Online monitoring of pore distribution in microporous membrane

8
Online monitoring of pore distribution in microporous membrane Ming Chang*, Juti Rani Deka, Chin Tao Tszeng, Pei Rung Cheng Department of Mechanical Engineering and R&D Center for Membrane Technology, Chung Yuan Christian University, Chungli 320, Taiwan Tel. þ886(3)2654323; Fax þ886(3)265439; email: [email protected] Received 27 July 2007; accepted revised 26 September 2007 Abstract The present investigation describes about the development of a cost effective system for real time monitoring of porosity of ion track etched membranes. The system uses a complementary metal oxide semiconductor (CMOS) image sensor to capture the image of the membranes from the optical microscope’s projector lens and a field-programmable gate array (FPGA) chip to process the captured image and transfer the pore distribution result to a LCD display. The porosity of the membrane is measured directly from the distribution of pores in the membrane. In the image processing by FPGA the pixels intensity greater than the threshold value is considered to be due to the presence of pores. This technique cannot be utilized in membranes with pore size less than 1 micron due to the limitation of magnification of the optical microscope. The porosities of three gamma ray irradiated thin track etched polycarbonate membranes of pore sizes 5, 8 and 10 mm are determined with the system and calcu- lated to be in the range 8.1–10.5, 2.0–9.8 and 4.9–9.6% respectively. The system is designed at minimum cost and provides more detailed data at a faster rate. Evaluation of the performance characteristics of the setup indicates that its performance is comparable with much costlier systems. Keywords: Polycarbonate membrane; Porosity; FPGA; CMOS image sensor 1. Introduction Porous membranes are produced by irradiat- ing polymer films with energetic heavy ions and subsequent etching of the ion tracks [1–3]. Polycarbonate (PC) membrane is produced by exposing microporous PC film to collimated charged particles from a nuclear pile. These charged particles leave sensitized tracks when it passes through the film. The polymer tracks are dissolved in an etching solution to form pores [4]. The pore size and density can be controlled *Corresponding author. Presented at the Fourth Conference of Aseanian Membrane Society (AMS 4), 16–18 August 2007, Taipei, Taiwan. 0011-9164/08/$– See front matter # 2008 Elsevier B.V. All rights reserved. Desalination 234 (2008) 66–73 doi:10.1016/j.desal.2007.09.071

Transcript of Online monitoring of pore distribution in microporous membrane

Online monitoring of pore distribution

in microporous membrane

Ming Chang*, Juti Rani Deka, Chin Tao Tszeng, Pei Rung Cheng

Department of Mechanical Engineering and R&D Center for Membrane Technology,

Chung Yuan Christian University, Chungli 320, Taiwan

Tel. þ886(3)2654323; Fax þ886(3)265439; email: [email protected]

Received 27 July 2007; accepted revised 26 September 2007

Abstract

The present investigation describes about the development of a cost effective system for real time monitoringof porosity of ion track etched membranes. The system uses a complementary metal oxide semiconductor(CMOS) image sensor to capture the image of the membranes from the optical microscope’s projector lens anda field-programmable gate array (FPGA) chip to process the captured image and transfer the pore distributionresult to a LCD display. The porosity of the membrane is measured directly from the distribution of pores in themembrane. In the image processing by FPGA the pixels intensity greater than the threshold value is considered tobe due to the presence of pores. This technique cannot be utilized in membranes with pore size less than 1 microndue to the limitation of magnification of the optical microscope. The porosities of three gamma ray irradiated thintrack etched polycarbonate membranes of pore sizes 5, 8 and 10 mm are determined with the system and calcu-lated to be in the range 8.1–10.5, 2.0–9.8 and 4.9–9.6% respectively. The system is designed at minimum cost andprovides more detailed data at a faster rate. Evaluation of the performance characteristics of the setup indicatesthat its performance is comparable with much costlier systems.

Keywords: Polycarbonate membrane; Porosity; FPGA; CMOS image sensor

1. Introduction

Porous membranes are produced by irradiat-

ing polymer films with energetic heavy ions

and subsequent etching of the ion tracks [1–3].

Polycarbonate (PC) membrane is produced by

exposing microporous PC film to collimated

charged particles from a nuclear pile. These

charged particles leave sensitized tracks when it

passes through the film. The polymer tracks are

dissolved in an etching solution to form pores

[4]. The pore size and density can be controlled*Corresponding author.

Presented at the Fourth Conference of Aseanian Membrane Society (AMS 4), 16–18 August 2007, Taipei,

Taiwan.

0011-9164/08/$– See front matter # 2008 Elsevier B.V. All rights reserved.

Desalination 234 (2008) 66–73

doi:10.1016/j.desal.2007.09.071

precisely by varying the temperature, strength of

the etching solution and the exposure time. All

particles larger than the pore size are captured

on its surface making it ideal for collecting

samples for blood test or for high-purity and gen-

eral filtration. The porosity of a membrane is func-

tion of pore density and pore size. Several methods

are available to study the physical properties of

membranes such as porosity, surface area, pores

size distribution and pore structure [5,6].

In general, the surface morphology of mem-

brane is studied with optical microscope, atomic

force microscope (AFM) and scanning electron

microscope (SEM) [7,8]. The present study

describes about the development of an experi-

mental system for real time monitoring of

surface morphology and porosity of track etched

PC membrane. The measurement system consists

of an optical microscope with broad spectral

width light source, a complementary metal

oxide semiconductor (CMOS) image sensor, a

field-programmable gate array (FPGA) and one

LCD monitor. The traditional charge coupled

device (CCD) camera employed to view the

image of the object directly is replaced by inte-

grating a CMOS image sensor with FPGA for

real time monitoring.

In order to determine the pore distribution in

PC membrane, it is required to capture the image

of the membrane by CMOS image sensor

initially from the projector lens of the optical

microscope. The optical microscope is used as

the prime apparatus of the measurement system

as it is cheaper and smaller in size in comparison

to other surface profile analyzing systems. In

this measurement system CMOS image sensor

capture the image and FPGA chip process on the

captured image and send the pore distribution

image to the LCD display. The CMOS image

sensor receives the superimposed light and con-

verted the optical signal to digital signal. FPGA

picked up this signal and allow the images to be

displayed on a high resolution LCD display.

Threshold image processing method is applied

to determine the threshold pixel value for pores

and background. From this digitized output, pore

density and porosity of PC membrane are esti-

mated from the image enhancement of unevenly

distributed gray scale values of image pixels.

2. Experimental

2.1. Materials

In the present investigation three different PC

membranes made by Millipore with pore dia-

meters 5, 8 and 10 mm called as C5, C8 and

C10 respectively are used. These membranes are

made up of thin sheet of PC crossed by almost

cylindrical parallel pores. The pores of the mem-

branes are created by irradiation of beam of heavy

ions on the raw PC sheets. These high speed ions

create tracks of molecular damage in the mem-

brane. By chemical etching and precession con-

trol of ultraviolet exposure pores are generated

along the tracks of that molecular damage.

2.2. Design of experimental system

Speed is a very important attribute for real

time application. Cost effective system on chip

(SOC) design approach based on FPGA technol-

ogy is used in the development of the experi-

mental setup for rapid data acquisition. The

experimental system consists of an optical

microscope, a CMOS image sensor, a FPGA

chip and a LCD display unit. The FPGA chip

processed on the CMOS image sensor captured

image and send the result to the LCD display.

Fig. 1 shows the schematic diagram of the

experimental system. The details of the design

consideration is mentioned in the following

subsections.

2.2.1. Optical microscope

The optical microscope is a type of micro-

scope which magnifies the image of small object

M. Chang et al. / Desalination 234 (2008) 66–73 67

image by sending a beam of light through the

object. The condenser lens of the microscope

shown in Fig. 1 focuses the light on the object

and objective lens magnifies it to the projector

lens so the image can be viewed by the observer.

The resolving power and resolution of the

microscope used in the present investigation is

determined to be 0.35 and 0.17 mm respectively

for monochromatic light source. As the image of

the pores of the membrane with few numbers of

pixels cannot be effectively recognized by an

optical inspection technique, this system can not

be utilized in membranes with pore size less than

1 mm. Accordingly, the pore density and porosity

of the PC membranes having pore sizes 5, 8 and

10 mm were investigated.

2.2.2. CMOS image sensor

Complementary metal oxide semiconductor

image sensor has been used in this investigation

to capture the image of the membrane. CMOS

image sensors consist of an integrated circuit

containing an array of pixel sensors, each array

containing a photo-detector connected to an

active transistor reset and readout circuit. Such

sensor is inexpensive and provides an accurate

digital image that can be easily stored by the

sensor’s interface unit.

TRDB_DC2 (DC2) board with one CMOS

sensor is used to capture the image from the

projector lens of the microscope. DC2 provides

frame grabber, high performance multiport

SDRAM frame buffer and image processing IPs.

The image sensor can produce an image of

(1280 pixel � 1024 pixel) size of 8-bit. The

size of each pixel is (5 mm � 5 mm). The buffer

stores every captured pixel and controls all

sensor signals. The image sensor is supported

with motion capture mode and illumination

can be controlled according to the light of the

surrounding area.

2.2.3. Field-programmable gate array

(FPGA)

Field-programmable gate array is an integrated

circuit that contains many identical logic cells

which can be viewed as standard components.

The FPGA integrated processing circuit consists

of four units, namely, image capture module,

image data transformation module, memory

control module and image processing module

[9]. Altera Cyclone II 2C35 FPGA is used in the

present investigation for image processing. This

FPGA consists of 35,000 logic elements (LE’s)

with 475 user IOs, 105 M4K RAM Blocks and

483 Kbit SRAM and with 35 embedded multi-

pliers. Fig. 2 shows the block diagram of the

FPGA integrated circuit along with CMOS image

sensor and display unit. The transmission of sig-

nal inside the core block represents transmission

of signals within the FPGA.

The master clock need by the CMOS image

sensor to start up is provided by the CMOS

sensor data capture block. The CMOS sensor

Fig. 1. Schematic diagram of the real time monitoring

system.

68 M. Chang et al. / Desalination 234 (2008) 66–73

transmits real time signals e.g., valid frame size,

valid line and pixel clock to the CMOS sensor

data capture block which sends the signal to next

block called Bayer colour pattern data to 30-bit

RGB block. The 12C bus controller in the 12C

sensor configuration block controls intensity and

frames per second of the image sensor. i.e., the

settings information about sensor exposure time,

frame size, filter selection etc. are stored in this

block corresponding to addressing register of

image sensor through look up table (LUT). Line

buffer and pipeline operates on the raw data

received by the Bayer colour pattern data to

30-bit RGB block and convert the process data

into standard 30-bit RGB data making it conve-

nient for image processing and display. The

SDRAM controller of multi-port SDRAM con-

troller block can simulate four data ports and can

save these 30-bit RGB data in SDRAM of 16 bit

data. These RGB data are simultaneously trans-

mitted to the SDRAM of next block which is a

complete frame buffer but has only two data

ports. VGA Controller and data request block

generate (640 pixel � 480 pixel) VGA signal

to display on VGA display.

2.3. Membrane’s image capture

The CMOS sensor lens has been focused on

the area where porosity is to be measured. By

moving the optical microscope continuously

over the sample surface along both X and Y axis

and adjusting the focus pore distribution in the

membrane is imaged. The frame grabber cap-

tures the image and save those image pixels in

the SDRAM frame buffer. The maximum clock

rate to run the FPGA on DE2 board is 50 MHz

and can achieve about 30 frames per second

under this clock rate. Although the sensor provides

30 frames per second, only one of these is

needed by system for porosity measurement at

a particular position. The image processing unit

read one image and store in the frame buffer.

The frame buffer size is 8 Mb. When a row of

pixels is completely captured and stored in the

frame buffer, the frame grabber will capture the

next row and store in the SDRAM buffer after

the first row. This process will continue until the

last row of the frame is stored. When one full

frame is stored, the image raw data is sent from

TRDB_DC2 to the Altera Cyclone II 2C35

Fig. 2. Block diagram of the FPGA integrated circuit along with COMS image sensor and display unit.

M. Chang et al. / Desalination 234 (2008) 66–73 69

FPGA on DE2 board through IDE cable. The

FPGA on the DE2 board is handling image

processing part and converts the data to RGB

format to display on the VGA monitor. Fig. 3

shows the pore distribution in PC membrane

having pore size 10 mm in (0.7 � 0.7) mm2 area.

The complete photograph of experimental setup

used to investigate the membranes porosity is

shown in Fig. 4.

2.4. Image processing

The image of the membrane is scanned for

locating pores in it. The colour image is then

converted to grey scale image as it is computa-

tionally inexpensive and sufficient for investiga-

tion of pore density and porosity of membrane.

Image thresholding method is used for image

processing and IP unit will determine the thresh-

old value. There are different image thresholding

techniques for different applications [10,11].

A simple method called threshold is used to get

the pores distribution as it is fast and easy to

implement. This image processing technique is

based upon the threshold value for converting

a greyscale or colour image to a binary image.

The pore density is determined from the inten-

sity variation of the pores from that of the back-

ground intensity. The threshold value of the

image is determined using histogram threshold-

ing where the histogram consists of ratio between

a pixel intensity value to the total number of

pixels in an image in one axis and the pixel value

in the another axis.

If a pixel in the image has an intensity value

less than the threshold value, the corresponding

pixel in the resultant image is set to black.

Otherwise, if the pixel intensity value is greater

than or equal to the threshold intensity, the

resulting pixel is set to white. i.e., if T is the

threshold value for converting the grey scale

image to digital image then,

gði; jÞ ¼ 1 for f ði; jÞ � T

¼ 0 for f ði; jÞ < T

where f (i, j) is the pixel intensity value and

g(i, j) is the resulting binary value. The threshold

value in this study is determined as 27.

The processing unit is allowed to scan the par-

ticular area. The average pixel value of the

image background is different from that of the

pores. The pixel values of pores are greater than

the threshold value and hence are white. Thus,

Fig. 3. Pore distribution having pore size 10 mm in

(0.7 � 0.7) mm2 area.

Fig. 4. Photograph of image capturing system.

70 M. Chang et al. / Desalination 234 (2008) 66–73

numbers of pores present in a particular area or

pore density can be determined from this thresh-

old image processing algorithm by counting the

number of pixels whose pixel value is greater

than the threshold value. The porosity (�) of the

membrane at that particular area is evaluated

from the pixels ratio given by,

� ¼ No: of pixels present in pores

Total pixels in the image

3. Results and discussion

The pore densities and porosities at different

locations of three ion track etched PC membranes

are measured with the real time monitoring

system mentioned above. It is already men-

tioned that the porosity at a particular location

determine from the pixel ratio.

It is observed from Fig. 5 that the porosity is

not exactly same at every measured locations of

a membrane. The porosity is measured to be

approximately 10% at most of the observed

locations of the membrane with pore size

5 mm. The almost uniform distribution of pores

in the membrane makes it uniform porous. But

small deviation of result e.g., 8.1% or 10.9%

are observed at some locations of the same

membrane which is due to the larger number

of pores at those places, as shown in the area

within the box of Fig. 6. Moreover, in some

cases the neighbouring pores coalesce which

may be due to temperature variation, strength

of etching solution or exposure time and in that

case the pore size is more than 5 mm. This may

results in deviation of porosity from that of the

average value of the membrane. The variations

of porosities from the average value of the

other two membranes at different locations may

also due to uneven distributions of pores. The

extremely low porosity value of 2% at the fifth

position of the membrane (0.049–0.059 position

in Fig. 5) may be due to scanning of area of the

membrane where very few numbers of pores are

present as shown within the box of Fig. 7.

The accuracy of the technique has been

studied by comparing the experimental results

with the data provided by Millipore, the manu-

facturing company. Millipore used bubble point

method to determine the porosities of the

membranes. The results obtained with the real

0

2

4

6

8

10

12

0.09

9–0.

010

0.08

9–0.

099

0.07

9–0.

089

0.06

9–0.

079

0.05

9–0.

069

0.04

9–0.

059

0.03

9–0.

049

0.02

9–0.

039

0.01

9–0.

029

0.00

9–0.

019

Area (mm2)

Poro

sity

(%

)

5 µm pores membrane8 µm pores membrane10 µm pores membrane

Fig. 5. Porosity of three membranes at different

locations.

Fig. 6. Pore density of pore size 5 mm in an area of PC

membrane.

M. Chang et al. / Desalination 234 (2008) 66–73 71

time monitoring and provided by Millipore are

tabulated in Table 1. It is observed from Table 1

that the porosity measured with the above real

time technique is almost within the range of

data provided by Millipore with small variations

at some places. This variation may be due to

scanning of some area containing more or less

number of pores than the average number as

described above. Additionally, the number of

pixel is always integer and hence the system can

calculate only the integer number of pixels from

those coalesce pores and unable to acquire the

data if there exist some fractions of a pixel.

As the technique uses digital image proces-

sing method, excellent repeatability can be

achieved with the system. The repeatability of

the results is examined by measuring the poros-

ity of the membrane with 5 mm pores by scan-

ning the same locations at three different

times. The repeatability plot as shown in Fig. 8

illustrates excellent repeatability of porosity

value with standard deviation of less than 4%.

As the technique is a non destructive as well

as real time monitoring one, the porosity of the

whole membrane can be determined accurately

and precisely. The measurement accuracy and

excellent repeatability using this system makes

it as an efficient technique for instantaneous

measurement of porosity of membrane.

4. Conclusions

The methodology presented here is capable

of achieving real-time monitoring of porosity

of membrane. The detection of pixel value more

than the threshold value in the area where pores

are presented indicates the existence of the pores

Fig. 7. Pore density of size 8 mm in an area of PC

membrane.

Table 1

Porosity of membranes determined with real-time

system

Membrane C5 C8 C10

Diameter (mm) 5 8 10

Pore density

(108 pores/m2)

42.9 10.5 7.9

Porosity (%) 8.1–10.5 2.0–9.8 4.9–9.6aPorosity (%) 5–10 5–10 5–10

aData provided by Millipore.

8

9

10

11

12

13

0.00

9–0.

010

0.08

9–0.

099

0.07

9–0.

089

0.06

9–0.

079

0.05

9–0.

069

0.04

9–0.

059

0.03

9–0.

049

0.02

9–0.

039

0.01

9–0.

029

0.00

9–0.

019

Poro

sity

(%

)

Area (mm2)

1st time2nd time3rd time

Fig. 8. Repeatability of porosity of PC membrane of

pore size 5 mm.

72 M. Chang et al. / Desalination 234 (2008) 66–73

which make the technique as acceptable for pore

distribution investigation. The SOC design and

integration of CMOS and FPGA technology to

the optical microscope reduces operation time

and makes the system faster for result analysis.

This faster result analyzing ability is an added

advantage of using this technique in comparison

to other techniques. The low cost of the CMOS

image sensor and use of FPGA chip reduces the

cost of the experimental system quiet low in

comparison to the system containing CCD cam-

era. As the system processing core and the other

required hardware are implemented on one board

there is no need of computer processing unit

(CPU) for experiment. Since it is a direct method,

so porosity of the membranes can be computed

more precisely and accurately with less time

without damaging the surface microstructure.

The same area can be viewed after several etch-

ing which is an added advantage of this non

destructive real-time monitoring system.

Acknowledgements

The authors gratefully acknowledge the

support of the Center-of-Excellence Program

on Membrane Technology, the Ministry of

Education (project number 956049303), Taiwan,

The Republic of China.

References

[1] R. Spohr, Nuclear track research activities at GSI.

Nucl. Instrum. Meth., 173 (1980) 229–236.

[2] S. Metz, C. Trautmann, A. Bertsch and P. Renaud,

Polyimide microfluidic devices with integrated

nanoporous filtration areas manufactured by micro-

machining and ion track technology, J. Micromech.

Microeng., 14 (2004) 324–331.

[3] M. Yoshida, N. Nagaoka, M. Asano, H. Omichi,

H. Kubota, K. Ogura, J. Vetter, R. Spohr and

R. Katakai, Reversible on-off switch function of

ion-track pores for thermo-responsive films based

on copolymers consisting of diethyleneglycol-

bis-allylcarbonate and acryloyl-l-proline methyl

ester, Nucl. Instrum. Meth. B, 122 (1997) 39–44.

[4] E. Faerain and R. Legras, Track-etch templates

designed for micro- and nanofabrication, Nucl.

Instrum. Meth. B, 208 (2003) 115–122.

[5] M. Mulder, Basic principles of membrane technol-

ogy, Kluwer, Dordrecht, 1991.

[6] S. Lowell and J.E. Shields, Powder surface area

and porosity, In: B. Scarlett (Ed.), Powder technology

series, Wiley, New York, 1984.

[7] M. Di Luccio, R. Nobrega and C.P. Borges, Micro-

porous anisotropic phase inversion membranes

from bisphenol-A polycarbonate: study of a ternary

system, Polymer, 41 (2000) 4309–4315.

[8] C. Torras, F. Ferrando, J. Paltakari and R. Garcia-

Valls, Performance, morphology and tensile

characterization of activated carbon composite

membranes for the synthesis of enzyme membrane

reactors, J. Membr. Sci., 282 (2006) 149–161.

[9] C.J. Chang, P.Y. Hsiao and Z.Y. Huang, Integrated

operation of image capturing and processing in

FPGA, IJCSNS International Journal of Computer

Science and Network Security, 6 (2006) 173–180.

[10] N.P. Pal and S.K. Pal, A review on image segmen-

tation techniques, pattern recognition, 26 (1993)

1277–1294.

[11] P.K. Sahoo, S. Soltani and A.K.C. Wong, A survey

of thresholding techniques, Comput. Vision Graph

Image Process, 41 (1988) 233–260.

M. Chang et al. / Desalination 234 (2008) 66–73 73