MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

download MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

of 102

Transcript of MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    1/102

    1

    Poznan University of Technology

    Faculty of Civil and Environmental Engineering

    Masters Thesis

    Pawe Mamrak

    Pressure evolution inside complex corridor

    system induced by blast action

    Supervisors: Marcin Wierszycki, PhD,

    Piotr Sielicki, MSc.

    Pozna 2011

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    2/102

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    3/102

    3

    Contents

    Contents ............................................................................................ 3

    1. Introduction ............................................................................. 52. Phenomenon of explosion ...................................................... 7

    2.1. Definition and classification of explosives ..........................................72.2. Blast phenomenon ...............................................................................7

    2.3.

    Blast-loading classification ................................................................ 12

    2.4. Calculation methods of blast-loading ............................................... 193. Finite Element Analysis techniques ..................................... 224. Tools and methods .................................................................23

    4.1. Software .............................................................................................. 23

    4.2.

    Creation of CEL model in Abaqus .................................................... 26

    4.3. Assumptions and restrictions in the script ...................................... 294.4. Simplified block scheme of the script ...............................................304.5. Structure of the script ........................................................................ 344.6. Problems and difficulties .................................................................. 404.7. Further development ......................................................................... 41

    5. Analyses ................................................................................. 425.1. Benchmarking analysis of blast wave propagation ......................... 42

    5.1.1. Description .......................................................................................................... 425.1.2. Analytical solution .............................................................................................. 435.1.3. Numerical solution ............................................................................................. 435.1.4. Discussion of the results .................................................................................... 44

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    4/102

    4

    5.2. Analysis of the cubicle ...................................................................... 465.2.1. Description .......................................................................................................... 465.2.2. Analytical solution .............................................................................................. 475.2.3. Numerical solution ............................................................................................. 565.2.4. Discussion of the results .................................................................................... 60

    5.3. Shock wave propagation analysis in corridor .................................. 655.3.1. Description .......................................................................................................... 655.3.2. Numerical solution ............................................................................................. 65

    5.4. Shock wave propagation analysis in tunnel shelter ........................ 745.4.1.

    Description .......................................................................................................... 74

    5.4.2. Numerical solution ............................................................................................. 755.4.3. Discussion of the results .................................................................................... 86

    6. Summary ................................................................................ 88Bibliography .................................................................................... 91

    Appendix A .................................................................................... 93Appendix B ................................................................................... 102

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    5/102

    5

    1. IntroductionPreface

    In recent years one can observe fast development of design methods in the field of civil

    engineering. The threat of terrorism (but not only) pointed peoples attention to the

    problem of resistance of the buildings to different, extreme kinds of accidental loading.

    Among them the most dangerous seem to be explosions, which can seriously damage or

    even destroy the structure of the object, forcing it to collapse. Also buildings, where

    explosive materials are manufactured and stored should be evaluated against structure's

    resistance to the effects of accidental explosions, as concentration of explosives increases

    this possibility.

    Nowadays computers are sufficiently powerful to successfully deal with these topics.

    It is possible to simulate the explosion taking place in a given space, as well as behaviour of

    the structure subjected to dynamic loading induced by the explosions (mainly air pressure).

    Both analyses are relatively new issue; therefore they are tersely described in national

    standards. In case of blast simulations the principal interest is value of loading subjecting

    the building. Many engineering methods have been introduced in the past, computing the

    load on a basis of weight of the explosive and distance to the structure. Many applications,

    e.g. ConWep for years has been serving people as useful tool to predict the influence of the

    blast wave. This approach, although widely used all over the world, e.g. to estimatepressure acting on a structure of tall buildings, is sometimes insufficient, as it does not take

    into account individual circumstances.

    In this thesis author presents a method of blast wave propagation analysis in Abaqus

    Explicit. In order to perform this kind of analyses, so called coupled Eulerian-Lagrangian

    analysis has been applied. The method allows detailed simulation of blast phenomena,

    requires however slightly different approach in modelling elements comparing to standard,

    Lagrangian analysis technique. The differences are described in next chapters.

    In order to automate the modelling process a special script has been prepared.

    Objectives

    Within the thesis four groups of simulations are conducted. In the first two, one seeks for

    appropriate mesh size on the basis of comparison between the analytical and numerical

    results. The parameters describing the shock wave are studied. In third simulation, the

    main objective is to obtain information on propagation of shock wave for different

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    6/102

    6

    conditions (e.g. arrangement and dimensions of corridors, location of explosives). The

    fourth simulation concerns analysing explosion in tunnel shelter. The results are compared

    to values obtained in experiments by Ishikawa and Beppu [27].

    Structure of the thesis

    The thesis consists of six chapters, followed by bibliography and one appendix. In the

    Chapter 1 the motivation and objectives are briefly discussed. Chapter 2 presents the

    theoretical background of explosion phenomenon. The types of explosions are described,

    followed by explanation of the crucial shock wave parameters. At the end, the historical

    and the current calculation methods of blast wave parameters are briefly described.

    Chapter 3 outlines the basic kinematical descriptions of the continuum that are used whiledealing with mechanics of fluids and solids. These issues are described in more detail in

    Appendix A. In the Chapter 4 the tools that were used for computations are described, that

    is: Abaqus/Explicit, Abaqus Scripting Interface, Python programming language. It is

    followed by a short overview of creating CEL model in the mentioned tools. Later, the

    script that was prepared for the purpose of easy and fast modelling of complex corridor

    systems is described. The issues such as assumptions that have been made in the script, its

    structure, as well as problems that have been encountered are discussed. Lastly authors

    thoughts on further development of the script are given. Chapter 5 concerns the numericalsimulations that have been conducted. The first simulation is a comparison of results

    obtained using Unified Facilities Criteria (UFC) methods [7] and Abaqus/Explicit. The

    second analysis deals with problem of blast wave propagation in a simple cubicle, where

    numerical results are compared to those calculated with use of analytical algorithms. The

    third simulation concerns behaviour of blast wave reflecting in a corridor broken by some

    angle. The fourth simulation is a numerical reconstruction of experiments performed by

    Japanese scientists in a tunnel shelter [27]. Chapter 6 is an ultimate summary of the

    information gathered in previous chapter. Appendix A discusses different Finite Element

    techniques, presents derivation of important equations of Eulerian, Lagrangian, Arbitrary

    Lagrangian-Eulerian and Coupled Eulerian-Lagrangian descriptions of continuum. Later

    the universal conservation laws in the non-conservative and conservative forms are

    provided. Appendix B contains UML class diagram of Python script.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    7/102

    7

    2.Phenomenon of explosion2.1. Definition and classification of explosivesExplosion is a rapid increase in volume and release of energy in an extreme manner [4].

    One can distinguish two types of explosive materials depending on the speed of

    propagation of the shock wave, namely "high explosives", if the detonation is supersonic

    and low explosives when the burning process is slower. Then this process is known as

    deflagration.

    Explosives, according to physical state can be classified as solids, liquids or gases.

    Solid explosives are mainly high explosives for which blast effects are best known. They can

    also be classified on the basis of their sensitivity to ignition as secondary or primary

    explosives. The latter are ones that can be easily detonated by simple ignition from a spark,

    flame or impact. Secondary explosives when detonated create blast (shock) waves which

    can result in widespread damage to the surroundings.

    The most common are explosions caused by chemical reactions. The first such a

    material was invented in 9th century in China and is known as black powder. In the

    Industrial Revolution of the 18th and 19th centuries, when a very significant and rapid

    development of chemistry was observed, new explosive materials have been invented. One

    can mention here nitro-glycerine, nitrocellulose, smokeless powder, and probably the most

    famous one, invented by Alfred Nobel, dynamite. Since World War II one of the most

    popular materials has become trinitrotoluene, called shortly TNT. It is a high explosive

    solid of yellow colour, an example of secondary explosive, which nature is a chemical event.

    The numerous applications of TNT caused that it is considered to be the standard measure

    of strength of bombs and other explosives. The method of quantifying the energy released

    in explosions is called TNT equivalent. The ton of TNT is a unit of energy equal to the

    amount of energy released in the detonation of one ton of TNT (which is approximately

    4.184 gigajoules). TNT equivalency is usually based on experimentally determined factors

    or the ratio of its heat of detonation to that of TNT [30].

    2.2. Blast phenomenonIn the moment of ignition, the explosive charge rapidly releases energy in the forms such as

    heat, sound and very dense and high pressure wave. From the viewpoint of influence on the

    structure the most important is the latter. The detailed descriptions of detonation process

    and blast wave propagation can be found in papers [6] and [7].

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    8/102

    8

    In this thesis author focuses on the most common chemically induced explosions. In

    the process of detonation a very rapid and stable chemical reaction takes place, which

    proceeds through the explosive material at a supersonic speed, called the detonation

    velocity. Detonation velocities range from 6700 to 8500 meters per second for most high

    explosives. The detonation wave rapidly converts the solid or liquid explosive into a very

    hot (about 3000- 4000C [6]), dense, high-pressure gas, and the volume of this gas which

    had been the explosive material is then the source of strong blast waves in air. Pressures

    immediately behind the detonation front range from 19000MPa to 33800MPa [7]. Only

    about one-third of the total chemical energy available in most high explosives is released in

    the detonation process. The remaining two-thirds are released more slowly in explosions in

    air as the detonation products mix with air and burn. This afterburning process has only a

    slight effect on blast wave properties because it is much slower than detonation. The blast

    effects of an explosion are in the form of a shock wave composed of a high-intensity shock

    front which expands outward from the surface of the explosive into the surrounding air.

    Figure 2-1. Blast wave pressure-time history [6].

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    9/102

    9

    The shape of the blast wave is very characteristic its front is very abrupt, reaching a

    very high value of pressure, then a rapid decrease is visible after which the pressure returns

    to initial value of the ambient atmospheric pressure (Figure 2-1).

    After a short time, the pressure behind the front may drop below the ambient

    pressure (Figure 2-2). During such a negative phase, called also negative-phase or suction

    phase a partial vacuum is created and air is sucked in. This is also accompanied by high

    suction winds that carry the debris for long distances away from the explosion source [6].

    Figure 2-1. Blast wave propagation [6].

    Along with propagation of the wave, following observations can be perceived: the

    wave decays in strength (the overpressure decreases steadily), lengthens in duration, and

    decreases in velocity. This phenomenon is caused by spherical divergence as well as by the

    fact that the chemical reaction is completed, except for some afterburning associated with

    the hot explosion products mixing with the surrounding atmosphere.

    The scheme describing shock wave propagation is very characteristic. It is usually

    called blast wave pressure-time profile, also referred to as overpressure curve. The main

    components describing the overpressure curve are (followed by [5]):

    - peak positive overpressure,- peak negative under pressure,- dynamic pressure,- positive and negative phase durations,- positive and negative phase impulses (integrals with respect to time of the

    respective pressures).

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    10/102

    10

    Figure 2-1 shows a typical blast pressure profile of ideal free air blast wave. The

    detonation can be considered to take place at time t = 0s. The first important component is

    the arrival time that is the time that it takes for the pressure wave to reach the point of

    interest. Pressure at that position suddenly increases to a peak value of overpressure; over

    the ambient pressure (ambient pressure is marked with dashed line). The pressure then

    immediately decays to ambient level (here the point of equalization is at time equal to 1.2

    milliseconds), then decays further to an under pressure (creating a partial vacuum) before

    eventually returning to ambient conditions at time. The peak pressure is usually referred to

    as the peak side-on overpressure, incident peak overpressure or merely peak overpressure.

    Throughout the pressure-time profile, two main phases can be observed - portion

    above ambient is called positive phase duration, while that below ambient is called

    negative phase duration. The negative phase is of a longer duration and a lower intensity

    than the positive duration.

    The shock wave overpressure curve is important from the standpoint of civil

    engineer as it a basis for determination of dynamic pressure. The dynamic pressure

    determines the value of loading that is subjecting the structure. Generally blast loading on

    a structure caused by a high-explosive detonation is dependent upon several factors:

    - the magnitude of the explosion,- the location of the explosion relative to the structure of interest (unconfined or

    confined),

    - the geometrical configuration of the structure,- the structure orientation with respect to the explosion and the ground surface

    (above, flush with, or below the ground).

    Experiments prove that for each pressure range there is a particle or wind velocity

    associated with the blast wave that causes a dynamic pressure on objects in the path of the

    wave. In the free field, these dynamic pressures are essentially functions of the air density

    and particle velocity. For typical conditions, standard relationships have been established

    between the peak incident pressure, the peak dynamic pressure, the particle velocity, and

    the air density behind the shock front. The magnitude of the dynamic pressures, particle

    velocity and air density is solely a function of the peak incident pressure, and, therefore,

    independent of the explosion size. Figure 2-3 gives the values of the parameters versus the

    peak incident pressure. The dynamic pressure is the most important parameter for

    determining the loads on structures.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    11/102

    11

    Figure 2-3. Peak incident pressure versus peak dynamic pressure, density of air behind the

    shock front, and particle velocity [7].

    Summarizing the size and material of the charge, as well as the stand-off distance

    (distance between the detonation point and examined point) and charges shape will all

    determine the magnitude and shape of the overpressure curve (the bigger stand-off

    distance, the longer duration of the positive-phase, lower amplitude and smaller intensity

    of the shock pulse). Additionally, the blast wave and the involved pressure can reflect off of

    surfaces (like ground or structures) in various directions, and cause further fluctuations

    (and reinforcement) in pressure at a single point. The reinforcement of the blast wave is

    described in next paragraph in more details.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    12/102

    12

    2.3. Blast-loading classificationTwo blast-loading categories have been distinguished (stated after [7]). The division bases

    on the confinement of the explosive charge, and so there are unconfined and confinedexplosions. These two categories can be further subdivided into next six basing on the

    location of the explosives and the structure. The Figure 2-4 presents full classification:

    Figure 2-4. Blast loading categories [7].

    Free air burst and surface burst are very important because they result in so-called

    ideal blast waves.

    Of the six categories, those from air bursts are seldom encountered and the free air

    burst is the least likely to occur. The possibility of such blast environments exists where

    potentially explosive materials are stored at heights adjacent to or away from protective

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    13/102

    13

    structures such as in manufacturing (process or storage tanks) or missile sites. In the latter,

    the rocket propellant would be a source of explosive danger to the ground crew and control

    facilities. The other four blast-loading categories can occur in most explosive

    manufacturing and storage facilities. In such installations, transportation of explosive

    materials between buildings either by rail, vehicle, or in the case of liquid or gases, through

    piping, is a possibility. Also, storage and handling of explosives within buildings are

    common occurrences.

    The Figure 2-4 shows also the five possible pressure loads associated with the blast

    load categories, the location of the explosive charge which would produce these pressure

    loads, and the protective structures subjected to these loads.

    Unconfined explosions

    An explosion, which occurs in free air, produces an initial output whose shock wave

    propagates away from the centre of the detonation, striking the structure without

    intermediate amplification of its wave (Figure 2-5), which takes place during ground

    impact (the main part).

    Figure 2-5. Free-air burst blast environment [7].

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    14/102

    14

    As the incident wave moves radially away from the centre of the explosion, it will

    impact with the structure, and, upon impact, the initial wave (pressure and impulse) is

    reinforced and reflected (Figure 2-6). The reflected pressure pulse of Figure 2-6 is typical

    for infinite plane reflectors.

    Figure 2-6. Pressure-time variation for a free-air burst [7].

    The variation of the pressure and impulse patterns on the surface of a structure

    between the maximum and minimum values is a function of the angle of incidence. This

    angle is formed by the line which defines the normal distance RA between the point of

    detonation and the structure, and line R (slant distance) which defines the path of shock

    propagation between the centre of the explosion and any other point in question on the

    structure surface (Figure 2-5).

    The dependency of peak reflected pressure on the angle of incidence can be seen in

    Figure 2-7 for a number of different shock strengths.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    15/102

    15

    Figure 2-7. Reflected pressure coefficient versus angle of incidence [7].

    The reflected overpressure can be much larger than the incident overpressure. For

    very strong shocks, where the ideal gas approximation is no longer valid, the predicted

    upper limit for reflected overpressure is much as 20 times the incident overpressure. It

    means that the objects surrounding a structure can create reflected waves that increase the

    loading seen by the structure. Additionally, there can be reinforcement in the corner

    geometries where multiple reflected waves can interact [19].

    Air Burst Explosion

    Air burst explosion is an explosion which is located at a distance from and above the

    structure, so that the ground reflections of the initial wave occur prior to the arrival of the

    blast wave at the structure. When the incident wave is reinforced by the ground effect, two

    phenomena can occur: a classical reflection (Figure 2-8) or a reinforcement reflection

    (called Mach Front, Figure 2-9).

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    16/102

    16

    Figure 2-8. Air burst blast environment classical reflection [8].

    Figure 2-9. Air burst blast environment reinforced reflection [7].

    The Mach front is formed by the interaction between initial (incident) and reflected

    pressure waves. This reflected wave is the result of the reinforcement of the incident wave

    by the ground surface. The occurrence of mentioned interaction depends on the angle of

    incidence between ground and incident wave. The critical angle is of around 40 (for < 40

    regular reflection takes place). The pressure-time variation of the Mach Front is similar to

    that of the incident wave except that the magnitude is somewhat larger [8].

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    17/102

    17

    Surface Explosion

    A surface burst explosion occurs when the detonation is located close to or on the ground.

    The initial wave of the explosion is reflected and reinforced by the ground surface to

    produce a reflected wave. Unlike the air burst, the reflected wave merges with the incident

    wave at the point of detonation to form a single wave, similar in nature to the Mach wave

    of the air burst but essentially hemispherical in shape (Figure 2-10) [7].

    Figure 2-10. Surface burst blast environment [7].

    Confined explosions

    Confined explosions are a very complex issue. When an explosion occurs inside a building

    the presence of the walls and ceiling greatly increases the number of blast wave structure

    interactions. Multiple reflections take place, and many waves coalesce to produce

    enhancements in corners and other local constrictions. The reflections can extremely

    amplify the peak pressures associated with the initial shock front. In addition, and

    depending upon the degree of confinement, the effects of the high temperatures and

    accumulation of gaseous products produced by the chemical process involved in the

    explosion will exert additional pressures and increase the load duration within the

    structure. The biggest difference between internal and external explosions, however, is the

    presence of the quasi-static gas pressure. Condensed explosives are approximately a

    thousand times denser than air. The detonation of an explosive in a building will introduce

    a quantity of hot gas into the building, as well as the shock waves mentioned above.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    18/102

    18

    Depending on the relative magnitudes of the mass of the explosive and the volume of the

    building, the gas pressure may be the dominant loading mechanism on the building

    elements. Because normal buildings (as opposed to containment buildings) have doors,

    windows, heating ducts, etc., which allow the gas to vent into adjacent rooms or the

    outside world, the gas pressure does not persist. The decay of the pressure, however, takes

    place in a time-scale much longer than the duration of the individual shock reflections, and

    the overall duration is typically much longer than the structural response time of elements

    loaded in the building. For this reason it is referred to as quasi-static gas pressure.

    When an explosion occurs within a confined area, gaseous products will accumulate

    and temperature within the structure will rise, thereby forming blast pressures whose

    magnitude is generally less than that of the shock pressure but whose duration is

    significantly longer [7]. The magnitude of the gas pressures as well as their durations is a

    function of the size of the vent openings in the structure.

    Figure 2-11. Combined shock and gas pressures for small and/or square chamber [7].

    Figure 2-11 illustrates an idealized pressure-time curve considering both the shock

    and gas pressures. As the duration of the gas pressures approaches that of the shock

    pressures, the effects of the gas pressures on the response of the elements diminishes until

    the duration of both the shock and gas pressures are equal and the structure is said to be

    fully vented.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    19/102

    19

    Fully Vented Explosion

    A fully vented explosion will be produced within or immediately adjacent to a barrier or

    cubicle type structure with one or more surfaces open to the atmosphere. The initial wave,

    which is amplified by the nonfrangible portions of the structure, and the products of

    detonation are totally vented to the atmosphere (there is no gas pressure build-up) forming

    a shock wave (leakage pressures) which propagates away from the structure.

    Partially Confined Explosion

    A partially confined explosion will be produced within a barrier or cubicle type structure

    with limited size openings and/or frangible surfaces. The initial wave, which is amplified by

    the frangible and nonfrangible portion of the structure, and the products of detonation are

    vented to the atmosphere after a finite period of time. The confinement of the detonation

    products, which consist of the accumulation of high temperatures and gaseous products, is

    associated with a build-up of quasi-static pressure (gas pressure). This pressure has a long

    duration in comparison to that of the shock pressure.

    Fully Confined Explosion

    Full confinement of an explosion is associated with either total or near total containment

    of the explosion by a barrier structure. Internal blast loads will consist of unvented shock

    loads and very long duration gas pressures which are a function of the degree of

    containment. The magnitude of the leakage pressures will usually be small and will only

    affect those facilities immediately outside the containment structure.

    2.4. Calculation methods of blast-loadingEmpirical and semi-empirical methods

    As it can be seen blast loading is a very complex issue. First attempts to describe it

    mathematically were started yet in the 1950s. The crucial notion in this matter is

    Hopkinsons law [9]. It is the most widely used approach for blast wave scaling. It

    establishes that similar explosive waves are produced at identical scaled distances when

    two different charges of the same explosive and with the same geometry are detonated in

    the same atmosphere [10]. Thus, any distance R from an explosive charge W can be

    transformed into a characteristic scaled distance Z:

    3

    RZ

    W (2-1)

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    20/102

    20

    where Wis the charge mass expressed in kilograms of TNT. The use ofZallows a compact

    and efficient representation of blast wave data for a wide range of situations.

    One of the first methods estimated peak overpressure Pso due to spherical blast was

    introduced by Brode [12] in 1955:

    3

    2 3

    6.71 10

    0.975 1.455 5.850.019 0.1 10

    so so

    so so

    P P bar Z

    P bar P bar Z Z Z

    (2-2)

    Next Newmark and Hansen [13] (1961) introduced a relationship to calculate the

    maximum blast overpressure, Pso, (in bars) for a high explosive charge detonates at the

    ground surface as:

    0.5

    3 36784 93so W WP

    R R

    (2-3)

    Another expression of the peak overpressure Pso (in kPa) was introduced by Mills [14]

    (1987), in which W is expressed as the equivalent charge weight in kilograms of TNT, and Z

    is the scaled distance:

    3 2

    1772 114 108soP

    ZZ Z (2-4)

    Then the maximum value of dynamic pressure qs can be calculated from the

    following formula:

    25

    2 7so

    s

    so o

    Pq

    P P

    (2-5)

    If the blast wave encounters an obstacle perpendicular to the direction of

    propagation, reflection increases the overpressure to a maximum reflected pressure Pr ,

    which can be obtained from RankineHugoniot relationships for an ideal gas:

    7 42

    7o so

    r so

    o so

    P PP P

    P P

    (2-6)

    Used symbols:

    Pso - peak overpressure,

    Po - ambient pressure.

    For more methods check paper [32].

    Beside methods that try to approximate the maximum value of peak pressure for

    different kinds of explosions (as free air burst or surface burst), the methods describing the

    shape of the overpressure curve have been developed. One of the methods represents the

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    21/102

    21

    pressure-time history by linear decay using approximate triangular equivalents. The other

    one, probably the most popular represents the pressure-time history by exponential

    functions. It is called Friedlanders method [11] and is described by following equation (only

    the positive phase):

    1

    A

    o

    t t

    tA

    s so

    o

    t tP t P e

    t

    (2-7)

    where t is the time, Pso is the peak overpressure, to is the duration of the positive

    phase, tA is the arrival time of blast wave, and is a positive constant called the waveform

    parameter that depends on the peak overpressure. It is depicted in Figure 2-12.

    Figure 2-12. Free-field pressure-time variation [7].

    Also many charts and tables predicting the blast pressures and blast durations have

    been developed. The method presented in UFC [7] is an example of semi-empirical

    method. It will be used in Verification Analysis to obtain peak incident overpressure and

    peak reflected pressure.

    Nowadays to simulate the blast one can use developed by US Army ConWep [16, 18],

    a blast loading predictive tool. It contains a big database of experimental records of blast

    loading parameters from a wide range of explosive charge masses and stand-offs produced

    by Kingery and Bulmash [15]. The main advantage of this model is that the loading is

    applied directly to the structure subject to the blast. There is no need to include the fluid

    medium in the computational domain. In the ConWep model, empirical data for two types

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    22/102

    22

    of waves are available: spherical waves for explosions in midair and hemispherical waves for

    explosions at ground level in which ground effects are included [21]. It is also available in

    Abaqus.

    Also AtBlast program [17] funded by US General Services Administration is available

    for free. Both of these applications can calculate free-field and reflected blast pressure

    histories (from free-air and surface burst explosions), average peak pressure and impulse

    from a hemispherical surface burst on a specified reflected wall area.

    Numerical methods

    A blast analysis using pressure time history predicted by the ConWep or AtBlast

    applications produces reasonable results. However it cannot consider the effects, such as

    confinement due to geometry of structure, reflection of multiple blast waves, and

    shadowing occurring when an object is blocking a surface of structure from direct blast

    wave. These factors can change the blast pressure dramatically. If more accurate results

    have to be obtained, also more accurate methods than the semi-empirical ones used in

    mentioned applications need to be applied.

    One of the possibilities is to use software which supports finite element analysis.

    One can mention here Autodyn, Dyna3D, LS-Dyna, and Abaqus. The mentioned software

    supports a coupled analysis, for which the blast simulation module is linked with the

    structural response module. In this type of analysis the CFD (computational fluid

    mechanics) model for blast-load prediction is solved simultaneously with the CSM

    (computational solid mechanics) model for structural response. By accounting for the

    motion of the structure while the blast calculation proceeds, the pressures that arise due to

    motion and failure of the structure can be predicted more accurately.

    The similar simulations can be conducted in Abaqus using the Coupled-Eulerian-

    Lagrangian (CEL) method.

    3.Finite Element Analysis techniquesChoice of an appropriate kinematical description of the continuum is crucial while dealing

    with mechanics of fluids and solids because it determines the relationship between the

    deforming continuum and the finite grid or mesh of computing zones. The algorithms of

    continuum mechanics usually make use of two classical descriptions of motion: the

    Lagrangian description and the Eulerian description [20, 21]. Additionally, for use in more

    sophisticated problems, when functionality of these two methods is insufficient, two other

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    23/102

    23

    techniques have been developed: Coupled Eulerian-Lagrangian and Arbitrary Eulerian-

    Lagrangian. The more detailed description is given in Appendix A.

    4.Tools and methods4.1. SoftwareThe shock wave propagation analyses have been performed with use of three basic tools.

    First one was Abaqus/CAE which provided a complete interface for creating simulations,

    which then were solved by Abaqus/Explicit. Second one was Python language, one of few

    supported languages for the purpose of use in Abaqus. Also Abaqus Scripting Interface was

    used. It is, as referenced in [1], an application programming interface (API) to the models

    and data used by Abaqus.

    Abaqus

    Abaqus is commercial software for performing numerical analyses of different physical

    phenomena. The most common ones in the case of civil engineering can be simulations of

    statics and dynamics of 2D and 3D structures. However Abaqus, originally released in 1978

    has a much wider range of application, as it is commonly and successfully used in

    automotive, aerospace and industrial products industries. Its important feature is that it

    provides numerous multiphysics simulations (that is such, which involve multiple physical

    models or multiple simultaneous physical phenomena). By the description of Abaqus often

    two shortcuts appear: FEA and CAE. First one is the name of numerical technique (Finite

    Element Analysis) for finding approximate solutions of partial differential equations (PDE)

    as well as of integral equations which this software uses. The second is abbreviation of

    Computer Aided Engineering and refers to a group of software applications that have been

    produced to help in engineering tasks.

    Abaqus/CAE is an environment that generates an input file (that is a file that

    contains all necessary data) that is next submitted to the Abaqus solver.

    Abaqus/Explicit

    Abaqus/Explicit as its name points uses explicit dynamics analysis procedure. It is based

    upon the implementation of an explicit integration rule together with the use of diagonal

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    24/102

    24

    (lumped) element mass matrices. The equations of motion for the body are integrated

    using the explicit central-difference integration rule:

    11 1

    2 22

    i iN N Ni

    i i

    t tu u u

    (4-1)

    1 1 12

    N N N

    i i ii

    u u t u

    (4-2)

    In the formulas (4-1) and (4-2) Nu is degree of freedom (a displacement or rotation

    component) and the subscript i refers to the increment number in an explicit dynamics

    step. The central-difference integration operator is explicit in the sense that the kinematic

    state is advanced using known values of1

    2

    N

    uu

    and Niu from the previous increment.

    Explicit procedure uses diagonal element mass matrices. Therefore the accelerations

    at the beginning of the increment are computed by:

    1N NJ J J

    i i iu M P I

    (4-3)

    where:

    NJM - Mass matrix,

    JP - Applied load vector,

    JI - Internal force vector.

    The benefits of lumped mass matrix are first of all significant computational

    advantages of calculations. Lumped mass matrix is a sparse matrix thus it gives much

    better performance of computer calculations than a full one. A diagonal mass matrix

    negates also the need to integrate mass across the deformed element and to build tangent

    stiffness matrix. The internal force vector JI is assembled from contributions from the

    individual elements such that a global stiffness matrix need not be formed.

    To assure the stability of the procedure, which integrates through time by using

    many small time increments, special conditions need to be fulfilled. The central difference

    operator is conditionally stable, and the stability limit for the operator (with no damping)

    is given in terms of the highest frequency of the system as:

    max

    2t

    (4-4)

    And with damping:

    2max maxmax

    21t

    (4-5)

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    25/102

    25

    where max is the fraction of critical damping in the mode with the highest frequency.

    An approximation to the stability limit is often written as the smallest transit time of

    a dilatational wave across any of the elements in the mesh:

    min

    d

    Lt

    c (4-6)

    where minL is the smallest element dimension in the mesh and dc is the dilatational wave

    speed in terms of 0 and 0 . The current dilatational wave speed dc is described by the

    formula:

    2dc

    (4-7)

    where is the density of the material and , are effective Lames constants. For an

    isotropic, elastic material Lames constants can be defined in terms of Youngs modulus E

    and Poissons ratio by:

    0

    0

    1 1 2

    2 1

    E

    E

    (4-8)

    This estimate for t is only approximate and in most cases is not a conservative

    (safe) estimate. In general, the actual stable time increment chosen by Abaqus/Explicit is

    less than this estimate by a factor between1

    22

    and 1 in a two-dimensional model and

    between1

    23

    and 1 in a three-dimensional model. Description is based on [25, 26].

    The equations presented above are based on CourantFriedrichsLewy condition

    (CFL condition), which is a necessary condition for convergence while solving certain

    partial differential equations (usually hyperbolic PDEs) numerically by the method of finite

    differences. It arises when explicit time-marching schemes are used for the numerical

    solution. As a consequence, the time step must be less than a certain time in many explicit

    time-marching computer simulations; otherwise the simulation will produce wildly

    incorrect results [31].

    Abaqus Scripting Interface

    Abaqus Scripting Interface is an application programming interface which is (after [2]) a

    particular set of rules and specifications that software programs can follow to communicate

    with each other. Entering the data is different - instead of clicking on specific icons and

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    26/102

    26

    fulfilling numerous boxes, user has to set every parameter using functions provided by ASI.

    This way traditional procedure of entering data can be omitted.

    As it is written in [1], the Abaqus Scripting Interface is a customized extension of

    standard Python. This means that in order to write scripts or programs one has to do it in

    Python language. Python is dynamic, interpreted and interactive high-level programming

    language. It enables object-oriented, structural and functional programming. It is

    developed as an open source project, what means it is free and its source code is publicly

    available. The feature distinguishing Python from other languages is its syntax, in which

    the scope of the loop or conditional statement is designated through indentation.

    4.2.Creation of CEL model in Abaqus

    Creating a model in Abaqus requires specifying numerous data. In this part, to show the

    concept of Coupled Eulerian-Lagrangian analysis technique, individual steps of creating a

    fully-working model are described.

    As it is written in Chapter 3 and Appendix A, there are significant differences in

    Lagrange and Euler descriptions. These differences influence significantly on the creation

    of model. The main differences are presented in following paragraphs.

    In Lagrange analysis technique nodes are fixed within the material, which fulfils it

    entirely, so the element boundary coincides with the material boundary. In Eulerian

    analysis nodes are fixed in space, so the material flows through non-deformable elements.

    In contrast to Lagrange elements Eulerian ones may not always be 100% full of material -

    they can be partially or completely void. The Eulerian material boundary must, therefore,

    be computed during each time increment and generally does not correspond to an element

    boundary. Eulerian models typically consist of a single Eulerian part, what is the notable

    difference comparing to Lagrange models, where several parts are created and then

    assembled into one model [21]. This part can be arbitrary in shape but typically it is a

    simple rectangular grid of elements. It represents the domain within which Eulerian

    materials can flow (area of simulation). The necessity of Eulerian model to consist of a

    single part induces the problem of defining different objects. It is solved by creating

    multiple regions within the Eulerian-type part instance. Then a particular material is

    assigned to a particular region. This way complex body geometry can be defined. Details

    can be found in paper [21].

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    27/102

    27

    Generally, the entire procedure involves accomplishing following steps (from [21]):

    1. Create Eulerian-type part/parts that defines/define the geometric region withinwhich Eulerian materials can flow. Create partitions that will represent the

    initial boundaries between different materials in the part (needed if materials

    are assigned uniformly across a region).

    2. Define materials.3. Define and assign an Eulerian section for the model.4. Create an instance of each Eulerian part. Merge instances to obtain single part

    instance.

    5. Create a field output request (in order to view the deformation of materials in anEulerian model, output variable EVF is necessary).

    6. Create a predefined field that defines the topology of materials in the initialconfiguration of the Eulerian part instance.

    7. Define any loads and/or boundary conditions acting on the model.8. Create a hexagonal mesh for the Eulerian part.

    Parameters

    During the preparation of the model for blast simulations, following material parameters

    have to be set (the given values are for typical air conditions and TNT as an explosive

    charge):

    1. Air described by following parameters:a. Density - 31.293 kg m ,b. EOS parameters for ideal gas:

    i. gas constant - 287 J kg K ,ii. ambient pressure - 2101325 N m ,

    c. Specific heat - 717.6 J kg K ,d. Data describing the viscosity kg s m - temperature dependency K

    (in tabular form),

    e. Physical constants:i. Stress - 2101325 N m ,

    ii. Specific energy between 193300 J kg and 219780 J kg ,

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    28/102

    28

    2. Explosive charge (TNT):a. Density - 31630 kg m ,b. Jones-Wilkins-Lee EOS parameters:

    i. detonation wave speed - 6930 m s ,ii. coefficients:

    1. A - 11 23.7377 10 N m ,2. B - 11 20.037471 10 N m ,3. R1 - 4.15 ,4. R2 - 0.9 ,5. - 0.35 ,

    iii. detonation energy density J kg ,iv. pre-detonation bulk modulus 2N m ,

    c. Physical constants:i. Stress - 2101325 N m ,

    ii. Specific energy between 3680 kJ kg (for classical TNT from1970s) and 5000 kJ kg (in the case of TNT-C4 compounds).

    Ideal gas law is used to describe the compressible flow phenomena that include

    microscopic properties such as density, pressure and internal energy. The formula of ideal

    gas:

    ZAp p R (4-9)

    where p is current pressure, pA is ambient pressure, is density, R is gas constant, is

    current temperature, Z is the absolute zero on the temperature scale being used.

    Jones-Wilkins-Lee equation of state is widely used to model both the detonation and

    the expansion phase of the explosion (unlike the ideal gas EOS which can be only used to

    model expansions phase) [22]:

    1 2

    int

    1 2

    1 1

    R R

    p A e B e eR R

    (4-10)

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    29/102

    29

    where p is current pressure, coefficients A, B, R1, R2 and are experimentally determined

    material constants, 0 is the ratio of densities: - of the detonation products, 0 - of

    the explosive. e

    int

    is internal energy per unit mass.All above values are specified in SI units, that is: length - metres [m], mass -

    kilograms [kg], time - seconds [s], thermodynamic temperature - kelvins [K]. Force is given

    in newtons (N = kgm/s2) while pressure in pascals (Pa = N/m2).

    4.3.Assumptions and restrictions in the scriptBasic elements building the structure are straight sections and connections. In the script

    following assumptions have been made:

    - Straight sections are defined in their axes, by specifying start and finish points(hereinafter referred to as important points),

    - Sections may intersect freely,- If two or more sections have to be connected, this connection has to take place

    in sections important points only,

    - To avoid unpleasant gaps, connection should be in horizontal or vertical planeonly,

    - Depending on the type of connection (vertical or horizontal) sections that arebeing joined should have respectively same width or height,

    - From the fact that it is impossible to obtain horizontal angle from sectionsdirected vertically (which would sometimes result in inadequate orientation of

    this sections faces) an additional, non-vertical section, from which the vertical

    one inherits the horizontal angle, has to be specified,

    - When sections of different width or height (later called important dimension)connect at angle 90, the precedent of them is one which has smallest important

    dimension it results in smooth connection.

    In process of modelling the structure, the user has to have in mind following

    limitations:

    - Although the sections can intersect at any angle, angles close to 180will resultin very long connection (because of fact that tan 180 ),

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    30/102

    30

    - If the connecting sections differ in value of important dimension, depending onthe mentioned difference for angles bigger than 90 some strange artefacts may

    appear,

    - In case of failure to comply with assumption number 5, unpleasant connectionwill be created.

    Meeting above requirements one can be sure about correctness of the model and its

    compatibility with his own assumptions and predictions.

    4.4. Simplified block scheme of the scriptThis script automates operations required to be done in Abaqus/CAE modules, requiring

    the user to specify just the key values, like points describing geometry of the structure,

    basic material properties and duration of analysis. Although some of the actions can be

    done at any time, in this application a certain order of execution of some modules has been

    introduced. This treatment results from the fact that for the correct execution of the script

    (as well as the Abaqus solver) it is necessary to enter all required data. Some of the

    functions are, however, non-obligatory, that is one can choose if any (if yes then which

    one) to execute. Figure 4-2 presents applications functioning block scheme.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    31/102

    31

    Figure 4-2. Block scheme presenting the functioning algorithm.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    32/102

    32

    The first main step is the initialization of applications main class called MyModel

    (A). Next one is to define data concerning the arrangement of corridors (B). The more

    detailed graph showing this procedure is presented in Figure 4-3.

    Figure 4-3. Block scheme presenting major steps of geometry definition.

    At the beginning one has to define number of geometry data sets. These sets should

    be filled with information on particular corridors and connections. To assure flexibility of

    the script, no check of correctness of the input data is performed. While setting the

    geometry, data is stored in complex list structures.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    33/102

    33

    Figure 4-4. Block scheme presenting draw method execution.

    Subsequent step is call to draw method (step C). After some computations each

    record from data structures becomes the basis for the creation of corridors and connections

    (in both cases firstly sketches, then parts and in the end instances are created). It was one

    of the assumed conventions to firstly define all the necessary geometry and then, finally

    draw it.

    In this place it is important to mention that creation of simple connection of two

    corridors is not always a process involving just creation of two cuboids. In order to avoid

    gaps or discontinuities in corridors other elements need to be placed. Then, in fact such a

    connection can be made even of several elements, which in the end are merged into new

    element. The complexity of the connection is dependent on the relative angle between the

    corridors. If angle is equal to zero, no additional elements are required. If the angle is

    smaller than or equal to 90, two elements are sufficient. The most complex case takes

    place when the angle is wider than 90, then four elements have to be used and the

    procedure itself involves operations such as merging and cutting, performed in a certain

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    34/102

    34

    way. When all the declared elements of each set have been drawn, they are merged into

    one element.

    Next step, which is partitioning of the elements (D), is non-obligatory (denoted as

    No partitioning in Figure 4-2). However, usually when the geometry is complex, it is

    necessary to execute it. We can choose from two different partitioning algorithms (Auto

    partitioning and Raw partitioning), it is recommended to check each one and compare

    which one produces better results for particular structure. Unfortunately author cannot

    assure that these algorithms will always produce elements appropriate for meshing, as the

    complexity of the different connections is unlimited, and list of options available in Abaqus

    is very short. Fortunately in many cases just a slight manual alteration may solve the

    problem. This issue is described in part Problems and difficulties. Except cutting the model

    into regions, this part of script merges all geometry data sets and creates calculation step

    (Explicit dynamic), loads acting on the walls (air pressure, including possibility of an inflow

    and outflow of air) and boundary conditions (fixed displacements).

    In step E user has to specify parameters describing the medium (air) and explosive

    charge, like number of explosives, their location and weight. Also moments in time when

    the detonations take place have to be specified.

    Subsequent step (F) covers declaring the output data one would like to obtain. Two

    methods differing in way of data recording are possible to execute. The data recording

    frequency can be chosen via appropriate attribute.

    The next step is to call mesh method (G). Though the suggestive name, the function

    does a bit more than just create a mesh of a specified size, additionally it modifies the

    special keywords, what is an important issue.

    The penultimate step (H) is optional and related to point (F). It concerns defining

    individual finite elements as output data.

    The last operation is a call to job method (I), which in this case does nothing else

    and nothing more beyond what its name implies. The method allows declaring number of

    CPUs and threads involved in process of calculations.

    4.5. Structure of the scriptOne of the properties of Python is that it is an object-oriented programming language. This

    important feature has been taken into account while creating the script. Thanks to object-

    oriented approach it was possible to divide the code into separate parts classes

    consisting of appropriate properties and methods. This way greater clarity of code

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    35/102

    35

    (comparing to procedural approach) has been achieved. Moreover, it allows easier and

    error-free further development. Also the usage is simpler, as the functions being used are

    precisely specified.

    The application consists of 23 classes, which one can divide into 4 groups: creation of

    geometry (marked as B, C in Figure 4-2), partitioning into regions (D), setting material

    properties (E) and output parameters (F and H), and job execution (G and I). The classes

    and methods, on which user can operate are briefly described below.

    The script has been designed in such a way, that all necessary commands (in order to

    create a functional and complete Abaqus model) are localized within the main class

    MyModel. The object of this class instance is responsible for managing of all data and

    subordinate classes. The user can call following methods:

    Model object initialization (A)

    __init__(name : str)

    Constructor method called when the MyModelobject is being initialized. The objects of all

    subordinate classes are created. The only attribute is the name of the model, under which it

    will appear in Abaqus. Exemplary call:

    mod = MyModel('Corridor')

    Methods for creating geometry (B, C)

    addCMs(GDSList : list)

    The method generates geometry data sets specified in GDSList list. GDSList should contain

    their names. Exemplary execution:

    GDSList = ['GeometrySet_1', 'GeometrySet_2']

    mod.addCMs(GDSList)

    The first line creates list of data sets. Then it becomes an input parameter to addCMs()

    method.

    addCorridor(name : str, startPoint : vec3f, endPoint : vec3f,

    width : float, height : float, oocc : str, addCorr : CorridorData,

    newName : str = '') : CorridorData

    Method is used to create straight corridor section assigned to set specified by its name

    name, with geometry defined by start and end points (startPoint and endPoint) and cross-

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    36/102

    36

    section dimensions width and height. By parameter ooccuser can specify corridors closure.

    Possible options are following:

    - oo corridor open at both points,- oc corridor open at start point and closed at end point,- c0 corridor closed at start point and open at end point,- cc corridor closed at both points.

    addCorridorIndexspecifies the corridors parent corridor, from which it will inherit vertical

    angle. This parameter shall take value other than None only if the corridor is pointing

    vertically. The last parameter is optional and is a new name of the corridor. This option is

    useful if geometry data set contains a single corridor section. Example:

    Corr1 = mod.addCorridor(GDSList[0], vec3f(0,0,0),

    vec3f(10,0,0), 2, 3.5, 'oo', None)

    The corridor is created from point [0,0,0] to [10,0,0], with width of 2m and height of 3.5m.

    The corridor has both ends open and no parent section. It will be assigned to geometry

    data specified by the name contained in GDSList[0].

    addConnection(name : str, corr1 : CorridorData, corr2 :

    CorridorData, verORhor : str, connType : str)

    This method adds connections between corridors. The input parameters are: name of the

    geometry data set, to which the connection will be associated, followed by the

    CorridorData objects of two corridor sections that the connection will concern. Next

    attribute is string specifying the plane in which the connections take places. It can take two

    values:

    - hor if the connection is in horizontal plane,- ver if the connection is in vertical plane.

    The last parameter concerns which parts of corridors to merge. Possible options:

    - sf the part will connect starting point of first specified corridor with endingpoint of the second one,

    - fs the part will connect ending point of first specified corridor with startingpoint of the second one,

    - ss the part will connect starting points of both corridors,- ff the part will connect ending points of both corridors.

    Exemplary use:

    mod.addConnection('GS1', Corridor2, Corridor1, 'hor',

    'sf')

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    37/102

    37

    The connection is created between starting point of corridor identified by Corridor2 and

    ending point of corridor Corridor1. The connection is in horizontal plane. It is saved to

    geometry data set specified by name GS1.

    draw()

    AddCorridor() and addConnection() methods append new elements to appropriate lists,

    respectively for sections of corridors and connections between them. Method draw()

    executes classes, which relying on the mentioned lists, generate the defined geometry

    elements. Finally, after all elements have been drawn, they are merged and the unnecessary

    instances are suppressed. The only possible call:

    mod.draw()

    Methods for partitioning (D)

    merge(name : str, timePeriod : float)

    First of three alternative methods, it merges all defined geometry sets into a new one. No

    partitioning takes place. First argument is name of resulting data set. The second one is

    duration of the simulation, needed for dataPrePartition() method, which is executed within

    the merge() method. It defines the analysis step (Explicit Dynamics), creates boundary

    conditions and loads. Example:

    timePeriod = 0.007

    mod.merge("Merged", timePeriod)

    The method creates data set Merged. The duration of the analysis is 0.007 second.

    autoPartitionAndMerge(name : str, PartHeights : list,

    timePeriod : float)

    Method performs same operations as the above one; additionally it creates regions based

    on Abaqus algorithm regarding auto partitioning of faces. This method enhances its usage

    to cells. First argument is name of resulting data set. The second is a list of floats, in whichone can define additional horizontal planes for partitioning. The last parameter defines

    duration of the simulation. Exemplary use:

    timePeriod = 0.07

    horPartitions = [-2.0, 2.0, 6.5]

    mod.autoPartitionAndMerge("Merged", horPartitions,

    timePeriod)

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    38/102

    38

    In the example data set named Merged is created. The geometry will be cut by three

    planes, at heights defined in horPartitions list. The analysis will be performed for 0.07s time

    period.

    rawPartitionAndMerge(name : str, timePeriod : float, angPart :

    bool)

    This method performs simple partitioning algorithm, based on planes defined by corridors

    faces. The arguments are: name of the new data set, length of the simulation and Boolean

    parameter specifying whether the angular partitions should be taken into account.

    Example:

    timePeriod = 0.07

    mod.rawPartitionAndMerge("Merged", timePeriod, False)

    In above example the method creates data set Merged. The analysis will be performed for

    time equal to 0.07s, the partitioning along bisectors of angles between intersecting

    corridors is disabled.

    Methods for setting material properties (E)

    dataPrePartition(timePeriod : float)

    The method shall be considered as private method. It is called by methods belonging to the

    group Methods for partitioning. It is responsible for creating Explicit Dynamics step, load

    and boundary conditions. The only argument it takes is the duration of analysis.

    dataPostPartition(tntLoc : list, tntWeight : list,

    detTime : list)

    This complex method creates fist of all air and TNT properties and Eulerian section

    consisting of these materials. Next it assigns them to the part. Subsequently the space for

    explosive charge is cut from the model, thus creating two sets - air and TNT. At the end, in

    Abaqus Predefined Field Manager, appropriate section materials are assigned to particularsteps. The method takes following arguments: list of vectors specifying centre points of

    explosives, lists of floats describing weight of charges and time of the explosion. Exemplary

    call:

    tntLoc = [vec3f(3,2.7,1.2), vec3f(15,14.8,0)]

    tntWeight = [270, 300]

    detTime = [0.01, 0.4]

    mod.dataPostPartition(tntLoc, tntWeight, detTime)

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    39/102

    39

    In this example two explosive charges are created. Their location within the model is set in

    tntLoc list, weight (270 and 300kg) in tntWeight list. The explosions take place in moment

    of time contained in detTime list.

    Methods concerning output data (F, H)

    HO(HOPoint : vec3f)

    Method for creating history output for specified point (HOPoint). Example:

    mod.HO(vec3f(20.3, 14, 27))

    FO(numIntervas : int)

    This method creates field output for the whole model. The parameter it takes is number of

    data recordings the application will take during the simulation. If the value is 0, then all

    increments are stored to output file. Example:

    mod.FO(700)

    Here the field output containing 700 records is created.

    FOSet(centrePoints : list, FOSSizes: list,

    numIntervals : int)

    Method for setting a region of the model as a field output domain. The localizations of

    points being centres of these regind and sizes of the sets are defined by the first two

    parameters the function takes. The last argument works in the same why as in the function

    above. Exemplary use:

    centrePoints = [vec3f(3, 0.4, -0.3), vec3f(5, 0.1, 0.3)]

    FOSSizes = [vec3f(0.15, 0.15, 0.15), vec3f(0.1, 0.1,

    0.1)]

    mod.FOSet(centrePoints, FOSSizes, 0)

    This call will effect in creation of two field output sets located in points specified by

    components of centrePoints list, and with sizes contained in FOSSizes list. ThenumIntervals value equal to 0 indicates that every simulations increment will be saved.

    elementFOSet(pointList:list, numIntervals : int)

    Unlike three methods described above, this one has to be executed after creating mesh. It

    generates field output for finite elements defined inpointList list. The second parameter it

    takes is number of data recordings the application will take during the simulation. If the

    value is 0, then all increments are stored to output file. Example:

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    40/102

    40

    pList = [vec3f(10,1,0), vec3f(1,1,0.5), vec3f(20,1,0)]

    mod.elementFOSet(pList, 123)

    In this example three element field output sets are created (defined in pList). The

    numIntervals specifies how many increments will be saved (here 123 increments).

    Methods related to job execution (G, I)

    mesh(meshSize : float)

    This method creates mesh of specified size. It edits keywords associated with Explicit

    Dynamics step in order to make calculations possible. Example:

    mod.mesh(0.15)

    The mesh of 15cm size will be created.

    job(jobName : str, writeInput : bool, numCpus : int, numDomains :

    int, activateLoadBalancing : bool)

    The method creates job with name specified byjobName. The second argument is a

    Boolean and determines whether to save .inp file. The three recent parameters describe the

    parallelization options available in Abaqus. The first of them defines number of CPUs

    which will be engaged in calculations, second denotes number of threads for each CPU, the

    latter one should be set to True if the dynamic load balancing should be activated. For

    more details check Abaqus documentation [21, 26].

    save(pathName : str)

    This method enables to save .cae file under the named specified bypathName parameter:

    mod.save("C:/myAnalysis")

    All classes are presented schematically in Appendix B.

    4.6. Problems and difficultiesDuring the preparation of the script several obstacles connected with Abaqus limitations

    have been encountered. Certain of them have been just partly resolved and in order to

    submit a job some modifications of the model can be required. In many cases they forced

    author to revise initial assumptions and modify the code to cooperate with Abaqus more

    properly. In this part these issues are extensively described.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    41/102

    41

    The script allows to define arbitrary geometry. However because of the necessity to

    use single Eulerian part instance, in order to distinguish different objects in its interior, it is

    required to cut it with datum planes (partitioning). In the script this method is used to

    separate explosives' cubes from the air domain. The problems with this approach arise

    during meshing. Eulerian parts use unique meshing rules, which are more restrictive

    comparing to those used in Lagrangian ones, that is the finite elements need to form right

    prisms (so the faces joining the two bases have to be perpendicular to the bases) with bases

    being tetrahedral. In certain cases the algorithm fails to mesh the object. First of it is

    impossible to mesh complex 3D shapes (e.g. if corridors are directing at different horizontal

    and vertical angles simultaneously). The author has taken efforts to solve these problems.

    Despite using three different partitioning procedures, finally not all of them have been

    eliminated. Starting from the assumption that the fundamental objective of every computer

    script is that it has to work, to make partitioning methods work properly, some usage

    limitation have been imposed. Therefore the following structure of the geometric data has

    been adapted: elements have to be grouped into so called sets, where each set should

    consist of corridors lying in one level (in horizontal plane). Also the vertical distance

    between sets should be greater than height of the storey, so that individual sets would not

    overlap each other. Then one of two available partitioning methods will result in model

    correctly partitioned. As it was mentioned, to assure flexibility, no special restrictions are

    set, so one can arrange the geometry arbitrarily. The partitioning methods are non-

    obligatory, but before creating mesh and running job, it is possible that one would have to

    manually adjust the model for calculations.

    4.7. Further developmentNow the application supports only cuboids as elements to define geometry. A good

    solution would be to add some other methods of the structure generation, for example to

    give the possibility to model rooms or halls of more complex shape. As far as creation ofthe script, which would handle it, seems not be a huge problem, it appears to be harder to

    figure out how to solve the problem of connection with corridors in order to avoid

    discontinuities.

    Also the script would be more functional if point 5 from part Assumptions and

    restrictions would lose its validity. However, this involves solving problem of this matter:

    how to model connection of corridors, when each has different height, merging at

    horizontal angle different than 0 or 180?

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    42/102

    42

    Another issue is that now the model consists of only rigid elements, which are fixed

    and unable to displace. Adding deformable materials inside the structure (e.g. building)

    would be a nice option, which would extend applicability of the script.

    The most problematic issue is one related to partitioning of the model.AutoPartition

    method is sufficient for structures which consist of horizontal layers of sets of corridors and

    additional corridors directed vertically. However for cases described in part Problems and

    difficulties this method is insufficient. Implementation of fully working code seems to be

    impossible; however this is author's private opinion.

    For now all faces of corridors (except the ones which normal vector coincides with

    the corridors longitudinal axis) have fixed outflow and inflow of the air. An interesting

    option would be to add windows and doors, so that the outflow would be enabled.

    5.Analyses5.1. Benchmarking analysis of blast wave propagation5.1.1. DescriptionThe analysis is aimed to compare peak incident overpressure obtained using Abaqus

    numerical approach and UFC analytical method. The subject of the analysis is a simple air

    cuboid with dimensions:

    9.2

    0.4

    0.4

    L m

    l m

    h m

    The TNT explosive weights 1kg and is located 1m from one end in centre of the cross

    section. The model as shown in Figure 5-1 has been prepared using script. The code can be

    found on the CD in folderModels under the name SimpleCuboid.py.

    Figure 5-1. Perspective projection of an air cuboid.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    43/102

    43

    5.1.2. Analytical solutionMethod described in UFC is used. For convenience calculations are performed for points

    based on scaled distance Z (here presented for1 3

    0.25 ftZlb

    ). According to UFC guidelines

    the charge is increased by 20%.

    1 3 1 3

    1.2 2.65

    0.25 0.0992

    0.105

    W kg lbs

    ft mZ

    lb kg

    R m

    Peak incident overpressure can be calculated from Table 2-7 [7]:

    0.25 4600 31.82soP Z psi MPa

    Some of the results are shown in Table 5-1.

    5.1.3. Numerical solutionThe numerical simulations have been performed in Abaqus. Dimensions of the model have

    been taken as in the analytical solution. The time of the analysis has been set to 0.1s, while

    air and TNT parameters have been set according to chapter 4. Internal energy has been set

    to 3680kJ/kg. The pressure has been measured uniformly across the model in elements

    located on the longitudinal axis (in red points shown in Figure 5-2). Five different mesh

    refinements have been studied: 10, 8, 6, 4and 2cm. The results are given in Table 5-1.

    Figure 5-2 shows distribution of points across the model and location of the explosive

    charge for model meshed with 2cm elements.

    Figure 5-2. Plan view of the air cuboid meshed with 2cm elements. Measuring points are in

    red, while blast charge in blue.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    44/102

    44

    5.1.4. Discussion of the resultsFigure 5-3 shows the comparison of maximum pressures obtained using UFC algorithm

    (blue line) and Abaqus.

    Figure 5-3. Comparison of peak pressure obtained for each distance from the charge for

    UFC solution and Abaqus.

    It can be seen that mesh refinement has significant influence on the results.

    Generally the finer the mesh the higher values of pressure one obtains. While for stand-off

    distance smaller than 2.0m the differences between Abaqus results for all kinds of meshes

    are visible, for the remaining part those for 10, 8, 6 and 4cm meshes are minimal, except

    0,1

    1,0

    0,80 1,80 2,80 3,80 4,80 5,80 6,80 7,80

    Peakpressure[MPa]

    Distance from the charge R [m]

    UFC

    10cm, 1473 FE

    8cm, 2875 FE

    6cm, 6874 FE

    4cm, 23000 FE

    2cm, 184000 FE

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    45/102

    45

    the 2cm mesh, which still provides higher results (minimum of 65%). The pressure

    differences for 2cm mesh are presented in Table 5.1.

    Table 5-1. Comparison of peak incident pressure (Pso) obtained by UFC method and in

    Abaqus for selected points.

    R [m]Pso [MPa] Pressure difference [%]

    with respect to UFCUFC Abaqus 2cm mesh

    0.8 1.77097 1.24762 30

    0.9 1.36714 1.03842 24

    1 1.15650 0.88089 24

    1.1 0.96297 0.74037 23

    1.2 0.75270 0.61475 18

    1.6 0.37646 0.26066 31

    2 0.22305 0.11699 48

    2.4 0.15001 0.05419 64

    2.8 0.10886 0.03204 71

    3.2 0.08225 0.02208 73

    3.6 0.06420 0.01511 76

    4 0.05218 0.01039 80

    4.4 0.04495 0.00749 83

    4.8 0.04077 0.00562 865.2 0.03658 0.00432 88

    5.6 0.03239 0.00342 89

    6 0.02820 0.00275 90

    6.4 0.02444 0.00227 91

    6.8 0.02248 0.00198 91

    7.2 0.02051 0.00192 91

    7.6 0.01856 0.00202 89

    8 0.01655 0.00200 88

    Basing on Table 5-1 it can be concluded that in the case of 1kg TNT charge, the

    convergence of results obtained using UFC method and Abaqus CEL simulations is similar

    for stand-off distance smaller than 1.6m. However for R outside this range the differences

    grow very significantly. For R = 3.5m pressure calculated using Abaqus with 2cm mesh is

    76% smaller than UFC value. For R = 8m pressure calculated using UFC is 14.5kPa, while

    using Abaqus is 2.1kPa. The discrepancy is enormous, for example in order to assess the

    damage that the pressure can induce. A pressure of 7kPa would already partially demolish

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    46/102

    46

    houses, while 2.1kPa would only damage windows and ceilings (for more details

    check [32]).

    It can be assumed that for 1cm mesh the numerical results would be even closer to

    the UFC, especially for small stand-off distances, where the biggest disturbances are

    observed.

    5.2.Analysis of the cubicle5.2.1. DescriptionThe analysis concerns comparing empirical and numerical computational methods of

    calculating blast wave properties induced by an explosion for a simple structure (shown

    schematically in Figures 5-4 and 5-5). The calculations presented in subsequent part

    Analytical solution are based on methods described in UFC [7]. The numerical simulations

    were conducted in Abaqus/Explicit. The main point of this example is to compare values

    obtained by these two methods. Also verification of influence of the finite element size on

    the accuracy of the results is performed.

    Figure 5-4. Cubicle in perspective projection. Explosive charge in red, analysed wall in blue.

    16H ft

    32L ft

    6h ft

    12l ft

    5.33AR ft

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    47/102

    47

    Figure 5-5. Plan and section view of the cubicle [7].

    5.2.2. Analytical solutionThe analytical solution is performed using method described in UFC [7]. This approximate

    method has been developed using theoretical procedures based on semi-empirical blast

    data and on the results of response tests on slabs [7].

    The subject of the task is to calculate average peak reflected pressure and average

    scaled reflected impulse on the side wall of a structure, which accordingly to the

    nomenclature given by UFC is described as a fully vented, three-wall cubicle. A TNT

    explosive charge weights 245lbs. Additionally peak incident pressure in line between the

    charge and the wall, and peak reflected overpressure on the side wall in 10 points are

    calculated.

    Data:

    16 4.87

    32 9.75

    6 1.83

    12 3.66

    5.33 1.62A

    H ft m

    L ft m

    h ft m

    l ft m

    R ft m

    Charge weight:

    W 245 111.13lbs kg

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    48/102

    48

    Calculation of peak incident overpressure

    For a freely chosen point located 1m from the charge, the maximum pressure is calculated:

    1 3 1 3 1 3

    1 3.283.28

    0.5243245

    R m ftftR

    ZW lb

    Peak incident overpressure can be calculated from Table 2-7 [7]:

    0.5243 2150 14.82soP Z psi MPa

    All the results are shown in Table 5-2.

    Table 5-2. Peak incident overpressure for different distances from the charge.

    Distance fromthe charge R [m]

    Scaled distance Z1 3

    ftlb

    Pso [psi] Pso [MPa]

    0.29 0.15 6850 47.23

    0.39 0.20 5500 37.92

    048 0.25 4650 32.06

    0.58 0.30 3950 27.23

    0.66 0.35 3370 23.24

    0.76 0.40 2950 20.34

    0.86 0.45 2630 18.130.95 0.50 2280 15.72

    1.05 0.55 2030 14.00

    1.14 0.60 1800 12.41

    1.24 0.65 1695 11.69

    1.33 0.70 1590 10.96

    1.43 0.75 1449 9.99

    1.53 0.80 1270 8.76

    1.61 0.84 1150 7.93

    Calculation of peak reflected overpressure

    Peak reflected overpressure has been calculated on the side wall in 10 points (9 of them are

    distributed evenly) as in the Figure 5-6. The distance from each point to the edge of the

    wall has been set to 0.5m.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    49/102

    49

    Figure 5-6. The view from the centre of the cubicle towards the wall on the left side

    (according to Figure 5-5) - scheme showing analysed points on the wall.

    Calculation scheme for point 1 is presented below. The results for all points are given in

    Table 5-5.

    Coordinates of point 1 relative to the charge:

    5.60

    1.62

    2.55

    x m

    y m

    z m

    Distance to the charge:

    2 2 2 6.36R x y z m

    Scaled distance Z:

    1 3 1 3 1 3 1 3

    6.361.323 3.335

    111.13

    ftR mZ

    W kg lb

    For Z=3.335 one can read from Figure 2-7 [7] peak incident overpressure:

    77soP psi

    Angle of incidence :

    1

    2 2tan 75.21y

    x z

    Basing on peak incident overpressure and angle of incidence a peak incident overpressure

    multiplier can be obtained (Figure 2-193 [7]):

    1.2rC

    Peak reflected overpressure:

    1.22 77 92.4 0.637r r soP C P psi MPa

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    50/102

    50

    Table 5-3. Summary of peak reflected overpressure in chosen points on the side wall.

    Pr [psi] Pr [MPa]

    Point

    1 92.4 0.6372 924 6.371

    3 283.9 1.957

    4 561 3.868

    5 4408 30.392

    6 119 0.820

    7 102.5 0.707

    8 3900 26.890

    9 462 3.18510 10005 68.982

    Calculation of average peak reflected pressure and scaled average unitreflected impulse

    Average peak reflected pressure and scaled average unit reflected impulse; unlike their

    peak reflected equivalents are assumed to act across the whole face.

    Calculation of chart parameters h/H, l/L, L/RA, and the scaled distance ZA:

    1 3

    1 3 1 3

    0.3750.375

    6.00

    2.00

    5.330.85

    245

    A

    AA

    h Hl L

    L R

    L H

    RZ ft lb

    W

    Interpolation is required for ZA, L/H, l/L, and h/H.

    Number of adjacent reflecting surfaces N = 2. Then:

    6.00

    0.8523

    A

    A

    L R

    Z

    The values of rP and1 3

    ri W can be determined and tabulated from Table 2-3 [7] for

    calculated above AL R and AZ and the following variables:

    0.625,1.25,2.50,and5.00

    0.10,0.25,0.50,and0.75

    0.10,0.25,0.50,and0.75

    L H

    l L

    h H

    The results are given in Tables 5-4 and 5-5.

  • 7/30/2019 MSC - Pressure evolution inside complex corridor system induced by blast action.pdf

    51/102

    51

    Table 5-4. Average pressure rP , part 1.

    h H 0.10 0.25

    l L 0.10 0.25 0.50 0.75 0.10 0.25 0.50 0.75

    0.625L H 462 569 598 569 533 665 701 665

    1.25L H 749 932 980 932 943 1178 1238 1178

    2.50L H 1200 1488 1562 1488 1432 1796 1881 1796

    5.00L H 2032 2519 2635 2519 1870 2334 2437 2334

    Figure [7] 2-64 2-65 2-66 2-67 2-68 2-69 2-70 2-71

    Table 5-4. Average pressure rP , part 2.

    h H 0.50 0.75

    l L 0.10 0.25 0.50 0.75 0.10 0.25 0.50 0.75

    0.625L H 546 681 718 681 533 665 701 665

    1.25L H 1017 1267 1333 1267 943 1178 1238 1178

    2.50L H 1609 2028 2120 2028 1432 1796 1881 1796

    5.00L H 1987 2456 2563 2456 2623 3119 3210 3119

    Figure [7] 2-72 2-73 2-74 2-75 2-76 2-77 2-78 2-79

    Table 5-5. Average unit impulses 1 3ri W , part 1.

    h H 0.10 0.25

    l L 0.10 0.25 0.50 0.75 0.10 0.25 0.50 0.75

    0.625L H 73 71 70 66 65 61 59 55

    1.25L H 96 92 90 84 96 92 90 83

    2.50L H 126 121 121 111 139 131 129 120

    5.00L H 172 164 164 153 167 153 154 143

    Figure [7] 2-113 2-114 2-115 2-116 2-117 2-118 2-119 2-120

  • 7/30/2019 MSC - Pressure evolution inside complex corr