinking. The detailed model is

explained in the following section, here we use compression

based integer representation based approach for classifying the

extracted words as positive or negative. The task of classifying

is a supervised task where we train the classifying algorithm

with terms belonging to two major classes positive and

negative. The result of this classification algorithm is to assign

binary values {0, 1} (1->if the application recognises the terms

properly else its 0). We are emphasizing at using integers

based compression due to the fact that text terms occupies

more space than integers. Once we are able to convert the

terms to integers it will be very easy to handle the integer

numbers and hence it contributes a lot to classification

algorithm. The detailed explanation for this is presented in the

corresponding subsections.

Classification Stage:

We first read the positive dictionary of words and negative

dictionary of words and save it. Given the query document we

first apply natural language processing methods and eliminate

stop words. After which the set of positive and negative words

of the query document will be compared with the dictionary. If

found we keep the count of positive and negative words and

based on the frequency of occurrence of words we classify the

content of text as positive or negative. If the positive or

negative word which is present in the query document is not

found in the dictionary we update the dictionary with the new

words.

Algorithm 2: Personality Analysis

Input: A query Document

Output: classifying the document as positive or negative

Method:

Positive <- read pos_dictionary

save Positive

Negative <- read neg_dictionary

save Negative

[Prow Pcol]=size(Positive)

[Nrow Ncol]=size(Negative)

[Trow Tcol]=size(Reg_Text)

for i<-1 to Prow

for j<-1 to Trow

if(strcmp(Positive(i), Reg_Text(j))=1)

pos++


25

end

end

for i<-1 to Nrow

for j<-1 to Trow

if(strcmp(Negative(i), Reg_Text(j))=1)

neg++

end

end

end

End

4. EXPERIMENTATION

We have performed experimentation on three types of datasets.

They include well-formed characters, partially well-formed

characters and non-identifiable characters. Here we have

mainly used two types of experimentation techniques to find

the accuracy of the handwritten characters that are recognised.

1. The first one is using optical character recognition

technique (OCR).

2. The second one is using Text Mining

4.1 Optical Character Recognition:

STEP 1: Here we initially give handwritten sample as input to

OCR and find out the characters that are recognized.

STEP 2: Next we generate the Confusion matrix as follows:

● Create Rows and Columns of matrix using English

Alphabets.

● Mark those cells in the matrix depending on how the

alphabets are recognised.

● Calculate Row sum and Column sum, this will be our

Recall and Precision for finding F Measure.

F measure calculated for partially well-formed characters is

shown in Table 1.

4.2 Text Mining:

Steps:

1. Once the OCR data gets converted into editable text

format and this will be stored in a .txt file.

2. The obtained data set will be divided into training and

testing sample.

3. Stop words are removed from the training sample and

integer representation is given to the data. Results of text

mining stage is presented in Table 2.

5. CONCLUSION

This project can be useful for conversion of old handwritten

documents into digital form. This will help a lot of

organisations who have legacy documents in need of

digitisation. This can also help students to digitize their notes.

Right now there are scanners which are used to scan printed

documents. But this is will be a single app which can do the

work of a hardware device and its separate software. If this

technology is used for other languages we can easily convert

old books which are in need of restoration. This will convert

the book to digital form and prevent the natural wear and tear

that physical books are often subjected to. The personality

analysis can be used by various organisations to judge a person

before taking them into their organisation or to check the

changes in a person’s state of mind over time. This particular

study can have huge applications in the field of psychology.

We can see the similar works done in the references [26] to

[51].

ACKNOWLEDGMENT

This research project was supported by Department of

Computer Science & Engineering, Sahyadri College of

Engineering & Management, Mangalore. We thank all the

teaching and non-staff for their continuous support and

encouragement.

REFERENCES

[1] Aurangzeb Khan, Baharum Baharudin, Lam Hong Lee,

Khairullah khan, A Review of Machine Learning

Algorithms for Text-Documents Classification,

Department of Computer and Information Science,

Universiti Teknologi PETRONAS, Tronoh, Malaysia.

[2] S. N. Bharath Bhushan, Ajit Danti and Steven Lawrence

Fernandes. Integer Representation and B-Tree for

Classification of Text Documents: An Integrated

Approach.

[3] Thorsten joachims, GMD Forsehungszentrum IT, AIS.KD

Schloss Birlinghoven, 53754 Sankt Augustin, Germany

[4] Mita K. Dalal, Mukesh A. Zaveri Automatic Text

Classification: A Technical Review ,International Journal

of Computer Applications (0975 – 8887) Volume 28–

No.2, August 2011

[5] Songbo Tan, An improved centroid classifier for text

categorization, 2007 Elsevier Ltd.

[6] Serafettin Tasc, Tunga Güngör , Comparison of text

feature selection policies and using an adaptive

framework, 2013 Elsevier Ltd.

[7] Elias Oliveira, Patrick Marques Ciarelli. Claudine Gon¸

calves. A Comparison Between a kNN based Aproach and

a PNN Algorithm for a Multi-Label Classification

Problem, Universidade Federal do Esp´ırito Santo,Brazil

[8] KIMURA Kazuhiro SUZUOKA Takashi AMANO Sin-

ya, Association-based Natural Language Processing with

Neural Networks, Information Systems Laboratory

Research and Development Center TOSHIBA Corp.

[9] Chin Heng Wan a, Lam Hong Lee b, Rajprasad Rajkumar

b, Dino Isa, A hybrid text classification approach with low

dependency on parameter by integrating K-nearest

neighbor and support vector machine, 2012 Elsevier Ltd.

[10] Cheng-Shiun Tasi, Yong-Ming Huang, Chien-Hung Liu,

Yueh-Min Huang, Applying VSM and LCS to develop an

integrated text retrieval mechanism, 2011 Elsevier Ltd.

[11] Alberto F. De Souza, Felipe Pedroni, Elias Oliveira,

Patrick M. Ciarelli, Wallace Favoreto Henrique, Lucas


26

Veronese, Claudine Badue, Automated multi-label text

categorization with VG-RAM weightless neural networks.

Elsevier 2009.

[12] Robert E Schapire, Yoram Singer, Boos Texter: A

Boosting-based System for Text Categorization, Machine

Learning, 39(2/3):135-168, 2000.

[13] Louise Guthrie Elbert Walker, Document Classification

by Machine: Theory and Practice.

[14] Rey-Long Liu, Dynamic category profiling for text

filtering and classification.

[15] P. Ponmuthuramalingam and T. Devi, Effective

Dimension Reduction Techniques for Text Documents,

IJCSNS International Journal of Computer Science and

Network Security, VOL.10 No.7, July 2010 .

[16] Setu Madhavi Namburu, Haiying Tu, Jianhui Luo and

Krishna R. Pattipati, Experiments on Supervised Learning

Algorithms for Text Categorization IEEEAC paper

#1260, Version 8, Updated December 10, 2004 .

[17] S. Subbaiah, Extracting Knowledge using Probabilistic

Classifier for Text Mining, Proceedings of the 2013

International Conference on Pattern Recognition,

Informatics and Mobile Engineering, February 21-22.

[18] Tomek Strzalkowski and Barbara Vauthey, Fast Text

Processing for Information Retrieval, Courant Institute of

Mathematical Sciences New York University 251 Mercer

Street New York, NY 10012

[19] Susan Dumais, John Platt, David Heckerman, Inductive

Learning Algorithms and Representations for Text

Categorization

[20] Jiana Meng, Hongfei Lin, Yanpeng Li, Knowledge

transfer based on feature representation mapping for text

classification, 2011 Elsevier Ltd.

[21] Hakan Altınçay, Using Linear Regression Residual of

Document Vectors in Text Categorization, 2013 IEEE

[22] Antonia Kyriakopoulou, Theodore Kalamboukis, Using

Clustering to Enhance Text Classification”, SIGIR.

Amsterdam, The Netherlands. ACM 978-1-59593-597-

7/07/0007.

[23] Wen Li, Duoqian Miao, Weili Wang, Two-level

hierarchical combination method for text classification,

2010 Elsevier Ltd

[24] Takahiro Yamada, Kyohei Yamashita, Naohiro Ishii,

Text Classification by Combining Different Distance

Functions with Weights”, 2006 IEEE

[25] S.Ramasundaram, S.P. Victor, Text Categorization by

Back propagation Network, International Journal of

Computer Applications (0975 – 8887) Volume 8– No.6,

October 2010.

[26] Bhushan Bharath S. N. and Danti Ajit. Classification of

text documents based on score level fusion approach.

Pattern Recognition Letters 94., 118–126. 2017.

[27] Danti Ajit and Bhushan Bharath S N. 2013, Document

Vector Space Representation Model for Automatic Text

Classification. In Proceedings of International Conference

on Multimedia Processing, Communication and

Information Technology, Shimoga. pp. 338–344

[28] Danti Ajit and Bhushan Bharath. Classification of Text

Documents Using Integer Representation and Regression:

An Integrated Approach.Special Issue of The IIOAB

Scopus Indexed Journal.Vol. 7, No.2, pp. 45–50. 2016.

Table 2: F measure calculated for recognized characters.

F Measure of recognized handwritten text in editable format

PRECISION RECALL F MEASURE

0.91 0.9225 0.9162

Table 1: F measure calculated for partially well-formed characters.

F Measure of recognized handwritten text in editable format

PRECISION RECALL F MEASURE

1 0.9285 0.9523



27

Recognition of Overlapping Sound Events

Jayalaxmi*, Hegde Abhijna Satish, Harshitha N Kotari, and Deeksha

Department of Computer Science and Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007


ABSTRACT

In this paper, we address the challenge of recognizing the isolated sound in the noisy background. Here we propose an

approach where we extract the local spectrogram features for each isolated acoustic sound events. The local spectrogram

features are extracted by using the keypoints which are unique for each sound event. The “keypoints” are the peak values

for each sound event where the sound is maximum. These local spectrogram features are then clustered to form a

codebook. The codebooks are used for training purpose. The features of the sound events which will be used for testing

are extracted separately using spectrogram. The extracted feature is then mapped with the local spectrogram features in

the codebook to recognize the sound event. The experimental setup has 12 isolated sounds, 12 overlapped sound events,

and 11 mixed noises to determine the accuracy of our approach.

Keywords: Cluster, Codebook, Keypoint detection, Local spectrogram features, Time Frequency Location, visual word.

1. INTRODUCTION

In any environment there can be many overlapping sound

events which will be present along with many background

noises. In many cases the background noise will be as

important as the structured sound events, so they cannot be

simply neglected and considered as unstructured sound events.

In cases like surveillance camera, hearing machine and also

automatic speech recognition the unstructured surrounding

background noises are as important and useful as the

structured sound events. So, the concept of Sound Event

Recognition (SER) is used to detect and also to classify the

sound events which are present in the unstructured

environment. Detecting and classifying these sounds based on

the sound events are helpful in case of security cameras,

monitoring of bioacoustics, meeting room transcription and is

also very helpful in case of “hearing machines”.

Different Technologies have been developed regarding

sound event recognition, most popular techniques among them

are based on frame-based features, such as Mel-frequency

cepstral coefficients (MFCCs) from ASR, or MPEG-7

descriptors (Casey,2001). All these techniques can then be

modeled with Gaussian Mixture Models (GMMs) and

combined with Hidden Markov Models (HMMs) in order for

recognition, and it can also be used to train SVM or Support

Vector Machine for the different classifications based on the

features. But these methods may not perform best in the case

of mismatched conditions which occur in sound events

recognition tasks.

To overcome these challenges Missing Feature Recognition

systems were developed. The task here is to identify how to

mask the sound so that it stands out separately from the

background noise. The performance of this system depends on

how well the mask can separate the sound from background.

This technique may not be helpful in case of overlapping

sounds as there will be information about two or more sounds.

In a research of humans understanding of speech traces of

frame-based feature is found, and it is also found that the

human auditory system may be based on the partial feature

extraction that are uncoupled and also local across the

frequency of the speech. This helps humans to recognize the

speech even if there is lot of disturbance and distortion across

the different regions of the spectrogram of the sound event.

Thus, based on Local Spectrogram Features we develop a

Sound Event Recognition system, where we will making use of

frame-based features.

Here we try to address a task of simultaneous recognition of

the sound events which are from single channel audio.

Conventional Frame-based methods cannot be used here as

each time the frame will contain information mixed as the

sound event will have different sounds or from multiple

sources.

Another method which can be used to detect the

overlapping sound is Missing Feature Recognition technique.

The drawback with this technique is that the recognition of the

sounds will be based on the way mask is created or in other

words the recognition of sound depends on the accuracy of the

mask. Here we try to make use of Local Spectrogram Features

which represent the local spectral feature of each sound and

this is extracted from spectrogram which is covered with

keypoints. “Keypoints” represents the peaks in the

spectrogram. Based on these keypoints we can form LSF

clusters and their occurrences can be shown using

spectrogram.

We have conducted experiments on the isolated sound

without background noise, and also on isolated sounds with

factory floor noise as background noise with the background


28

noise taken in different decibels, and finally on overlapped

sound events.

2. RELATED WORK

Sound event classification is used for applications like

security surveillance [1], bioacoustics monitoring [2], meeting

room transcription [3] and mainly in machine hearing [4].

Sound events original feature can be extracted by using visual

signature which is the representation of sound’s frequency.

These features can be extracted using spectrogram which is

gray scale normalized [5]. Another way of extracting local

spectrogram features is by making use of keypoints concept.

The keypoints are the peak values of the sound. The local

spectrogram features can be extracted by extracting the values

around keypoints. These extracted values are unique for each

sound. These extracted local spectrogram features along with

the label name is used to train the SVM model [1].

To separate the sounds, audio event detection is used. To

classify the audio events the system uses two parallel GMM

classifier. The classifier is trained initially with the audio

features which are obtained using 2-step process. At different

signal to noise ratio such as 0dB, 10Db, and 20dB the

experiments were done. The approach is applicable to separate

two sounds for the noisy background. Firstly, the features are

extracted from the audio events and these features are used for

analysis [6].

Spectral subtraction is used to separate the noise from the

sound events. Training is used to improve the performance of

automatic speech recognition. In multi condition training the

system is trained in different situations where it can work.

Hierarchical spectro-temporal processing is used to extract

features from the noisy background [7].

Invariant features present in the sound event does not

change its application under any circumstances. These

invariant features of the sound should match with the objects

in the surrounding. Feature should be correctly matched with

features in the database of features of the known sound [8].

Hearing machines are present to detect the speech from the

musical environment and background noises. Using these

features the machine can recognize the speech of the sound in

the noisy environment [9]. Built on the key developments in

statistical modelling of natural language processing and

involuntary identification schemes, there is extensive

submission in works which need a humanoid mechanism

interface, such as auto call processing in the telephony system

and enquiry-based data system which does work such as

providing upgraded portable data, stock rate extracts, climate

report [10].

Based on the description of the probabilistic mixed

prototype for a frame of speech the recognition of speech is

performed. Every part of prototype is the naming stage of

Hidden Markov model grounded identification of speech [11].

3. IMPLEMENTATION

Figure 1. Overview of proposed method

3.1 Keypoint detection

The feature of the sound is first extracted using MATLAB and

is stored into mat file. This mat file is then converted into a csv

file. This csv file is then given as an input to spark program.

The spark program identifies the peak values of the given

sound. The identified peak values are then used to detect the

keypoints corresponding to the sound. These keypoints are

used to extract local spectrogram features.

The Keypoint Detection is summarized as follows:

Algorithm: Detecting Keypoints

Input: csv file contenting the feature of the sound.

Output: Detected Keypoints for that sound.

1. Extract the values from .csv file and split them as

comma-separated value.

2. Make frames of size -6 to 6.

3. Find max for each frame.

max(float(x) for x in x.split())

4. Find the sum of each frame which is divided by 40.

5. Keypoints should be greater than max value and sum

value.

6. Print the keypoints.

The frame size -6 to 6 and the sum value divided by 40

gives more accurate Keypoint values.


29

3.2 Local spectrogram feature extraction

Local spectrogram features for a sound are the values

surrounding the keypoints. Once the keypoints are detected we

can then extract the local spectrogram features using these

keypoints. The extracted local spectrogram features is then

grouped into clusters. Clustering is done using K-Means

clustering algorithm. This makes use of Euclid’s distance to

form the clusters. The Euclid’s distance formula is as follows:

…. (1)

Where (x1,y1) and (x2,y2) are the co-ordinates of keypoint

selected for clustering. The clusters are grouped based on the

distance obtained using Euclid’s formula. In K-Means, K

refers to number of clusters. In our approach we are making

use of 500 clusters which are formed using 50 iterations

3.3. Codebook

For each of the clusters formed mean is calculated which is

called visual word. The collection of all those visual word is

called codebook. Then we are matching extracted features or

the input sound with this codebook to predict the sound event.

Fig 2 shows the generation of codebook.

The extracted local spectrogram features for the input is

matched with the codebook line by line and features with least

distance will be plotted in the histogram along with the

corresponding labels. The label for which the histogram value

is highest is predicted as the output. Fig 3 shows the plotting

of histogram. In the figure, x-axis represents labels and y-axis

represents values.

Figure 2. Codebook generation

3.4. Training and testing

From each sound class 32 sounds are taken for training and 8

sounds are taken for testing. During training the features and

the labels are given to support-vector machine (SVM).

Figure 3. Histogram

During testing only, features are given and the

corresponding labels are given as output from trained SVM.

To classify the sound events and to predict the accuracy SVM

Chi-squared Kernel is used. Chi-squared Kernel is very

popular for training non-linear SVM. The Chi-squared Kernel

is calculated as follows:

.… (2)

Where x and y need to be non-negative and should be

normalized.

Accuracy is calculated by predicting how correctly the sound

is recognized. SVM classifiers are used to maintain a balance

between accuracy of training and the strength of the classifier.

4. EXPERIMENTAL RESULTS

4.1. Datasets

For our experiment we are using the following 12 classes of

sound: Applause, Cup Jingle, Chair Moving, Cough, Door

Slam, Key Jingle, Knock, Keyboard Typing, Phone Ringing,

Paper Work, Steps, and Laugh. For convenience in table 2 it is

represented as Ap, CJ, CM, Co, DS, KJ, Kn, KT, PR, PW, St,

La respectively. Each of this single class consists of 40 sounds.

Firstly, we have considered all the above classes in its isolated

form amongst which 32 sounds were used for training and 8

sounds were used for testing. Next we mixed the above

isolated sounds with the factory floor noise which is from

NOISEX’92 database. The sounds were mixed in 0dB, 10dB,

and 20dB. Next, we have considered single sound from 1 class

(i.e. Laugh) and mixed with other classes to form mixed sound

events.

4.2. Results

As estimation we are measuring the accuracy of the

recognized sound events. The results of our experiment can be


30

found in Table I. As shown in the Table 1 isolated sounds have

more accuracy compared to overlapped sounds. The factory

floor noise of 20dB has more accuracy compared to the 10dB

and 0dB.The sound events with 20dB has less noise compared

to the 0dB.Mixed events have less accuracy compared to the

other overlapped sound since there will be confusion to

recognize the sounds. Here we are taking a single sound from

single class and mixing with all the other 11 classes. Our

approach finds it difficult to distinguish the two sound events

present in the mixed environment. In Table 2 has a confusion

matrix for mixed sound events where the confusion gives the

clear picture of how well the sound has been recognized

correctly and where it has been mis-predicted. In table 2 we

have taken confusion matrix of mixed sound events that has

given the accuracy of 71.591%. In this 8 applause sounds that

are given for testing is correctly recognized as applause itself.

But in cup jingle sound only 4 out of 8 sounds were

recognized correctly, remaining 4 sounds are mispredicted one

as chair moving, one more as keyboard typing and the

remaining 2 as paper work. Because of these types of

mispredictions, the accuracy has been decreased.

5. CONCLUSION

In this paper a technique to recognize the event is proposed in

overlapped noisy form. Our motivation is derived from social

observation, which has been recommended for human listening

built on confined evidence, and also after picture entity

identification that will make parallels with overlying SER. The

methodology we made use of is to discover keypoints in the

spectrogram, later portray the sound conjointly via the LSF

and the key-point dispersal in relation with the sound

inception. Further as future deed, our goal is to improve the

accuracy of mixed sound events. Also the work may comprise

reconstruction of the recognized acoustic events.

Table 1 Experimental Result under various test conditions

Experimental

setup

Acoustic events

Isolated

Sound

Events

Noisy sound Mixed

Sound

Events 0

dB

10

dB 20dB

Accuracy 88.542 74.583 79.167 82.292 71.591

Table 2 Confusion Matrix for mixed events

Ap CJ CM Co DS KJ Kn KT PR PW St

Ap 8 0 0 0 0 0 0 0 0 0 0

CJ 0 4 1 0 0 0 0 1 0 2 0

CM 0 0 5 1 0 0 0 0 1 1 0

Co 0 0 0 7 0 1 0 0 0 0 0

DS 0 0 0 0 8 0 0 0 0 0 0

KJ 0 0 0 1 0 6 0 0 0 1 0

Kn 0 0 0 0 1 0 6 0 0 0 1

KT 0 1 0 0 0 1 0 4 0 2 0

PR 0 0 0 3 0 0 1 1 3 0 0

PW 0 1 0 0 0 1 0 0 0 5 1

ST 0 0 0 1 0 0 0 0 0 0 7

ACKNOWLEDGMENT

This research was supported by Sahyadri College of

Engineering and Management. We are grateful to Mr. Sunil

B.N and Dr. Pushpalatha K who moderated this paper and in

that line improved the manuscript significantly.

REFERENCES

[1] Gerosa, L., Valenzise, G., Antonacci, F., Tagliasacchi,

M., Sarti, A., 2007. Scream and gunshot detection in

noisy environments, in: 15th European Signal Process.

Conf. (EUSIPCO-07), Sep. 3–7, Poznan, Poland.

[2] Bardeli, R., Wolff, D., Kurth, F., Koch, M., Tauchert, K.,

Frommolt, K., 2010. Detecting bird sounds in a complex

acoustic environment and application to bioacoustic

monitoring. Pattern Recognition Lett. 31, 1524–1534.

[3] Temko, A., Nadeu, C., 2009. Acoustic event detection in

meeting-room environments. Pattern Recognition Lett. 30,

1281–1288.


31

[4] Dennis, J., Tran, H., Li, H., 2011. Spectrogram image

feature for sound event classification in mismatched

conditions. IEEE Signal Process. Lett. 18, 130–133.

[5] Dennis, J., Tran, H., Chng, E., 2012. Overlapping sound

event recognition using local spectrogram features with

the generalised hough transform, in: Proc. Interspeech

2012.

[6] Heckmann, M., Domont, X., Joublin, F., Goerick, C.,

2011. A hierarchical framework for spectro-temporal

feature extraction. Speech Comm. 53, 736–752.

[7] Lowe, D., 2004. Distinctive image features from scale-

invariant keypoints. Internat. J. Comput. Vision 60, 91–

110

[8] Lyon, R., 2010. Machine hearing: an emerging field.

IEEE Signal Process. Mag. 27, 131–139.

[9] Nádas, A., Nahamoo, D., Picheny, M., 1989. Speech

recognition using noise-adaptive prototypes. IEEE Trans.

Acoustics Speech Signal Process. 37, 1495–1503.

[10] O’Shaughnessy, D., 2008. Invited paper: automatic

speech recognition: history, methods and challenges.

Pattern Recognit. 41, 2965–2979.



32

Janardhana D R*, Ashreen, Anusha, Ashoora and Gladys Merlyn Dsouza

Department of Information Science and Engineering, Sahyadri College of Engineering & Management, Mangaluru - 575007


ABSTRACT

Over the centuries, humans have tried to reduce energy and one of the most efficient way to achieve this is to build a

smart city. Many modernist cities have started using different methods to reduce energy consumption and also to create

healthy surroundings. Few methods are used specially regarding road safety of passengers and waste management;

thereby we are introducing a simplest way to avoid such problems using an idea of Internet of Things (IoT) i.e. an “IOT

Based Energy and Waste Management System for Smart Cities”. Internet of Things is a simple way of connecting the

software and hardware components to the internet. IoT refers to an emerging model consisting of combination of

uniquely addressable things communicating with one another to form dynamic networks. Here this idea is used on street

lights and dustbins where street lights promote security across city and improves safety for drivers, riders and

pedestrians. Similarly, nobody wants to be in a place which is encompassed with waste or has awful smell, hence we

introduce smart dustbins, which are capable of sending alerts when they are full and can be emptied immediately, giving

no chance for the dustbin to be over flown .

Keywords: Arduino Microcontroller, Dustbin, ESP Module, IoT, Sensor, Streetlight

1. INTRODUCTION

The Internet of Things (IoT) is the network of hardware and

software devices connected on to the ever available internet.

Each factor is unambiguously recognizable through its

embedded automatic data processing system. At times IoT is

connected to an object and then they are invoked and

controlled remotely. The IoT will usually be connected to

associate information processing network to the worldwide

net. IoT focuses on reducing human intervention. Commercial

IoT, where local communication is usually either Bluetooth or

LAN (wired or wireless), the IoT device will typically

communicate only with local devices. The IoT promotes

increase the level of awareness concerning our world and a

platform to observe the reactions to the dynamic conditions.

India is a developing nation with massive population, due to

which the energy consumed and waste produced is huge. Thus

there is need for saving these resources and also to keep the

city clean. Since energy is one in all the foremost vital

resource in our life so it should be employed in economical

ways to make use of the energy and save energy for future

requirement. Over the century, humans have tried to cut back

energy and one among the foremost economical manner is to

create a sensible green home. Several modernist cities started

adopting this technique to cut back energy and environmental

pollution over the past few years. Few of the major reasons

are basically throughout night time all the lights on the main

road stay ON for the passengers, similarly as for vehicles,

however ton of energy is wasted once there's no vehicle or

traveler movement. Dominant of street lamp is of utmost

important in developing country like India to cut back the

power consumption. Saving of this energy could be a vital

issue lately as energy resources are getting reduced day by

day.

Secondly, waste management is additionally one among the

first drawback that the globe faces irrespective of the case of

developed or developing country. Within the contemporary

situation, many times we see that the dustbins are placed at

public places within the town are seen flooded because of

increase in the waste daily. It creates insanitary circumstances

for the folks and creates awful smell which leads in spreading

of some deadly diseases.

To avoid such issues, we are progressing to vogue “IoT

Based Energy and Waste Management for Smart Cities”. This

project describes about the circuit that switches the road lights

ON during dedicated amount of time and remains OFF after

steady time. The dynamic street lamp management depends on

an inter-connectable IoT. Novel methods are adapted to make

the connection of hardware units to speed up the connection to

the internet faster. These details are usually accessed by the

IoT Based Energy and Waste Management for

Smart Cities


33

municipal authorities from their place with the help of

software. Admin has the authority to read the status of

streetlight that is if the light is active or inactive and also read

the dustbin that is he can check the level of waste. He/she can

also view the graph that is the execution time.

2. RELATED WORK

Nabil Ouerhani et al. [1], the paper affords a real-global

established solution for dynamic street light managing which is

based on the fancy internet. Enormous interest is delivered on

interoperability degree, the use of novel tool connection

concept based totally on version-driven communique

marketers to connect the sensors to the internet. The paper

shows the outcomes from actual-international assessments with

deployed dynamic road lighting fixtures in cities. The

proposed model will trying save the energy up to an estimate

of 50% of the power consumed. Zigbee unit used in this

system will ensure the safety in the operations.

Rohaida Husin et al. [2], the paper gives an idea about

automatic streetlamp system supported by an inexpensive

microcontroller. This method consists of a microcontroller,

lightweight device, rain device, optical device and a collection

of the LED units. The system was programmed to

mechanically shut at the daytime and work during the night.

Many numbers of tests are conducted to check and validate the

projected image within the completely different environment.

As conclusion, around at the max there is 80% of reduction in

electricity usage can be achieved. This paper makes a case for

the uses microcontroller and detector to browse the input and

processes it to get needed output. It uses low price

microcontroller. It doesn’t use IoT.

Deepak Kapgate [3], the paper describes the employment of

wireless sensing element network for streetlight observation

and management. As we glance at existing systems the ability

consumption and maintenance price of streetlight management

department is high. This technique would provide associate in

nursing optimum cost accounting for lamp maintenance and

control. System utilized use of network process device for

sensing of sunshine then gathered information is used for

dominant lamp ON/OFF. Lifetime of streetlight depends on

the length for which they get used. During this analysis, this

will be able to try and scale back the required length that lamp

ought to get on, focuses on increasing band for network nodes

to urge most possible rate and additionally discuss the

parameters needed to automatic detection and removal of

nodes within the network. This paper discusses concerning the

usage of Wireless Sensor Network (WSN) to control and

monitor the road light-weight. The control center will manage

and monitor all street at real time. It uses Zigbee that may not

secure and conjointly the coverage is restricted.

Dr. N. Sathish kumar et al. [4], The problem that people face

in current days are the garbage is overflown and there is no

proper waste management. It in turn ends up in numerous

hazards appreciate dangerous odour and create unhealthy

atmosphere which could be seen as a number one cause for

uninvited diseases. To prevent all these dangerous situation

and to maintain a healthy surrounding, the proposed work is to

have a feasible garbage management system. This paper tries

to tackle the problem by cleaning the garbage as soon as it is

full with keeping in mind some of the criteria, primarily based

on level of garbage filling. The garbage system works in very

simple way, once the bin is full to certain level the alert is sent

the respective municipal authority sever and the cleaning is

taken care of. The proposed system is taken care with the help

of RFID tags, these tags help in verifying the signals that are to

be sent to the system. The IoT system is segregated with the

other components and the process is completed.

Twinkle Sinha et al. [5], the paper discusses relating to the

design of model for a ‘Smart Dustbin’ that indicates directly

that the waste bin is crammed to a particular level by the

rubbish and cleansing or evacuation them may be a matter of

immediate concern. This prevents lumping of garbage within

the edge waste bin that finally ends up giving foul smell and

health problem to folks. The design of the sensible waste bin

includes one directional cylinder associated an arduino UNO.

Here the garbage system uses the alerting way to notify the

municipal through some method of alerting system. There is a

level sensor which indicates the level of the garbage, it sends

notification when a defined level is reached. The signals are

received as in a format of glowing diodes.

3. SYSTEM DESIGN

System design is the process of describing the data,

architecture and modules for a system to fulfill the specified

requirements.

3.1 Architecture Diagram

Architectural diagrams are used to illustrate the relationship

between different components of a system. It is very important

to understand the complete concept of the system. The figure1

shows the architectural diagram of our proposed system. The

proposed system consists of IR sensors which will identify the

level of the garbage bin. When the garbage bin is full it sends

the level details to the arduino microcontroller, which sends

alert to the admin. The streetlights are monitored for their

states and their status is sent to the admin. ESP module is used

for communication with the web server.

4. IMPLEMENTATION

Implementation is the process of carrying out an execution,

application of a plan in particular manner. The process of

implementation is done to see how information will be

processed, installed, deployed into a working operational unit.

4.1 C#

C# is a programming language that encompasses functional,

imperative, generic, object-oriented (class-based), and

component-oriented programming disciplines. C# is intended

to be a simple, modern, general-purpose, object-oriented

programming language. We have used c# to design a software.

4.2 Microsoft Visual Studio 2010

Microsoft visual studio is an integrated development

environment from Microsoft. It is used to develop console and


34

graphical user interface applications along with websites and

web pages. Visual Studio supports different programming

languages by means of language services. Visual studio is used

for front end to design our software.

4.3 Microsoft Visual Studio 2010

Microsoft visual studio is an integrated development

environment from Microsoft. It is used to develop console and

graphical user interface applications along with websites and

web pages. Visual Studio supports different programming

languages by means of language services. Visual studio is used

for front end to design our software.

4.4 MySQL

MySQL is an open source Relational Database Management

System (RDBMS) that runs as a server providing multi-user

access to a number of databases. The MySQL development

project has made it source code available under the terms of

the GNU General Public License, as well as under a variety of

proprietary agreements. MySQL is a popular choice of

database for use in web applications, and is a central

development of the widely used LAMP open source web

application software stack. My Sql is used for back end to

design our software.

4.5 Ultrasonic Sensor

In this project we have made use of Ultrasonic sensor.

Ultrasonic sensor mainly works by using ultrasonic signal, it

sends the signal to check the distance between the objects.

Sound wave is sent at particular frequency to measure the

distance there is some frequency at which sound bounce back.

When the object is detected it send those values to the

processing unit.

4.6 Arduino Uno

Arduino is an open source software and hardware company

that manufactures single board micro controllers. Arduino

boards have an Atmel 8-bit AVR microcontroller. Arduino

Uno is one of the most common one these day, they have six

pins for pulse-width modulated signals and another six analog

inputs, which can also be used as six digital I/O pins and 14

digital pins. Which is main unit in our project that combines

the two circuit.

4.7 IR Sensor

IR Sensors works by selecting light wavelength in the InfraRed

(IR) spectrum by using a specific light sensor. One can look at

the intensity of the received light by comparing the intensity of

received light.

4.8 LCD

LCD is used to display the percentage of dustbin fill in our

project.

Figure 1: Architecture diagram of proposed system.

4.9 ESP 8266

It connects microcontroller i.e arduino uno to wi-fi network

and creates a simple TCP/IP connection. The two pins of esp

module was used to create the Communication i.e is Rx and

Tx. One is used for transmitting and other one is used for the

receiving purpose

5. RESULTS AND ANALYSIS

The hardware prototype of our project “IoT based Energy

and Waste Management for Smart Cities” is been developed

using arduino as the microcontroller, IR as level sensor, There

is a screen to monitor the garbage level. When the level is

25%, 50%, or 75% the level message is shown to the admin.

Streetlights are turned ON during the night time and switched

OFF during day, the status of streetlight is displayed to the

admin.

Figure 2. Graph indicating the status of dustbin

6. CONCLUSION AND FUTURE WORK

IOT based energy and waste management aims at keeping the

environment clean by continuously monitoring the dustbin.

Here the level of waste filled in the bin can be detected by the


35

admin thus helping in cleaning the bin immediately once it is

filled and also helps in monitoring the streetlight hence

preventing the accidents caused at the night and any other

filthy acts. Hence this project would be of great benefit to the

society.

The proposed system is built in such a way that there is

room for further enhancement. New additional features can be

added without any hassle. In future this system can be

designed as mobile/IOS application so that the user can

publish the online publication without cost per click.

REFERENCES

[1] Nabil Ouerhani, Nuria Pazos, Marco Aeberli, Michael

Muller, “IoT Based Dynamic Street Light Control for

Smart Cities”, IEEE University of Applied Sciences,

Switzerland, 2016.

[2] Rohaida Husin, Syed Abdul Mutalib Al Junid, Zulkifli

Abd Majid, “Automatic Street Lighting System for Energy

Efficiency Based on Low Cost Microcontroller”,

International Journal of Simulation Systems, Science and

Technology, Vol.13, 2012.

[3] Deepak Kapgate, “Wireless Streetlight Control System”,

International Journal of Computer Applications, Vol.41,

2012.

[4] Dr.N.Sathish kumar, B.Vijayalakshmi, R.Jeniferprarthana,

A.Shankar, “IoT Based Smart Garbage Alert System

Using Arduino UNO”, 2016.

[5] Twinkle Sinha, K.Mugesh Kumar, P.Saisharan, “Smart

Dustbin”, International Journal of Industrial Electronics

and Electrical Engineering, Vol.3, 2015.



36

Ads Recommendation Using Data Mining

Akhila Thejaswi R*, Mohammed Mohiddin, Deshik, Jnanesh Shetty and Nilesh Kumar

Department of Information Science and Engineering, Sahyadri College of Engineering & Management, Mangaluru – 575007


ABSTRACT

Every vendor wants to publicize their products. They do it by showing ads on various platforms like social media,

television etc. These ads will sometimes be of use to customer or it won’t be of any use to them. Sometimes unwanted ads

may even annoy the customers. So in order to avoid such cases we develop a project that provides ads based on location,

time and interest of the customer. The interest will be given by the user via social media and the location and system time

will be calculated by the software. Therefore if a customer likes something in social media, that will be taken as interest

and whenever user comes across such a location based on his interest, ads of such vendors or shops will be given. This will

help both vendors to provide its ad and customers to get the ads it likes.

Keywords: User Interests, Location based, Recommendation, Data Mining, IP Address, Offers and Vendors.

1. INTRODUCTION

Data mining is a process that allows sorting of large data sets

to identify patterns and establishing connections to solve a

particular problem using data analysis. Data mining tools also

allows the enterprises to predict their future trends.

In data mining, association rules can be created by analyzing

the data that requires frequent use of if/then patterns, and using

the confidence criteria one can locate the most important

connections within the data. The other data mining parameters

can include Classification, Clustering, Forecasting and Path

Analysis.

Data mining techniques can also be used in many research

areas including marketing, mathematics, cybernetics and

genetics. While data mining techniques are meant to drive

efficiency and predict the behavior, if correctly used, any

business can set itself apart from other computational business

through the proper use of predictive analysis. Data mining is

primarily used by the companies which have a strong

consumer focus like retail, financial, marketing organizations

and communications.

In general, the benefits of data mining comes from the ability

to uncover hidden patterns and relationships in data that can be

used to make predictions that impact businesses.

Likewise, the development of data warehouse also uncovers

the immediate way you are currently practicing: The

requirement for an intense, easy-to-use and economic data

warehouse created for the cloud to bank all your data in one-

single point and use and analyze it later. Therefore, the modern

data warehouse came as an effective data solution.

The proposed work is a web based application using location

and interest of the user. First, we collect the interests of the

users and store it in a file. The location and the ads of the

vendors are provided to the admin by the vendors itself. The

ads are verified based on certain criteria by the admin. The

verified ads are added to the LIT software.

When the user reaches a certain location near to that of the

vendor at a particular time, the ad will pop up on the screen.

The user can now use the help of the LIT software and use the

ad or can just ignore it. The decision is completely based on

the user whether or not to use the ads.

2. RELATED WORK

In the recent years, wireless networks and mobile technologies

have shown a rapid growth. This has led to opportunities for

marketers and advertisers which include satisfaction of the

customer, customer engagement etc. The main challenge for

advertisers and marketers is to analyze huge amount of data

emitted by mobile devices. And also provide customer

engagement from the mobile data. Lei Deng et al. [1],

addresses this challenge by introducing a framework that

recommends ads by using big data analytics. The algorithm

used here is clustering algorithm. GEO information integration

with profile datasets is also been used. The framework

proposed here provides a decision based approach to handle

various cases associated in pushing ads towards the end users.

Ananthi Sheshasaayee and H. Jayamangala.[2], focuses

mainly on a node in a network called the influential node or an

information hub. This influential node has a large number of

contacts in social networks. So, any recommendations of ads

will reach a vast number of users. The technique used in order


37

to identify the information hub or influential node is beSt

sPreadeR Identi_catioN using K-shelL graph structurE

(SPRINKLE) approach. In future, more focus is given on

popularity of the item relevant to the preferences, as they are

considered as the main factor for recommendation of ads.

Social networking is not only seen as a means of

entertainment, but also an active tool for marketing. Andy

Bengal et al.[3], proposes a framework that empowers

marketers and analysts by providing them with all the

information related to tags that are deployed on websites in

real time. In order to find any information about the behavior

of a user of any webpages a “tag" is used. A tag is a small

piece of code which will be embedded in the webpages in

order to find any information related to the user of the

webpage, for example, the visitor's browsing behavior. The

method used here is real-time tag discovery and an analysis

tool.

In business field, Internet has been used as a primary tool

for advertising and marketing. Most of the advertising

solutions these days use Behavior Targeting (BT) technology

to provide services to the end users. Lei Deng and Jerry Gao.

[4], proposes a system with the help of data mining algorithms

and machine learning solutions. The main objective in this

paper is to provide static services for advertisers. Services

such as when, where what and how to place advertisements

will be provided to the advertisers. This not only recommends

the advertisements but also predicts trends. The system also

uses NoSQL database technologies. This will allow advertisers

to reduce cost and improve effectiveness. The algorithm used

here is Synthesis Index Strategy (SIS) Algorithm. Ads will be

recommended using location and current state of art

technologies.

Nico Neumann. [5], discusses the development of

advertising and marketing technology. The system is based on

Automated buying processes and micro-targeting

personalization. It also uses Location-based targeting

combined with profile data.

Badrish Chandramouliet al.[6], presents a big data

application for demand-side plat-form(DSP), for mobile

display advertising. The algorithm used here is behavior

targeting (BT) Algorithms with TiMR. In order to increase the

efficiency of the campaigns through information collected

based on visitors browsing data, by advertisers and marketers.

Validate the approach by proposing a new end-to-end solution

using temporal queries for BT.

3. ARCHITECTURE OF THE PROPOSED

MODEL An architecture model is concerned with a set of tradeoffs

inherent in structure and design of a system.

The figure 1 shows the architecture of our system. The system

is divided into two phases; train phase and test phase.

Initially, the file will contain all the user details like user

interests. The Ads can be viewed by the user once their

location is fetched by the LIT software.

The locations of the vendors are verified and also their Ads

based on admin's strategies. The admin can discard or use the

given Ads. The verified Ads are added to the LIT software.

The user when arriving at a certain location at a certain time,

he can view the Ads based on his interests. The user can ignore

or visit the store of the vendor.

Figure 2: Architecture diagram for ads recommendation

4. IMPLEMENTATION

4.1 Front End

Front end is through which all end-users interact with the

software. There are many styling languages because it is

mainly concerned with the design. We have used JSP to design

our front end. JSP is nothing but JavaServer Pages which is

used to create web pages dynamically. It is similar toASP and

PHP but here Java programming language is used.

4.2 Pseudo code

The pseudocode for Admin end is as follows:

1. BEGIN

2. IF admin login is successful

3. IMPORT JSON and CSV

4. IF successfully imported

5. EXTRACT keyword

6. UPDATE user interest table with user interest and

weight

7. ELSE DISPLAY error message

8. ADD Vendors and provide login credentials to

vendor

9. Admin LOGOUT

10. ELSE DISPLAY invalid credentials.

11. END

The pseudocode for Vendor end is as follows:

1. BEGIN

2. IF vendor login is successful

3. ADD offers,DELETE offers,UPDATE offers

4. Vendor LOGOUT

5. ELSE DISPLAY invalid credentials

6. END


38

The pseudocode for User end is as follows:

1. BEGIN

2. IF user registration is successful

3. ENABLE Location

4. LOGIN to view ads

5. IF login is successful

6. IF offer is available

7. DISPLAY OFFERS based on user location and

interest

8. ELSE DISPLAY no offer available

9. ELSE DISPLAY invalid credentials

10. User LOGOUT

11. END

4.3 Back End

The back end is where the data is stored and is accessed by

end-users from the front end. In our system, the software

database is kept at back end and mySQL is used in order to

access or manipulate data.


The software has been used at various places in order to check

the output which the software has to yield.

The performance which is measured in time here is

compared with the size of dataset.

A graph is obtained which is shown in figure 4.

Figure 1: Represents the place - 1

6. CONCLUSION

The proposed system recommends ads based on three factors

i.e., Location of the user, Interest the user has shown in social

media and until the time the offer is valid for. The system will

provide details about the ongoing offers around him. The

objective of the system is to avoid unwanted ads which one

gets while he is nowhere concerned about it.

Figure 3: Represents place - 2

Figure 4. Graph: time vs dataset

ACKNOWLEDGMENT

We are grateful to Dr. R Srinivasa Rao Kunte, Principal,

Sahyadri College of Engineering & Management, Dr. Umesh

M. Bhushi, Director Strategic Planing, Sahyadri College of

Engineering & Management and Dr. D. L. Prabhakara,

Director, Sahyadri Educational Institutions, who have always

been a great source of inspiration. Finally, yet importantly, we

express our heartfelt thanks to our family & friends for their

wishes and encouragement throughout the work.

REFERENCES

[1] Deng, Lei, Jerry Gao, and Chandrasekar Vuppalapati,

"Building a big data analytics service framework for

mobile advertising and marketing", 2015 IEEE

First.International Conference on Big Data

Computing..Service and Applications.


39

[2] Sheshasaayee, Ananthi, and H. Jayamangala. "A study on

the new approaches for social network based

recommendations in digital marketing", 2017

International Conference on Innovative Mechanisms for

Industry Applications (ICIMIA).

[3] Bengel, Andy, Amin Shawki, and Dippy Aggarwal.

"Simplifying web analytics for digital marketing", 2015

IEEE International Conference on Big Data (Big Data).

[4] Deng, Lei, and Jerry Gao. "An advertising analytics

framework using social network big data", 2015 5th

International Conference on Information Science and

Technology (ICIST).

[5] Neumann, Nico, "The power of big data and algorithms

for advertising and customer communication,

“International Workshop on Big Data and Information

Security (IWBIS)”, IEEE, 2016.

[6] Chandramouli, Badrish, Jonathan Goldstein, and Songyun

Duan. "Temporal analytics on big data for web

advertising", 2012 IEEE 28th International Conference

on Data Engineering (ICDE).

[7] Michael Blaha and James Rumbaugh, “Object-Oriented

Modeling and Design with UML”, 2nd Edition, Pearson

Education, 2005, pp 21-157.

[8] Sommerville, “Software Engineering”, Eighth edition,

Pearson publication.

Sheshasaayee, Ananthi, and H. Jayamangala. "A Study on

Advertising." 2012 IEEE (ICDE).

Michael



40

Multilevel Encryption for Cloud Storage

Deepti Rai*, Roopa Desai, Tripti P S and Vinutha B

Department of Information Science and Engineering, Sahyadri College of Engineering & Management, Mangaluru – 575007


ABSTRACT

InformationCloud storage has easy access anytime, anyplace, anyhow due to its scalability, cost efficiency, and high reliability of the

data. Cloud computing uses internet for computing services. Organizations are moving their data to cloud. So we have to protect

uploaded data against unauthorized users from data access, modification etc. In this paper, a multilevel encryption and decryption for

cloud storage is proposed. Here a combination of AES and Rounded shift algorithm is used. Thus, only a valid user will access and

modify the data file. If an intruder takes the confidential data intentionally or accidentally, one must have had to decrypt the data for at

each level. So, there is less probability of getting original data. It is expected that using double level encryption and decryption will

provide more security of cloud storage that using one level for encryption and decryption.

Keywords: Cryptograph, Security algorithm, AES, Rounded shift, Symmetric, Asymmetric

1. INTRODUCTION

In this new era, Cloud computing provides large number of

services of internet. For cloud services allows user to utilize

software and hardware that will be managed by unauthorized

or invalid users. Cloud services mainly used for file storage,

webmail and business application. Security to the data that

resides in the cloud is provided by cryptographic algorithms.

By using cryptography original data called plain text is

converted into non readable form called cipher text. Existing

cryptographic algorithm uses single level encryption and

decryption so cyber criminals can easily break single level

encryption.

Hence we propose the system that consists of multilevel

encryption and decryption to provide security to cloud data. In

our proposed system we implement two algorithms in which

first level of encryption is done by Advanced Encryption

Standard (AES) and second level encryption is done by

Rounded Shift algorithm which is of Caesar Cipher type. AES

will process huge amount of data and that has high speed of

performing encryption and decryption which is more secure.

Caesar Cipher also known as shift cipher which consists of left

and right shifts. Here each bit of plain text is shifted in Caesar

box to a certain position using key. In our paper, we use

modified Caesar cipher for better security purpose. When user

uploads file it undergoes first level encryption using AES

algorithm and here plain text is converted into cipher text. This

scrambled form again undergoes second level encryption using

Rounded Shift algorithm and these encrypted data stored in

cloud database. When user wants to retrieve data from cloud

decryption is done in the reverse order of encryption. Thus

user gets the original data. In multilevel encryption it is

difficult to guess the key for intruder.

Architecture design shows the conceptual model of the

application. A graphical representation of concepts, their

principles, elements and components that are part of

architecture. The general architectural diagram of Multilevel

Encryption/Decryption for Cloud Storage is shown in Figure.

This design consists two level of encryption and decryption.

Initially when user uploads the file, it undergoes first level of

encryption using AES algorithm. During this original data is

converted to cipher text and this cipher text undergoes second

level of encryption using Rounded Shift algorithm. This

encrypted data is stored in cloud database. When the user

wants to download the file, the file is retrieved from cloud

database and it undergoes two level of decryption using

Rounded Shift and AES algorithm respectively. During each

level of decryption cipher text is converted to plain text. Thus

user gets original data file.

Figure 1: Architecture diagram of multilevel encryption / decryption for

cloud storage


41

2. LITERATURE SURVEY

Data security in the cloud will be increased for using RSA and

AES encryption and decryption algorithm. Here it uses key

size of 1024 bit and 128 bit, so attacker cannot determine

private key even if public keys are generated. The performance

can be analysed with the help of file size and computation cost.

This paper uses symmetric and asymmetric algorithm wherein

asymmetric public of key is used by encryption to and private

for key is for used by decryption. In symmetric one key is used

for either an encryption or a decryption. Security can be

enhanced using both symmetric and asymmetric algorithm.

Hash and signature algorithms used to compress data [1]. The

major issue related for cloud security of data integration. As a

solution Byzantine for fault-tolerant protocol across over

multiples of clouds is used. Another major concern is service

availability. It prevents loss of customer private data as about

result for malicious insiders out of clouds. This paper uses

cloud computing of model includes of five characteristics

features [2]. Security can be gained by applying cryptographic

for methods by enclosing data decryption key only of the

registered users. But here solution produces many computation

times over the data owner on key distribution with

management. This problem can be overcome by using

attribute-based encryption like proxy decryption and lazy

decryption. Data access control is developed by implementing

fine grained to access by control, with leads to edibility of

differential access with rights on individual users [5]. The

system architectures used concatenating of digitalized

signature algorithm of Diffie Hellman and AES as encryptions.

Block tag form of authentication is used to maintain data over

cloud storage. So there was need to remoting data integrity

along provides security regards of user data. The combination

with authentications techniques and key exchanged algorithms

is implemented and that leads to three way of mechanisms.

Here key distribution done in decentralized manner. Data

slicing was doing through data for fragmentation technique to

create segments of data. Datasets get slice onto three segments

with using vertical, horizontal of mixed fragmentation with

techniques. [6].

3. IMPLEMENTATION

Functional modules of multilevel encryption for cloud storage

are:

1. Registration

2. Login into the system

3. Upload files

4. My files

3.1. Registration

Here new user signup into the system by entering username,

mobile number, email, password which will be stored in the

database for further reference.

3.2. Login into the system

Registered user can login into system for the upload, download

and viewing the files which are stored in cloud.

3.3 Upload files

Here user chooses the file to be stored in the cloud. This file

will undergo two level of encryption using rounded shift and

AES algorithms before uploading it to cloud.

3.4. My files

Here get two options viz download and view. The file can be

searched by its uploaded date and with file title. During

download and view decryption is done in the reverse order of

encryption and OTP is sent to the authorized user to access the

file.

Here we are using two algorithms such as rounded shift

algorithm and advanced encryption standard.

1) Rounded shift algorithm: Rounded Shift algorithm is

Caesar-cipher type algorithm which uses the shifting of bits to

encrypt the plain text. In this paper, we have used nine cross

nine matrix which is further divided into nine blocks which is

of three cross three matrix.

In first step this algorithm shifts fixed number of blocks as it

defined in an algorithm. Later, in second step it shifts the bytes

within the selected block based on the original length of the

plain text. This generated cipher text is given as input to AES

algorithm which is further encrypted using random generated

keys. The decryption is done just by reverse order of

encryption.

2) Advanced encryption standard (AES): Most popularized

and with widely used symmetric of encryption algorithm is

advanced encryption standard. AES is much faster than DES.

The size of key used in DES is very small. It needed to be

replaced by an algorithm. The feature with AES are symmetric

of key, block cipher, 128 bit data, 128/192/256 bit keys. This

algorithm treats the 128 bits in a plane text as a block of 16

bytes. The 16 bytes are arranged in the form of a matrix

consisting of 4 rows and columns.

There are 4 steps in the encryption process.

1. Bytes substitution: Byte substitution is a step in which 16

bytes of input is substituted after looking up a fixed table

which gives a matrix of 4 rows and columns.

2. Shifting of rows: Shifting of rows includes shift of the

matrix’s rows to the left, whichever entry which falls of re-

inserted to the right of the row. The shifting is done randomly

picking each row and shifting towards the left with random

number of shifts. This step occurs in numerous number of

times.

3. Mixing of columns: Mixing of columns is done by

transforming each column which consists of 4 bytes using

some mathematical function. This function replaces the

original column of 4 bytes into completely new set of 4 bytes.

This gives with result upon another unique in matrix which

consists of 16 new with bytes.

4. Adding of round keys: Adding of round keys has 16 from

bytes of matrix which are considered with 128 bits and is OR,

forming a round key of same number of bits. This step is

performed many number of times to get the output as cipher

text. The decryption of cipher text is done by reversing the

order of encryption process.


42


Fig. 2. Encryption speed of AES algorithm

Figure 2 shows the encryption speed of the algorithm. If the

encryption speed is less then it gives better security. By using

rounded shift algorithm it’s difficult to predict encryption key.

The above graph deals with the encryption time. It compares

the AES algorithm with other encryption algorithms.

5. CONCLUSION AND FUTURE WORK

Multilevel Encryption for Cloud Storage provides security for

the confidential data. The operations like upload, download

and view are performed by registered user. Searching of file is

provided with the help of file title and uploaded dates. It

includes two levels of encryption by using Rounded shift

algorithm which uses shift key for encryption. Here plain text

is converted into cipher text of same length. Second level of

encryption is done through AES algorithm which uses

randomly generated key, it converts cipher text of rounded

shift into 124 bit cipher text. Decryption is done in the reverse

order of encryption. During view and download OTP is gen

erated randomly and sent to the registered mobile number to

authenticate whether valid user is logged in or not. Our

proposed project support only for text and word file. So future

enhancement can be encryption of images and pdf files.

REFERENCES

[1] Akashdeep Bharadwaj, GVB Subrahmanyam, Vinay

Avasthi, Hanumat Sastry, “Security Algorithms for Cloud

Computing”, ELSEVIER, vol. 85, pp. 535-542, 2016.

[2] Mohammed A. AlZain, Eric Pardede, Ben Soh, James A.

Thom, “Cloud computing Security: From Single to Multi-

Clouds”, 45th Hawaii International Conference on System

science, 2012.

[3] Eman M. Mohamed, Hatem S. Abdelkader, Sherif El-

Etridy, “Enhanced Data Security Model for Cloud

Computing”, The 8th International conference on

Informatics and systems, May 2014.

[4] Meiko Jensen, Jorg Schwenk, Nils Gruschka, Luigi Lo

Iacono, “On Technical Security Issues in Cloud

Computing”, IEEE International Conference on Cloud

Computing, pp. 109-116, 2009.

[5] Shucheng Yu, Cong Wang, Kui Ren, and Wenging Lou,

“Achieving Secure, Scalable, and Fine-grained Data

Access Control in Cloud Computing”, in Proc. of SP’02,

2009.

[6] Akhil Behl, Emerging Security Challenges in Cloud

Computing, “An insight to cloud security challenges and

their Mitigation”, pp. 217-222, 2011.

[7] Mr. Rupesh R Bobde, Prof. Amit Khaparde, Prof. Dr. M.

M. Raghuwanshi, “An Approach for Securing Data on

Cloud Using Data Slicing and Cryptography”, IEEE

sponsored 9th international conference on intelligent

system and control, 2015.



43

War Field Spy Robot

Megha N*, Pratham Shet, Salian Veena Suresh, Shashidhar, Susmitha K.

Department of Electronics and Comm. Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007


ABSTRACT

This paper presents a design scheme to develop robotic vehicle which provides surveillance in the war field. The spy robot

is very useful and capable of performing jobs in situations which are hazardous for humans. The robotic vehicle is attached

with a wireless camera which wirelessly transmits real time videos with night vision capabilities. This robot is used in

monitoring purpose in the war field by spying the activities taking place in the war field. The android application device

used at the transmitting end acts as a remote control to control the motion of the robot. The robot is capable of detecting

the bomb underneath and sending a message to android application device.

Keywords: Night vision wireless camera, Android application device, PIC microcontroller.

1. INTRODUCTION

Robots help in performing repetitive and dangerous task which

humans won’t consider to do. This project deals with the control

of robotic vehicle using android application. The robot can

move in required directions using android application and

captures the real time images and videos. The Microcontroller

is used to interface with Bluetooth Module, Camera, and

Proximity sensor. The Bluetooth Module is used for

transmitting the data to Android Application device. The aim of

this project is to develop an android controlled robotic vehicle

using Bluetooth Module. Robots can do jobs in areas where

human cannot go and in situations that pose risk to human life.

Small holes make human movement difficult inside it and hence

robots prove beneficial under such circumstances [1]. The

robotic vehicle moves according to the commands given by

android application. The motor drivers act as the wheels for

robot which is programmed using Microcontroller. Bluetooth

module is used to interfacing the robot with android device. The

camera mounted on the robot transmits real time happenings of

its surroundings. The path followed by the robot is being seen

in the camera by remote user, accordingly the video signals are

sent to the PC or laptop. Wired robots were developed in 1940’s

and were used by expertise trained for the same. A new class of

robots controlled by remote can be accessed now on the site:

the online robots. These allow users from all over the world to

museum, tend gardens, find way undersea, or handle crystals of

protein. The first generation of online robots came into

existence in 1994. In contrast, research on the second

generation of Internet robots has of late focuses on independent

mobile robots that steer in a forceful and unsure environment.

Remote controlled robot had problem in their range restriction

and also they were very high-priced in terms of safety and use

[1]. We are exploring how a robot can be controlled using

android application. The Secret Robot will be able to replace

human. It has camera used to capture the real time videos. The

robot vehicle work like military tank, moving forward, turning

left direction and turning towards right direction. Wireless

camera will send back the existent time videos of surrounding

happenings which can be seen on a distant monitor in the PC

where the robot is being inhibited and action can be taken

accordingly.

The research paper [1], proposed by Darshan Dayma,

Bhushan Chavan Authors presents a Bluetooth and cell phone

controlled robot. The system uses bluetooth module, four dc

motors, microcontroller, temperature sensor and humidity

sensor. PIC Microcontroller used is the heart of this structure

which is used to organize and process several functions based

on coding. The humidity sensor senses the humid environment

around the robot. Based on the numerical temperature value the

temperature sensor senses the weather changes. In April 2016

[2] Authors have developed a Smart spy robot system using RF

technology. The system consists of wireless camera, Bluetooth,

8051 microcontroller, RF technology. It is an RF based spying

robot attached with a wireless camera that reduced human

victim. Microcontroller is the main controller that decodes all

the instructions received from the transmitter unit. The

commands to control the motion of the robot is transmitted

through the Bluetooth module and the microcontroller decodes

to manage the movement of the robot. The wireless camera

transmits the real time videos to the android application device

at the receiving end. The wireless camera is mounted on the

robot. In [3] the Authors have developed a Bluetooth controlled

robot. An 8051 series microcontroller is used as control device

in the system. The robot design is controlled by the mobile

application. The project’s purpose is scheming a robot that can

be managed using Android cellular phone. The robot is capable

of programming again and can be used for multiple


44

applications. In the research paper [4], proposed by Tushar

Maheshwari Upendra Kumar presents operation of wireless spy

robots which controls large operating ranges. The camera is

based on WiFi technology which helps in appearing of

suurounding area live through mobile, laptop, etc.. In paper [5],

the Authors have designed a Bluetooth controlled automated

vehicle. The controller utilized here is an 8051 arrangement

microcontroller used to control the framework. The utilization

of Android cellular phone to control a mechanical vehicle is the

fundamental reason behind outlining this undertaking.

Bluetooth Module is implemented in the android device to

operate the wireless Robot. An advanced mobile phone with

Android application is utilized to control the mechanical

vehicle.

2. METHODOLGY

2.1 Block Diagram

Figure 1: Block diagram of proposed system

As shown in Fig. 1, robot is controlled using android based

gadget. The robot moves according to the commands given by

Android application i.e. front, back, left and right. The system

uses four DC motors which act as the wheels for the robot. The

DC motors are controlled using motor driver which is interfaced

with Microcontroller.

A supply of 12V DC is given to the regulator and the 5V DC

output is applied to each of the modules. The system uses a

Proximity sensor to determine the bomb underneath which are

interfaced to the Microcontroller and message will be

transferred to Android Application Device through Bluetooth

Module. The robot is controlled manually.

A Wi-Fi camera is used in this project. The advantage of

digital Wi-Fi camera systems is that they do not require a line-

of-sight connection between the transmitter and the receiver.

Bluetooth device is used to interface with PC, mobile phone.

It acts as a gateway between android device and the

microcontroller. The microcontroller used here is PIC16F877a.

The system uses HC-05 Bluetooth module. It operates at a

frequency of 2.5GHz. It is applied with a 3.3V DC supply. The

working temperature is from 20oC to75oC.

Android is user friendly and works effectively with all

applications. Android application can be used to control the

movement of robotic vehicle. The system uses a Sealed Lead-

acid rechargeable battery AP12-1.3 with 12V/1.3Ah supply, the

battery is mounted on the robotic chassis and the power is

supplied to the whole system through Microcontroller Board.

Since the battery is rechargeable, a Switch mode power adapter

with input voltage range AC100-240V - 50/60Hz 0.3A and

output voltage DC12V - 1A is used to recharge the battery.

2.2 Robotic Motion Control

The robot moves according to the commands given by Android

application i.e. front, back, left and right. The system uses four

DC motors which act as the wheels for the robot.L293D is a

motor driver circuit which is used to steer the DC motor in the

requisite direction. It has 16 pins which controls two set of DC

motor. This single IC can control two DC motors. It works on

the notion of H-bridge circuit which allows the voltage to flow

in both directions. It has four input pins. Input pins on left will

rotate the motor connected on the left side and input pins on

right rotate motor on the right side. Based on the inputs given

such as logic 0 or logic 1 to the input pins motors will rotate.

2.3 Data Transmission

Bluetooth module is used to interface robot with android device.

It acts as a gateway between android device and the

Microcontroller. The system uses HC-05 Bluetooth module.

HC-05 is a 6 pin module out of which only 4 pins are used to

interface with Microcontroller. The four pins include VCC

(Supply), GND (Ground), RXD (Receiver pin), and TXD

(Transmitter pin). The VCC is provided with 5V DC supply,

GND is grounded, RXD is connected to TX0 of Microcontroller

and TXD is connected to RX0 of the Microcontroller pins R6

and R7 are transmitter and receiver pins.

The receiver pin of Bluetooth is given to the transmitter pin

RC6, side transmitting pin of the Bluetooth is connected to the

receiver pin RC7, hence we establish a wireless network. The

transmitting pin of the Bluetooth is connected to the receiving

pin RC7 of the PIC.

3. FLOW CONTROL OF THE ROBOT

As shown in the fig.2, the motion control of the robot is shown.

Based on the commands received by the android application

device the motion of the robot is controlled. When L, R, F, B

are received the robot is moved left, right, front and backward

direction respectively. When S is received the robot is stopped.

In the figure.2, the motion control of the robot is shown.

Based on the commands received by the android application

device the motion of the robot is controlled. When L, R, F, B

are received the robot is moved left, right, front and backward

direction respectively. When S is received the robot is stopped.


45

Figure 2: Flow of control of the robot.

Figure 3 shows the proximity sensor sensing the bomb

underneath. Whenever a bomb is detected proximity sensor

provides a logic high signal to the RC0 pin of the PIC16F877a

and the controller sends message to the android application

device “Bomb is detected” through the Bluetooth module.

4. CONCLUSION

The primary need of our project would be accuracy. The robot

is able to move in required directions with help of the

commands given by the android application. The things

happening currently in the area can be clearly and accurately

seen to locate the spy robot as to where it is situated. The robot

is able to detect the bombs that are placed underneath in the war

field and notify the detection of bombs. By keeping the circuit

uncomplicated and effortless many users will be able to use it

without difficulty.

REFERENCES

[1] Darshan Dayma, Bhushan Chavan, Suraj Kale, Assoc.

Prof. B. S. Tarle “SMART SPY ROBOT” International

Journal of Science, Technology and Management, (IJSTM)

Volume 4, Issue 02, February-2015.

[2] Ankit Yadav, Anshul Tiwari, Divya Sharma

RatneshSrivatsava, Sachin Kumar, O.P. Yadav “SMART

SPY ROBOT” International Journal of Science,

Engineering and Technology Research (IJSETR) volume

5, Issue 4, April-2016..

[3] Arvind Kumar Saini,Garima Sharma, Kamal

KishorChoure, “BluBO: Bluetooth Controlled Robot”

International Journal of Science and Research (IJSR), April

2015

[4] Rowjatul Zannat Esita, Tanwy Barua, Arzon Barua, Anik

Mahammod Dip “Bluetooth Based Android Controlled

Robot” American Journal of Engineering Research

(AJER), Volume 5, Issue 3, 2016.

[5] Rahul Kumar, Ushapreethi P, Pravin R. Kubade,

Hrushikesh B. Kulkarni “Android Phone Controlled

Bluetooth Robot” IRJET Volume 3, Issue 04, April- 2016.

Figure 3. Flowchart of bomb detection



46

Adnan Aslam Hasan Haji Shaikh*, Prajwal M, Leandra Sonal, Kanishka Madan Gaonkar, Athar Ali

Department of Civil Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007


ABSTRACT

Precision farming is a field which is gaining more importance every passing day. It involves finding out the accurate top

soil constituents for farming application. The application of chemical and natural farming additives with a main purpose

of increasing and preserving the yield, but it will be prone to spatial variability across the farm. To obtain accurate values

sampling all across the field is not economic. Hence the need for interpolation to obtain accurate values. In this study we

have used four interpolation techniques to study the variation of Soil organic carbon (SOC) in a 3 acre farm. They were

namely Inverse Weighted Distance (IDW), Spline, Ordinary Krigging and Natural Neighbour. A total of 35 samples were

taken across the farm, 30% of this data set was used for validation and 70% of the data set was used for calibration. The

results obtained showed that IDW method had a deviation of less than 10% compared to the other methods. Hence out of

the four techniques IDW is most suited for SOC variability for precision agriculture applications.

Keywords: Soil organic carbon, Precision farming, Inverse weighted distance, Spatial variation, Interpolation, farming,

Ordinary krigging, Natural Neighbour, Spline.

1. INTRODUCTION

Soil is the basic element of all living beings on earth. The top

loose layer of the earth's surface, consisting of minerals and

rock particles blended with decomposed organic matter

(humus), and capable of holding water. Soils regulate

ecosystem services [1] and assume a noteworthy role in the

global system managing major biogeochemical cycles and

energy.

Soil is a standout amongst the most important of agricultural

production [2] and has dominant effect on crop yields and

quality [3]. In-field soil data has been utilized for quite a long

time by agriculturists to settle on choices concerning crop

management practices.

Topsoil (0 to 20cm) has the highest concentrations of

nutrients and microorganisms [4] and is the framework for most

of the earth’s soil biological activity. Soil properties are

neither static nor homogenous with space and time. Topsoil

has its major application in agriculture as plants obtain most of

the nutrients from it. Information on soil properties at finer

resolution are essential in many fields, more so in precision

agriculture [5].

When it comes to precision agriculture accurate and precise

values of the top soil constituents is a necessity. One such

important constituent is soil organic carbon (SOC). In the past

different geostatistical approaches have been used to estimate

the spatial distribution of SOC [6]. Sampling all across the field

is not economical and it a time intensive task. Geostatistics is

an efficient method [7] for the study of spatial allocation of

SOC content and its irregularities and reducing the variance of

assessment error and execution costs.

In this paper an attempt has been made to assess the

interpolation techniques to predict the variability of soil

organic carbon across a farm plot. The farm had lateritic soil.

The interpolation techniques used are Inverse Weighted

Distance (IDW) [8], Spline [9], Ordinary Krigging [10] and

Natural Neighbour [11]. The data set has been divided into

calibration dataset (70%) and validation dataset (30%) and the

accuracy of the results are compared.

2. AREA OF STUDY

The area chosen for this study is Saripalla situated in

Mangalore, Karnataka, India. The site is 3 acres in area. It is

an open flat surface that’s consisting of lateritic soil. There

used to be a rubber plantation before but now it is barren.

Sites for sampling were chosen in a gridded pattern of 25 m

X 25 m for each cell.

Figure 1: Study area

Accessing Spatial Variability of SOC Content

Using GIS Based Interpolation Techniques


47

3. METHODOLOGY

Figure 2: Methodology

The soil sample was collected from 35 sample sites. These

sites were initially marked on Google earth and with the help

of its coordinates were located. The sampling [12] was done by

digging into the ground up to a depth of 15 cm using pickaxe

and shovel. The sample was collected and stored in zip lock

bags to avoid contamination.

Figure 3: Sample collection

Soil organic carbon was tested using the standard muffle

furnace test [13]. The parameters used were 500 degree Celsius

for 30 minutes of heating. The carbon content for each sample

was noted.

Figure 4: SOC determination

The samples were calibrated into two data sets namely

calibration and validation data set. 30 samples were used for

calibration and 5 samples were used for validation for the

purpose of ground trothing.

The technique for estimation of unknown value between two

known value and deducing missing values from a set of known

values is called interpolation. Interpolation comes in use where

the values around the missing values are known and its

seasonality, repetition and long-term cycle is known.

On the calibration data set four types of interpolation

techniques are used namely Inverse distance weighed, Spline,

Natural neighbour and ordinary krigging.

Inverse distance weighted assumes that the values around a

specific unknown value is more likely to be similar than that

are further apart. That is the nearest values around unknown

value have most influence on the unknown value.

Spline keeps low regards for the curvature of the surface

and uses a mathematical equation [14] to assess the unknown

value. Basically the surface is assumed to be a smooth one

where the surface moves exactly over the input points.

The following equation (1) is used in for spline interpolation:

(1) Where: j= 1, 2…..N.

N is the number of points.

λj are coefficient found by the solution of the system of a

linear equations.

rj is the distance from the point (x,y) to the jth point.

T(x,y) and R(r) are defined differently, depending on the

selected option.

Ordinary Kriging is an advanced geostatistical method that

produces an estimated surface from a scattered set of points

with z-values [15]. It assumes that the distance or direction

between data points show a spatial correlation that can be used

to understand variation in the surface. The Kriging tool assigns


48

a mathematical function to a known number of points, or all

points within a known radius, to find the output value for each

location.

Kriging is similar to IDW in a way that it weights the

surrounding known values to predict values of unknown

location. The following equation (2) is used in for krigings

interpolation:

(2)

where:

Z(si) = the measured value at the ith location

λi = an unknown weight for the measured value at the ith

location8

s0 = the prediction location

N = the number of measured values

The technique used by Natural neighbour interpolation tool

searches the nearest subset of input values to a query point and

applies weights to them based on proportionate areas to

interpolate a value[16]. This method is also known as Sibson or

“area-stealing” interpolation. Its base identity is that its local,

using only a subset of samples that surround a query point, and

interpolated heights are guaranteed to be within the range of

the values used. It does not regard trends and will not produce

peaks, ridges, pits or valleys that are not already represented

by the input values. The surface passes through the input

values and is smooth throughout except at points of the input

value.

4. RESULTS

The results obtained for the interpolation techniques are

depicted in the figures below. In IDW method the power used

was 2. In ordinary krigging spherical semivariogram was used.

Figure 5: IDW output

Figure 6: Spline output

Figure 7: Krigging output

Figure 8: Natural neighbor


49

The results were tabulated using the validation data set which

was not used for calibrating the results. The average error was

found using the difference between the original and

interpolated SOC content. Root mean square deviation

(RMSD) represents the sample standard deviation of the

differences between predicted values and observed values. The

RMSD serves to aggregate the magnitudes of the errors in

predictions for various times into a single measure of

predictive power. RMSD is a measure of accuracy, to compare

forecasting errors of different models for a particular data and

not between datasets, as it is scale-dependent [17].

(3)

The Error and RMSD is tabulated in the table below

along with the percentage change with the actual value.

Interpolation

method

Error

(gms) RMSD

(gms) Percentage

change (%)

IDW

0.028 0.1392

9.87

Spline

0.028 0.1392

11.22

Krigging

0.036 0.1392

10.87

Natural

neighbour 0.028 0.1443

11.18

5. CONCLUSION

From the results, we can infer that there is very little change in

interpolation techniques used. IDW method gives the best

results in terms of RSMD and percentage change. Other than

IDW all the other methods give a percentage error greater than

10%. For a change of distance of 25 meters the interpolation

deviation of greater than 10% may not be acceptable for

precision farming applications.

This is an indication of interpolation taking only SOC

content into consideration. More studies have to be conducted

to assess the accuracy of interpolation for different top soil

constituents.

REFERENCES

[1] Dominati, Estelle, Murray Patterson, and Alec Mackay. "A

framework for classifying and quantifying the natural capital and

ecosystem services of soils." Ecological Economics 69.9 (2010):

1858-1868.

[2] Matson, Pamela A., et al. "Agricultural intensification and

ecosystem properties." Science 277.5325 (1997): 504-509.

[3] Cassman, Kenneth G. "Ecological intensification of cereal

production systems: yield potential, soil quality, and precision

agriculture." Proceedings of the National Academy of

Sciences 96.11 (1999): 5952-5959.

[4] Li, Feng-Min, et al. "Dynamics of soil microbial biomass C and

soil fertility in cropland mulched with plastic film in a semiarid

agro-ecosystem." Soil Biology and Biochemistry 36.11 (2004):

1893-1902.

[5] McBratney, Alex, et al. "Future directions of precision

agriculture." Precision agriculture 6.1 (2005): 7-23.

[6] Bhunia, Gouri Sankar, Pravat Kumar Shit, and Ramkrishna

Maiti. "Comparison of GIS-based interpolation methods for

spatial distribution of soil organic carbon (SOC)." Journal of the

Saudi Society of Agricultural Sciences (2016).

[7] Goovaerts, Pierre. "Geostatistics in soil science: state-of-the-art

and perspectives." Geoderma 89.1-2 (1999): 1-45.

[8] Lu, George Y., and David W. Wong. "An adaptive inverse-

distance weighting spatial interpolation technique." Computers

& geosciences 34.9 (2008): 1044-1055.

[9] Wahba, Grace. Spline models for observational data. Vol. 59.

Siam, 1990.

[10] Bhat, Vishwanatha, et al. "Spatiotemporal Relationship Linking

Land Use/Land Cover with Groundwater Level." Groundwater.

Springer, Singapore, 2018. 41-54.

[11] Boissonnat, Jean-Daniel, and Frédéric Cazals. "Smooth surface

reconstruction via natural neighbour interpolation of distance

functions." Computational Geometry 22.1-3 (2002): 185-203.

[12] Tan, Kim H. Soil sampling, preparation, and analysis. CRC

press, 2005.

[13] Dean Jr, Walter E. "Determination of carbonate and organic

matter in calcareous sediments and sedimentary rocks by loss on

ignition: comparison with other methods." Journal of

Sedimentary Research 44.1 (1974).

[14] Hengl, Tomislav, and Ian S. Evans. "Mathematical and digital

models of the land surface." Developments in soil science 33

(2009): 31-63.

[15] Childs,Colin. "Interpolating surfaces in ArcGIS spatial

analyst." ArcUser, July-September 3235 (2004): 569.

[16] Sibson, Robin. "A brief description of natural neighbour

interpolation." Interpreting multivariate data (1981).

[17] Hyndman, Rob J.; Koehler, Anne B. (2006). "Another look at

measures of forecast accuracy". International Journal of

Forecasting, 22(4):679688 doi:10.1016/j.ijforecast.2006.03.001



50

Study on Strength of Hybrid Concrete Beam

Sachin K S, Vinayak Patgar, C S Darshan, Murigeppa, Manohar K*

Department of Civil Engineering, Sahyadri College of Engineering & Management, Mangaluru-575007


ABSTRACT

The construction industry place a very important role in the economic development of a country. The main intention of

these materials is to improve the quality of the materials used in the concrete. Hybrid concrete beam is one of the method

of improving the quality of the materials. This hybrid beam consist beam consist of a special reinforcement called arch

reinforcement other than concrete and main reinforcement.

Keywords: Arch Reinforcement, Flexural strength, Hybrid Beam, Partial Beam.

1. INTRODUCTION

The basic of all the construction work depends up on a term

called Concrete. Concrete is the most consumed material in the

world. It is made up of three basic ingredients that is cement,

aggregates and water. And these materials get in to a liquid

which can be turned into almost any shape and later on gets in

to a hard rock. The materials used for concrete are easily

available. It does not include any complicated system. The

concrete structures can be formed in to any shapes, size and

height. Depending up on the mixture of cement, aggregate and

water we can get different grades of concrete mixtures.

Concrete is strong in compression and weak in tension, in

order to overcome this we are providing a material which can

withstand the tensile forces called reinforcements. Steel is

mainly used as the reinforcing material. With the use of steel

in concrete we can obtain a material which is strong in

compression and tension

1.1 Partial Beam

Beam is a horizontal structural member which takes the load

from the super structure right angle to its axis. Beams while

bending develop tension and compression zone which are

divide by the neutral axis. As we know that concrete is good in

compression and steel reinforcement is good in tension. Here

we are going to reduce the amount of cement content in the

tension zone that is we are going for two grades of concrete. A

high grade concrete in the compression zone that is above

neutral axis and low grade of concrete in the tension zone that

is below the neutral axis. These types of beams are called as

partial beams. The definition of partial beams says that using

two different grade of concrete or creating a hollow or vacuum

section in the tension zone of the beam, having flexural and

shear strength similar to homogenous solid beam

1.2 Arch Reinforcement

Flexural test is done in order to find the tensile strength of the

concrete. And it is a measure of calculating the bending of

beams or slabs. As load is applied on the beam the beam start

bending. In order to resist the bending nature of the beams we

are introducing a special type of reinforcement called arch

reinforcement. This special reinforcement tries to reduce the

bending nature of beams and helps to carry more loads due to

bending forces. This arch reinforcement can be applied in the

longer span of the beam where the bending action is more

visible. In this project the arch reinforcement is made with the

help of PVC pipe with an infill of cement mortar in which the

pipe is made to bend in to the shape of an arch.

2. METHODOLOGY

In this project we are casting 13 beams specimens of 6 varities.

The beam mould is of size 150mm *150mm *700mm. Table

2.1 represents the different type of beams and the type of

concrete grades used in the work.

Table 1: Types of beam with different grades of concrete

Sl.No. Type of Beam Grade of

Concrete

No. of

specimens

1 Conventional M25 3

2 Partial beam M25 & M15 3

3 Conventional with single arch M25 3

4 Partial beam with single arch M25 & M15 3

5 Conventional with double arch M25 3

6 Partial beam with double arch M25 & M15 3

2.1 Bar Bending

The reinforcement are made with 10mm and 8mm dia. Bars,

10mm dia. bars as main reinforcement & 8mm dia. bars as


51

vertical reinforcement (Shear reinforcement @ 100 mm c/c).

There are 2 bars at the top (Anchor bar) and 3 main bars at the

bottom (Tension reinforcement).

2.2 Materials

The OPC cement is used of 53 grade with a fine aggregate

confirming to IS standards, and coarse aggregate is used in

different percentage i.e., 60% of 10mm down size and 40% of

20mm down.

2.3 Placing of Reinforcement and Testing

The reinforcement is provided with 20 mm cover from all the

sides. And clover Blocks are used for the providing cover. The

arch reinforcement is made with a PVC pipe which has got an

infill of 1:3 cement mortars and it is bent in to the shape of an

arch. The arch reinforcement is supported with help of hook

of the main reinforcement which is present at the bottom.

Figure 1: Arch action in placing of reinforcement

After placing the reinforcement inside the mould the concrete

is filled inside the mould. After placing the concrete inside the

mould tamping is done with the help of tamping rod. Tamping

is done in order to remove the air voids inside. The specimen

is cured for 28 days. The beam specimen is tested for the

flexural strength. Two point loading is applied for the

specimens. The point of application of load is at l/3 distance

from the ends. 50mm from the end is given for the support.

Marking on the specimen is done as per the above

requirements. Beams specimen is kept on the loading frame.

Reading is noted with the help of proving ring.

Figure 2: Two point loading

Figure 3: Test Setup

Figure 4: Cracked beam

2.4 Proving Ring Reading Calculation

Proving ring consists of two scales one is main scale reading

and the other one is small scale reading. In the main scale,

there are 25 divisions and each division value is 0.2 mm. In the

small scale reading there are 100 divisions and the value of

each division is 0.002mm. When the load is applied if the

needle of the small scale rotates one cycle the deflection value

is taken as 0.2 mm which can be seen in the main scale. Least

count of the small scale: 1 division = 0.002mm by using this

values the deflection value of each specimens can be

calculated. For each crack the divisions on the proving ring

has to be noted down and the corresponding deflection has to

be calculated. To calculate load, if the needle covers 103

division in the small scale then the load value is taken to be 20

kN. Corresponding values has to be noted down.

Figure 5: Variation of strength for different type of beams


52

Figure 6: Deflection of conventional beam with 20mm

down aggregate

Figure 7: Deflection variation of partial beam

3. CONCLUSION

According to the test results, the conventional beam with

single arch and the partial beam with single arch takes load

which is almost near to that of the conventional beam. The

load at which cracks developed in the conventional and partial

beams having single arch reinforcements is almost near to that

of conventional beam. Since arch reinforcement helps to resist

the deflection load, these arch reinforced beams mainly fails

due to shear failure. By increasing the number of stirrups at the

supports and minimizing at the centre the shear failure can be

reduced and the beam will be capable to resist more loads.

REFERENCES

[1] M.A. Abeol Seoud and J.J. Myers, “Implementation of

Hybrid Composite Beam Bridges In Missouri, USA.” 11-

13 December 2013.

[2] John R. Hillman, “Product Application of A Hybrid

Composite Beam System”, The Idea Program

Transportation Research Board, National Research

Council, March 17, 2008.

[3] John R. Hillman, “Hybrid Composite Beam – Design and

Maintenance Manual”, The Missouri Department of

Transportation, August 27, 2012.

[4] Fathoni Usman, “Flexural Behavior of Hybrid Concrete

Beam”, The 3rd National Graduate Conference,

University Tenaga National, Putrajaya Campus, 8-9 April

2015.

[5] Prof. Pravin B Shindhe, Prof. Sangita V Pawar, Prof. V P

Kulkarni, “flexural behavior of hybrid fiber reinforced

concrete deep beam and effect of steel and polypropylene

fibre on Mechanical properties of concrete”, IJARSC,

Vol. No.4, Issue No.02, Feb.2015

Guidelines for Submission

• Authors are requested to provide full details for correspondence: postal

address, phone numbers and email address. (The email address of

corresponding authors will be published along with the article).

• Authorsarerequestedtopreparetheirsoftcopyversionintextformatsand not in

PDF version.

• Authorsarerequestedtofollowthemanuscripttemplatewhilepreparingthe

manuscript.

SIJR is devoted to the publication of original research work. The journal also

accepts review papers and scientific articles. SIJR welcomes the submission of

manuscripts that meet the journal standard which includes novelty, significance

and excellence. Every volume of the journal will have two issue in a year. All

articles published in the SIJR are peer-reviewed.

Contact for any query

or

Information on

SAHYADRI International Journal of Research

Email: [email protected]

Web: www.sijr.in

Sahyadri Journal of ResearchInternational

SAHYADRI International Journal of Research

PublisherManjunath Bhandary - President

Bhandary Foundation, Sahyadri Campus, Adyar, Mangaluru - 575 007

PrintersPrakash Offset Printers, 164, Industrial Area, Baikampady, Mangaluru - 575 011

| Vol 4 | Issue 1 | June 2018

Sahyadri Campus, Adyar, Mangaluru - 575 007

COLLEGE OF ENGINEERING & MANAGEMENTSAHYADRI

(Affiliated to VTU, Belagavi and Approved by AICTE, New Delhi)

Empowering Young Minds

+ 91-824-2277222 | [email protected] | www.sijr.in | www.sahyadri.edu.inSAHYADRI Journal of Research | December 2015 | 3

Sahyadri College of Engineering & Management is in existence since a decade, being recognized by AICTE,

Government of India, affiliated to VTU, Government of Karnataka, NAAC accredited with “A” Grade and

certified by ISO.

The Sahyadri college has Academic MoU with various National, International Universities and also has

collaborations with leading Corporate & Industries that have given a tremendous boost to the students for

innovation, incubation, internships, research, projects and hands on experience.

Sahyadri encourages students to “Walk-in with an Idea and Walk-out with the Product”. The College aims at

imparting Project based learning, thus enabling students to understand the process of learning concepts.

Students right from day one are offered an opportunity to take part in the events held at IIT’s, NIT’s and SAE

events, thus improving their analytical skills, connectivity and gain exposure across the globe. The total

student and staff strength at campus is nearly 4000.

Journal Research - SIJRsijr.in/articles/june2018/Sahyadri-Journal-Research-Vol-4-Issue-1.pdf · Dr....

Documents

Transcript of Journal Research - SIJRsijr.in/articles/june2018/Sahyadri-Journal-Research-Vol-4-Issue-1.pdf · Dr....