Anda di halaman 1dari 74

Numerical simulation of ballistic

impacts on ceramic material


A.P.T.M.J. Lamberts
MT07.33

Committee:
prof. dr. ir. M.G.D. Geers
dr. ir. J.A.W. van Dommelen
dr. ir. H.C. de Lange
ir. A.T.M.J.M. Huizinga
Eindhoven University of Technology
Department of Mechanical Engineering
Materials Technology
PDE Automotive B.V.
Computer Aided Engineering
Eindhoven, August 22nd , 2007

Contents
Samenvatting

IV

Summary

VI

1 Introduction

2 State of the art: ceramic armouring

2.1

Ceramic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Ceramic armour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3

Bullets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Material characterisation

13

3.1

Plate impact experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.2

Bar impact experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.3

Penetration experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4 Bullet impact experiments

17

5 Material modelling

21

5.1

Simha model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5.2

JohnsonHolmquist ceramic models . . . . . . . . . . . . . . . . . . . . . 22

5.3

Microphysical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
I

II

CONTENTS
5.4

Material model selection . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

5.5

JohnsonHolmquistBeissel model . . . . . . . . . . . . . . . . . . . . . 25

5.6

Material models for metals

. . . . . . . . . . . . . . . . . . . . . . . . . 28

6 MSC.Dytran code

31

7 Model validation

35

7.1

Plate impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

7.2

Thick target impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

8 Numerical issues in MSC.Dytran

43

8.1

Voids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

8.2

Bullet impact fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

8.3

Depth of penetration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

9 Conclusion and recommendations

53

9.1

Conclusion

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

9.2

Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Bibliography

59

Acknowledgement

61

A Experimental results

63

B Additional pressure

65

Samenvatting
Binnen PDE Automotive, onderdeel van de Benteler Automotive Company, is een
speciale divisie die zich bezig houdt met het ontwikkelen en bouwen van gepantserde
voertuigen. Hiervoor wordt een zelf ontwikkeld pantserstaal gebruikt. Om het
ballistisch beschermingsniveau te verhogen, is het toepassen van eenvoudigweg
dikkere staalplaten geen optie vanwege een te hoge toename van het voertuiggewicht.
Keramische tegels daarentegen bieden het voordeel een lagere dichtheid te hebben,
maar toch zeer hard te zijn. Door deze tegels op pantserstaalplaten te lijmen, kan
eenzelfde of zelfs hogere beschermingsgraad verkregen worden met een lager gewicht
per oppervlakte eenheid, in vergelijking tot het gebruik van pantserstaal alleen. Om
te voorspellen wat er gebeurt als een kogel een keramiek/pantserstaal doel treft, wil
PDE Automotive gebruik gaan maken van eindige elementen analyses. In het eindige
elementen programma MSC.Dytran, dat binnen PDE wordt gebruikt voor ballistische
simulaties uit te voeren, is echter geen geschikt keramiek model aanwezig. Het doel
van deze afstudeeropdracht is dan ook, om een geschikt keramisch materiaal model te
implementeren, dit vervolgens te valideren en uiteindelijk keramisch pantsermateriaal
te simuleren.
Uit een literatuurstudie kwam naar voren, dat het JohnsonHolmquistBeissel
model (JHB) het meest geschikt was om het gedrag van keramisch materiaal te
beschrijven. Daar keramisch materiaal ook nog enige sterkte onder compressie bezit
nadat het gefaald is, werden de intacte en gefaalde materiaal sterktes door twee
afzonderlijke vergelijkingen beschreven. Een schade parameter bepaalde wanneer
welke vergelijking gebruik dient te worden. Daarnaast was ook een specifiek model
aanwezig om de hydrostatische drukvolume respons te beschrijven. Met behulp van
usersubroutines, werd het JHB model in MSC.Dytran gemplementeerd.
Om de implementatie van het JHB model te valideren, werden zogeheten plate
III

IV

CONTENTS

impact simulaties uitgevoerd. Een keramische plaat werd op een andere keramische
geschoten, waardoor een schokgolf gegenereerd werd in beide platen. De resultaten
verkregen uit deze simulaties toonden een goede overeenkomst met experimentele
resultaten, verkregen uit de literatuur. Daarnaast werden ook thick target impact
simulaties uitgevoerd, waarin een lange staaf werd afgevuurd op een dik blok keramiek.
De gegenereerde resultaten vertoonden wederom grote gelijkenis met experimentele
waarden. Ook het karakteristieke schadeverloop in het keramiek, zoals radiale scheurgroei en het ontstaan van een schadeconus in het gebied vo
or het projectiel, was
duidelijk zichtbaar. Tijdens deze simulaties, kwam ook een beperking naar voren van
het gebruik van een Euler gebaseerde aanpak. Materiaal dat met elkaar in contact
staat, maar geen enkele binding met elkaar heeft, kan niet correct gemodelleerd
worden in MSC.Dytran bij gebruik van de Euler methode, omdat materiaal met
samenvallende oppervlaktes geacht wordt aan elkaar vast te zitten.
Tijdens simulaties van kogelinslagen op pantserstaalplaten, om nieuwe materiaal
parameters voor het pantserstaal te verkrijgen, bleek dat MSC.Dytran onjuiste
resultaten genereerde. Bij gebruik van de axiaal symmetrie optie traden onjuiste void
fractie berekeningen op. Dit resulteerde in onverwacht materiaal falen. Daarnaast was
het niet mogelijk een eenduidige materiaal parameterset aan te wijzen. Resultaten
van depth of penetration simulaties, uitgevoerd in MSC.Dytran met gebruik van het
JohnsonCook model, vertoonden significante verschilllen met resultaten uit andere
eindige elementen programmas. Na onderzoek werden twee onjuistheden gevonden.
Het opnieuw uitvoeren van de berekeningen met de aangepaste programmatuur
resulteerde echter niet in betere resultaten.
Voor model validatie of voor het verkrijgen van keramische material parameters, werden ballistische experimenten op keramiek/pantserstaal doelen uitgevoerd.
Vanwege het feit dat het staal niet correct kon worden gemodelleerd, was het echter
niet mogelijk deze experimenten te simuleren.

Summary
Within PDE Automotive, a part of the Benteler Automotive Company, armoured passenger cars are developed and built with in house developed specialised armour steel.
To improve the ballistic protection level, ceramic tiles are used as addon armour. The
advantage of using ceramic material instead of simply adding more armoured steel,
is the fact that it is less dense but still providing excellent armouring capabilities.
Therefore an equal or even higher ballistic performance can be achieved with a notably
reduced weight per area. To predict the impact of various projectiles on the ceramic/steel armouring, PDE Automotive wants to use finite element analyses. Within
the FEcode MSC.Dytran, which is used by PDE for ballistic simulations, a ceramic
material model was not available. Therefore this thesis objective was to implement a
suitable ceramic material model and to validate and simulate ceramic armour material.
A literature study learnt that the JohnsonHolmquistBeissel model (JHB) was
to be the most suitable for describing the response of ceramic material. As ceramic
material remains to have some compressive strength after failure, this model described
the intact and failed material strength by two separate analytic expressions, both
being strain rate dependent. A damage parameter distinguished between the intact
and failed response. Furthermore, a specific model was used for the hydrostatic
response. By use of usersubroutines, this model has been implemented into the
MSC.Dytran code.
To validate the JHB model implementation, plate impact simulations on silicon
carbide have been performed. In this simulation a ceramic plate was impacted by
another ceramic plate, creating a shock wave in both disks. By using laser, the
material velocity at the back of the impacted plate was measured during experiments.
The generated results were in good agreement with experimental results, obtained
from literature. Next to this, thick target impact simulations, where a thick ceramic
V

VI

CONTENTS

block is impacted by a long rod, also showed satisfying results. The distinctive ceramic
damage propagation, like the forming of radial cracks and the creation of a damage
conoid in front of the projectile, was clearly observed. These simulations, however,
also revealed a limitation of an Eulerian based approach for this type of target setup.
Material interfaces without any bonding cannot be modelled correctly in MSC.Dytran
using an Eulerian approach, since materials with coinciding surfaces are considered to
be bonded together.
During simulations of bullet impacts on armour steel plates, for obtaining new
steel material model parameters, the MSC.Dytran code generated incorrect results.
When using the axial symmetry option, inconsistencies in the void fraction calculation
created unexpected material cracking. Beside this, it appeared not to be possible of
obtaining unique material parameters. Depth of penetration simulations, performed
in MSC.Dytran using the JohnsonCook material model, did not generate the same
results as other FEcodes. Two inconsistencies in the code were found, but calculations
with the adjusted program did not compute more agreeable results.
For validation purposes or for obtaining material parameters, bullet impact experiments on various ceramic/steel targets have been conducted. Because the backing
material could not be modelled correctly in MSC.Dytran, these experiments could not
be used for model validation nor for obtaining ceramic material model parameters.

Chapter 1

Introduction
Wars, criminal activities and violence have always been part of our society since the
beginning of mankind. Thereby all kinds of weapons are used to harm the opponent.
Because of this, there has always been the need for protection against these attacks.
Throughout the centuries, weapons have been improved to be able to attack from a
larger distance, to improve the accuracy of attacks and to increase injury capabilities.
In return, armouring techniques needed to be upgraded to follow these developments
in order to withstand the increasing impact forces. Today, specialised armoured steel
is used to provide adequate protection against bullet impacts. Police and military
personnel are equipped with lightweight armour vests made of Kevlar or ceramics. But
not only police or military forces are faced with the danger of being attacked by use of
fire arms. For example representatives of governments, public figures or UN employees
have to be protected from being assaulted by activists or terrorist groups as well. This
need for armoured protection is still increasing, since these acts of violence are not
constrained by any country borders. Car manufacturers and companies, specialised in
ballistic protection, anticipate to this by providing the necessary armour protection.
One of these suppliers is PDE Automotive, which is part of the Benteler Automotive Company, where armoured passenger cars are being developed and built
with specialised armour steel developed in house. Since many types of bullets exist,
they are categorised into classes by their ballistic impact severity. Armoured vehicles
therefore have to be specifically certified according to the class they need to comply
to. The experiments required as well as the certification itself are rather expensive.
The development of these products can be facilitated by prediction of the material
response to bullet impacts on armour steel plates before performing real life testing
1

CHAPTER 1. INTRODUCTION

and certification. Beside the fact that with the use of Computer Aided Engineering
(CAE) the development process is less expensive, it also gives better understanding
of the dynamic behavior and variables at high velocities. In a previous graduation
project at PDE, B. Adams successfully simulated several bullet impacts on armour
steel plates [1]. Also, important knowledge was gained about the bodywork effect.
Having some experience in the field, PDE now intends to develop armoured
civil vehicles with ballistic protection of a higher level. By using only steel for this,
the increased weight of the car would be unacceptable. The complete body and
suspension of the car would have to be modified and a much stronger engine would be
needed. This is hard to achieve with a standard civil vehicle without compromising
the objective not to affect the normal appearance of the vehicle. To increase the
ballistic performance combined with a smaller increase in vehicle weight, the use
of addon ceramic tiles is being considered. As with the steel armour, there was a
need to simulate the behaviour of this type of armour impact situations as well.
In MSC.Dytran, the software code used by PDE for simulating bullet impacts, no
ceramic material model was available. The objective of the current graduation project
was to implement a suitable material model and to validate and simulate ceramic
armour material.
A literature study has been conducted about the state of the art concerning
the use of ceramics in several applications, especially ballistic protection and the
sort of experiments that are performed to investigate the ceramic material behaviour
at high impact conditions. Various ceramic material models have been compared
from which one is selected to be implemented into the software code. After this, a
validation of the predictive capabilities of the model for impact situations is presented.
Subsequently several issues with the software programme will be discussed and
probable causes of incorrect computational results will be given. Finally, the present
state of the ability at PDE Automotive to simulate bullet impacts on ceramic/steel
armour is given as well as recommendations for future work.

Chapter 2

State of the art: ceramic


armouring
2.1

Ceramic material

Next to metals and polymers, ceramics constitute one of the three main material classes.
They can be defined as compound material between metallic and non-metallic elements
having interatomic bondings that range from purely ionic to totally covalent having a
hard but brittle character. The most ordinary and well-known ceramic materials are
traditional ceramics. They primarily consist of raw materials such as clay, cement or
glass. Products that are considered to be made of traditional ceramics are for example
pottery, porcelain, bricks and tiles. Since the 1950s significant progress has been made
in the understanding of the fundamental character of ceramic material and the phenomena that are responsible for some of their unique properties [2]. Consequently, the
variety of applications has been greatly extended. In table 2.1, an overview is given of
the most common categories with some examples of applications for each category.
Manufacturing process
Ceramics are formed from a variety of compounds, usually a metallic and nonmetallic
element, such as aluminium and oxide (alumina), calcium and oxygen or silicon and
nitrogen. Ceramic products that use naturally occurring minerals first must undergo
special processing in order to control purity, particle size and heterogeneity, before
going to production. This is an important part of the manufacturing process, since
the material structure greatly influences the final properties of the finished material.
For chemically prepared powders no such treatment is needed, because they can be
3

CHAPTER 2. STATE OF THE ART: CERAMIC ARMOURING

Table 2.1: Industrial use of ceramics [3].


Category
Structural clay products

Examples
Brick, sewer pipe, roofing tile, clay floor and wall
tile

Whitewares

Dinnerware, floor and wall tile, sanitaryware ,electrical porcelain, decorative ceramics

Refractories

Brick and monolithic products are used in iron and


steel, nonferrous metals, glass, cements, energy
conversion, petroleum and chemical industries

Glasses

Flat glasses (windows), container glasses (bottles),


pressed and blown glasses (dinnerware), glass fibres
(home insulation) and advanced/specially glass
(optical fibres)

Abrasives

Natural (garnet, diamond, etc.) and synthetic (silicon carbide, diamond, fused alumina), abrasives
are used for grinding, cutting, polishing

Cements

Used to produce concrete roads, bridges, buildings,


dams and the like

Advanced Ceramics

Structural (bioceramics, armouring, engine components)


Electrical (insulators, integrated circuit packages)
Coatings (engine components, cutting tools)
Chemical and environmental (membranes, filters,
catalysts)

controlled with precise composition and particle size.


As-mined raw material usually has to go through a milling or grinding operation to reduce the particle size and create a powdered product. By the addition of
water and other ingredients, the minerals become highly plastic and pliable and may
therefore be formed without cracking. Two common shaping techniques are utilised:
hydroplastic forming (e.g. extrusion) and slip casting. The liquid which was added
to assist in the forming operation is removed in a drying process. This is manifested
as shrinking. Finally, this so-called green body is baked in an oven between 900
and 1400 , called firing or sintering, where diffusion processes cause the body

2.1. CERAMIC MATERIAL

Figure 2.1: Overview of production methods for ceramic materials [2].


to shrink further and strength is increased. Another production method is powder
pressing. A powdered mass is compacted into a desired shape by pressure. Differently
sized particles are mixed in appropriate proportions to maximise the compaction and
minimise the fraction of void space. As the powder particles do not deform plastically,
usually a small amount of water or other binder is added to lubricate the particles
as they slide along each other. After the pressing operation, the product needs to be
fired. The powder particles coalesce into a more dense mass by a sintering process. A
complete overview of ceramic production methods is shown in figure 2.1.

Ceramic properties
The properties of a material are dictated by the types of atoms present, the types
of bonding between the atoms and the way the atoms are packed together. Ceramic
materials can be crystalline or amorphous and are usually ionic or covalently bonded.
A material held together by either bonding type will tend to fracture before significant plastic deformation takes place, which results in a poor toughness. Because these
materials also tend to be porous, the pores and other imperfections lead to stress concentrations, decreasing the toughness even further. These imperfections combine as
the material is loaded, resulting in brittle failure. The chemical bond for metals on the
other hand, called metallic bond, is much weaker than the covalent and ionic bonding,
leading in general to a much more ductile failure.

2.2

CHAPTER 2. STATE OF THE ART: CERAMIC ARMOURING

Ceramic armour

Advanced ceramics are used for a broad range of innovative applications, from artificial
bones or complete engines to space shuttles. They must meet the highest quality
criteria because they are specially selected for their many excellent characteristic properties: high melting point, oxidation resistance, high hardness, being non magnetic,
chemically stable and low weight. The use of ceramics in ball bearings, for example,
has proven to be very useful because they are much less dense than other materials,
which helps to reduce centrifugal forces and increase the maximum rotation speed.
Also in comparison to the steel alternative, a ceramic bearing is smoother, harder,
has a higher tolerance, shows less wear because of reduced friction and has a longer
operating life.
Another application of ceramics if the use for ballistic protection of military
personnel and vehicles against gun and rifle ammunition, exploding mines and mortar
fragments. In a hostile situation, it is most important not to be hit by a projectile at all.
If this did not succeed, proper armouring is required. Key feature of ceramic material
is the low density in comparison to armoured steel; An equal or even higher ballistic
performance can be achieved with a notably reduced weight per area. This results
in a better manoeuvrability and consequently increases the ability of a rapid escape
when being attacked. Whereas conventional armour protection vests are not sufficient
for direct gun fire, ceramic monolithic plate inserts do provide adequate protection.
Ceramicpolymer composites fully comply to the requirements for protection levels
of the upper body. To provide some additional comfort to the wearer, the plates can
be adapted to the body shape by means of single or doublecurved tiles (see figure
2.2). For the use in armoured vehicles, like tanks and transport vehicles, ceramics are
applied as a socalled addon armour (see figure 2.3). The front side of the composite
armour consists of ceramic material. For the backing different materials can be used,
depending on the application. The small ceramic tiles are glued on the backing plate in
a stretcher bond with broken joints. On top of this, different types of top layers can be
used, like foil or a thin metallic plate. In armour vests, highperformance polyaramide
(e.g. Twaron) or polyethylene fibres are used. In vehicles, a metal backing of steel
or aluminium is used. The contact layer between the ceramic layer and the backing
material is of critical importance for the whole system. Exclusive highperformance
adhesives are to be used, to prevent the individual layers from loosening when a bullet
strikes the armour system.

2.2. CERAMIC ARMOUR

Figure 2.2: Armour vest with ceramic insert [4].

Figure 2.3: Addon armour setup, consisting of ceramic tiles on top of a layer of glue
and a metal backing [4].
To provide maximum protection in the case of multihit firing, the ceramic front is
based on 50x50 mm or 100x100 mm tiles. When impacted, only single tiles fracture
whereas a monolithic insert can develop a major crack, which reduces the impact
resistance considerably for subsequent shots. The thickness to be chosen depends on
the ballistic performance required and is usually between 5 and 15 mm. Different
shapes, bore holes, or chamfered edges are also available, but need to be custom
produced by a diamond cutting machine.

CHAPTER 2. STATE OF THE ART: CERAMIC ARMOURING

Ceramic materials used for ballistic protection are required to meet the following properties:
High hardness
High fracture resistance
High E-modulus
Low weight per area

The requirements for a ceramic material which is to be used for ballistic applications
are met by several materials. The most important materials which are currently used
are:
Boron Carbide (B4 C)
Silicon Carbide (SiC)
Silicon Nitride (SiN)
Alumina (Al2 O3 )
Aluminium nitride (AlN)

From these, Aluminium Oxide, also known as Alumina, is the most important. This
is due to the fact that its raw material is relatively low cost, that the production
technique is well mastered and that the sinter process can be carried out in air
atmosphere. The only disadvantage is that the density is higher than for the other
materials. Aluminium oxide is the main component of bauxite, the principal ore (i.e.
raw mining material) of aluminium. It is available in different purities ranging from
85% to 99.9%. For ballistic protection, the higher grades (starting from 95%) are more
favourable because of their higher penetration resistance, but are also more expensive.
For ballistic protection of objects which need to be extremely light, for example
helicopters and airplanes, the more expensive but lower density materials as Boron
Carbide and Silicon Carbide are used. Two kinds of silicon carbide are available in
the market today; S SiC (sintered silicon carbide) and LPS SiC (liquid phase sintered
silicon carbide). The difference is that for the LPS process, the sintering temperature
is above the melting point of some of the elements.
The ceramic production technologies should meet high requirements when ballistic applications are concerned. For all components there should be an absence of

2.3. BULLETS

cracks and pores and the finished tiles are only allowed to show very little damage
at the edges. Important is also the reproducibility and errorfree production of
all individual tiles to ensure that extreme tight tolerances regarding dimension,
plane parallelism and evenness are achieved. Only if all these requirements are met,
closejointed packing can be guaranteed. A joint wider than 0.3 mm will already be a
weak point which may have a negative influence on the ballistic strength. Beside the
geometrical properties, the physical properties also have little tolerance.

2.3

Bullets

There is a wide variety of bullets, each designed for a specific purpose or target type.
In general, two major groups can be distinguished: bullets for wound ballistics and
for armour piercing. Bullets designed to incapacitate personnel or kill animals cause
an extensive wound by crushing, lacerating or displacing body tissue. When a bullet
strikes into body tissue, a temporary cavity is formed around the crushed tissue
which results from the acceleration and stretching of soft tissue radial to the wound
track. This temporary cavity affects solid organs such as the liver and kidneys and
the suction effect draws contaminants such as clothes and bacteria into the wound
track. Wholly lead bullets deform on impact, resulting in high levels of wounding.
The later developed full metal jacketed bullets were more stable and therefore more
suitable for the longer ranges, but had the disadvantage that they did not deform on
impact. Hollow nose semi-jacketed bullets, the so-called DumDum bullets, caused
very extensive wounding and were therefore outlawed by the The Hague Declaration
of 1899 [5].
Military bullets are only allowed to have a full metal jacket. To still achieve a
high wounding potential, the bullets are specially designed to yaw and tumble in the
body after a certain penetration distance. Bullets designed for armour piercing are
used to penetrate hardened armour targets such as body armour, vehicle armour,
concrete, tanks and other defences. They typically consist of a brass or ordinary steel
jacket with a very hard and stiff penetrator like hardened steel, tungsten-carbide or
depleted-uranium. Because of the diversity in bullets, armour materials have to be
classified by their ballistic protection level. Therefore different standards are developed
like EN 1063 (European), NIJ (US), UL (US), DIN or STANAG. In table 2.2 an
overview is given of the European EN 1522 class, which is used within PDE Automotive.

10

CHAPTER 2. STATE OF THE ART: CERAMIC ARMOURING

Table 2.2: Euronorm: EN1522 [6].

Class Type of
weapon
FB 1
FB 2
FB 3
FB 4
FB 5
FB 6
FB 7
FJ1 =
FJ2 =

Calibre

Rifle
22 LR
Hand gun 9 mm Luger
Hand gun 357 Mag.
Hand gun 357 Mag.
Hand gun 44 RemMag
Rifle
5.56 x 45
Rifle
5.56 x 45
Rifle
7.62 x 51
Rifle
7.62 x 51
Full steel jacket (plated)
Full copper alloy jacket

Type

Bullet
Mass (g)

Test condition
Test
Bullet
Range(m) velocity
(m/s)
10 0.5
360 10
5 0.5
400 10
5 0.5
430 10
5 0.5
430 10
5 0.5
440 10
10 0.5
950 10
10 0.5
950 10
10 0.5
830 10
10 0.5
820 10

L/RN
2.6 0.1
FJ1 /RN/SC 8.0 0.1
FJ1 /CB/SC
10.2 0.1
FJ1 /CB/SC
10.2 0.1
FJ2 /FN/SC
15.6 0.1
FJ2 PB/SCP1 4.0 0.1
FJ2 PB/SCP1 4.0 0.1
FJ1 /PB/SC
9.5 0.1
FJ2 /PB/HC1 9.8 0.1
L = lead
CB = coned bullet
FJ = full metal jacket bullet
FN = flat nose bullet
HC1 = steel hard core (>63 HRC)
PB = pointed bullet
RN = round nose bullet
SC = soft core(lead)
SCP1 = soft core (lead) with steel penetrator

For certification, the bullet impact speed has to correspond to that defined by
the standard of table 2.2. But not only the impact speed, also the entire experimental
procedure should meet specific requirements which are described in [6]. These
requirements range from room temperature and shooting distance to the number of
shots to be fired. Not all the bullets presented in table 2.2 are of interest for PDE
Automotive when ceramic material is to be used. Especially the higher classes FB 6,
FB 7 or even higher are modelled for impact simulations.
Since bullet material components are classified, they have to be obtained by
analysing the material with a Scanning Electron Microscope. From that the material
composition is known which makes it possible to identify the material. Although only
the composition is known, material properties can be obtained from literature. In
figure 2.4 an overview is given of 4 types of bullets used within PDE Automotive.
The SS109 bullet (see figure 2.4(a)) has a steel penetrator and a soft lead core with a
brass jacket. The diameter is 5.56 mm and the total weight is 4.0 gram. The average
shooting velocity is 950 m/s. Having the same dimension but a tungsten carbide core,
the 5.56x45 WC bullet has armour piercing capabilities (see figure 2.4(b)). The M80

2.3. BULLETS

11

bullet (see figure 2.4(c)) has a complete lead core and a brass jacket. With a diameter
of 7.62 mm it is bigger and heavier (9.55 gram) than the SS109 bullet. The average
shooting velocity is 830 m/s. The P80 bullet is the same a the M80, but having a hard
steel core, it is used for armour piercing.

(a) SS109

(b) 5.56x45 WC

(c) M80/P80

Figure 2.4: Overview of bullets [7].

12

CHAPTER 2. STATE OF THE ART: CERAMIC ARMOURING

Chapter 3

Material characterisation
The brittle nature of ceramic material and the dynamic nature of the application makes
it very difficult to perform traditional material characterisation experiments as used for
metals and polymers at low and moderate strain rates, such as a tensile test. Two types
of experiments can be distinguished for obtaining information on ceramics in armour
applications: experiments to evaluate material properties in high strain-rate environments and penetration experiments to evaluate the material behaviour under ballistic
impact conditions. The first type attempts to subject the material to a controlled
high rate loading with a one or twodimensional stress-strain state. The penetration
experiment is a less controlled complex threedimensional loading condition.

3.1

Plate impact experiments

In plate impact experiments [8] a propellant gas gun accelerates a projectile carrying a
disc-shaped sample (i.e. impactor) of the ceramic of interest to velocities up to several
kilometres per second. This impactor undergoes a planar impact with a stationary
sample of the same shape and material, generating a shock wave into both disks. The
impactor and sample dimensions are chosen so that reflected waves from the edges do
not reach the centerline in time to interfere with the recording of data. That way a
uniaxial strain experiment is created. The diameter of the disks thus depends on the
speed of sound of the ceramic used. A thick disk of lithium fluoride is bonded to the back
of the stationary disk (see figure 3.1(a)) and performs as a transparent laser interface
window for a VISAR (Velocity Interferometer for any Reflecting Surface) system, that
measures interface velocity versus time. Some distinct areas can be pointed out. The
initial part of the wave profile with almost zero rise-time is the elastic wave. It travels
13

14

CHAPTER 3. MATERIAL CHARACTERISATION

(a) Plate impact

(b) Wave profile

Figure 3.1: Plate impact experiment [9].


with the longitudinal speed of sound of the ceramic. The break in the curve indicates the
Hugoniot elastic limit (HEL), which separates the elastic wave from the plastic wave.
The HEL is analogous to the yield point in a onedimensional tensile experiment and
thus represents the onset of inelastic behaviour. To calculate the stress at the HEL,
the velocity at the HEL in the wave profile is converted with HEL = 0 Cl v, where
0 is the initial density, Cl the material speed of sound and v the material particle
speed. As variation of this experiment, a spall experiment can be conducted, which is
used to determine the tensile fracture strength of the ceramic. As the impactor hits the
sample disk, a compression waves travels into both disks. In the sample disk, after some
time, the compression wave reaches the lithium fluoride window and, due to the higher
impedance of the ceramic, is reflected as a release wave back to the impact interface
(see figure 3.2(a)). In the impactor the compression waves reaches the free surface and
is also reflected as a release wave. At some point these two waves interact and cause a
fracture in the target. In the velocitytime profile this point can be identified by the
dip v seen in figure 3.2(b), also called the pull back signal. Because the height of the
dip is proportional to the spall strength, it can be related to the tensile fracture stress.

3.2

Bar impact experiments

In this experiment, a long ceramic bar (length/diameter ratio about 46) is impacted by
either another bar or a plate. The impacted bar can have two setups; confined and unconfined (see figure 3.3). Just after impact this experiment has much resemblance with

3.3. PENETRATION EXPERIMENTS

(a) Wave interaction

15

(b) Pullback signal

Figure 3.2: Tensile fracture strength experiment [9].


the plate impact, but release waves from the edges of the bar cause a twodimensional
stress and strain loading. Further down the bar, the propagating wave has become a
onedimensional stress wave travelling at the bar wave speed. In figure 3.4 a typical
stresstime profile is shown where the peak stress is a measurement of the maximum
stress the rod can support. Results of unconfined and confined experiments are different because the confinement delays the arrival of unloading waves from the edges of
the bar and therefore delays failure.

(a) unconfined

(b) confined

Figure 3.3: Bar impact experiment [9].

3.3

Penetration experiments

Penetration experiments have such a setup that the investigated ceramic can qualitatively be tested for real armour applications The goal of penetration experiments
is to analyse the actual ceramic armour application behaviour. A long rod projectile

16

CHAPTER 3. MATERIAL CHARACTERISATION

Figure 3.4: Stresstime profile from bar impact experiment.


penetrates a ceramic body, which can have different dimensions and setups. In a semi
infinite penetration experiment, as shown in figure 3.5(a), the target consists of a thick
ceramic block covered with a steel or aluminium plate. During testing, the length of
the penetrator can be measured in time by Xray. Also the final depth of penetration is
measured. A different type of penetration experiment is a depth of penetration (DOP)
experiment, shown in figure 3.5(b). A ceramic tile is bonded to a thick metal backing in
a confined or unconfined configuration. Experimental analysis is done by measuring the
ballistic performance of the ceramic by the residual penetration depth. This means the
final penetration into the metal block is compared to the reference penetration depth
of the same projectile into the block without the ceramic tile. A special feature is the
socalled interface defeat, when the impactor is completely defeated by the ceramic
tile, resulting in no residual penetration into the steel backing plate.

(a) Semi-infinite

(b) DOP

Figure 3.5: Penetration experiments [9].

Chapter 4

Bullet impact experiments


For model validation purposes bullet impact experiments on ceramic/steel targets
have been conducted by the TNO Defence, Security & Safety department at the
Laboratory for Ballistic Research in Ypenburg.
The ceramic tiles, 95% pure aluminium oxide used were provided by Morgan Advanced
Ceramics. The tile thicknesses were 5 mm, 7 mm and 10 mm, all having the same
surface area of 50x50 mm. The pink colour does not originate from the chemical
composition but is only used for identification of the specific ceramic. In table 4.1
an overview is given of the ceramic material properties. The steel armour backing
Table 4.1: Material properties of Alumina 95% [10].
Material
Density
Bulk modulus
Shear modulus
Porosity
Tile size
Tile thickness

Al2 O3
= 3740 kg/m3
K = 320GPa
G = 130GPa
0% (fully dense)
50x50 mm
5, 7, 10 mm

material was BSEC510 from Benteler Automotive Company. Plates of 400x400 mm


were cut, having three different thicknesses of 3.0 mm, 5.0 mm and 6.3 mm. Material properties of this material are classified and are therefore not given in this report.
Different ceramic/steel combinations were impacted by 2 types of bullets (P80
and 5.56x45 WC) with varying velocities. These combinations were chosen specifically
to generate results varying from a complete bullet stop by the ceramic tile to a full
17

18

CHAPTER 4. BULLET IMPACT EXPERIMENTS

Table 4.2: Experimental procedure.


Backing
thickness
[mm]
3.0

Ceramic
thickeness
[mm]
5.0

Bullet
type
5.56x45 WC
P80

5.0

5.5

5.56x45 WC
P80

6.3

7.0

5.56x45 WC
P80

6.3

10.0

5.56x45 WC
P80

impact
speed
[m/s]
850, 900,
950
750, 800,
850
850, 900,
950
750, 800,
850
850, 900,
950
750, 800,
850
850, 900,
950
750, 800,
850

Objective

Complete
bullet
penetration

Bullet stop
penetration

and

Complete
bullet
stop with bulging
backing plate
Complete
bullet
stop by ceramic
tile

bullet penetration. The difference in speed between these extremes should be as


small as possible, because the material model can than be validated with the highest
precision. In table 4.2 an overview is given of the ceramic/steel combinations tested,
bullets used and the speeds of the bullet impacts.
Polyurethane was used to glue the ceramic tiles onto the steel backing. Because
of the limited number of ceramic tiles they are not packed close together. Also from
a simulation point of view this setup is better defined than when close packed tiles
were used. For the same reason, the bullets impact a bare ceramic tile, without any
plate or tape in front of the tile. Between the individual tiles wooden boards are glued
to prevent ceramic debris from damaging intact neighbouring tiles. In figure 4.1 the
experimental configuration is shown.
During testing it became clear that the target material responded differently
than the objective given in table 4.2. Therefore the bullet impact speeds were adjusted
to generate, when possible, perforations and stops for all target combinations in a close
speed range., which is useful for simulation purposes. In appendix A all experimental

19

Figure 4.1: Experimental setup for ceramic/steel target.


results are given as well as a discussion of the results.
Whether the ceramic tile completely stopped the bullet or if penetration occurred, the complete tile broke up in pieces and fell off the backing plate. Although
larger pieces could be recovered, they could not be used for later analysis, since they
did not originate from the area of interest on the tile, which is located closely around
the crack.

20

CHAPTER 4. BULLET IMPACT EXPERIMENTS

Chapter 5

Material modelling
The response of ceramic material to impact loading is of a complex nature. Material
models should be able to represent the evolving macromechanical material properties
which result from the detailed and complex micromechanical structure. Damage evolution, due to stress concentrations at triple points of grains and in the intergranular
glassy phase, should be taken into account. Also the model should have the ability to
provide some compressive strength after the material has failed.
First, an overview of existing ceramic materials model is given. Then, the models will
be compared and finally one is selected. This selected model will be discussed in more
detail in section 5.5.

5.1

Simha model

A phenomenological model of the response of a ceramic material under impact loading


conditions is developed by Simha et al. [11] (see figure 5.1) The yield stress, y , is a
d
function of the pressure p and the deviatoric strain rate . During plastic deformation,
damage accumulates until failure (D = 1). The yield stress of the material is a weighted
sum of the intact material strength, intact , and the failed strength, f ailed . Because
0 D 1 all intermediate states of failure can be described and the material is
softened gradually.

y = intact (1 D) + f ailed D +

3
,
2

where
f ailed = min[p, max ]

= 0 exp[1 (p pHEL )],


21

(5.1)

22

CHAPTER 5. MATERIAL MODELLING

Figure 5.1: Simha strength model (without strain rate effect).


with the slope of the curve shown in figure 5.1, 0 and 1 material parameters and
pHEL the pressure at the Hugoniot elastic limit, which is the onset of inelastic behaviour. The term controls the contribution of the effective deviatoric strain rate,
d
to the strength of the material. Taking this rate-dependent term into account is
a phenomenological contribution of micro-crack sliding, dislocation activity and grain
boundary sliding within the material during deformation. In [11] this model is used for
plate impact simulations, depth of penetration simulations and interface defeat computations. A very characteristic feature of this model, is the higher material strength
of failed material with respect to intact material.

5.2

JohnsonHolmquist ceramic models

One of the most widely used models for ceramic materials in the ballistic research
field is the JohnsonHolmquist model. The past several years, these authors developed
three ceramic models [12, 13, 14] (see figure 5.2). All these models (JH1, JH2 and
JHB) are based on two sets of curves of yield stress vs pressure, i.e. intact and failed.
Each curve depends on plastic strain and plastic strain rate. A damage variable, D,
defines the level of fracture. For the JH1 and JHB model, the intact material curve
is used prior to fracture (D < 1.0). Once fracture has occurred (D = 1.0) the failed
material curve is used. The JH2 model also has an intact and failed material curve,
but the model is gradually softened as damage accumulates.
Most recently Johnson, Holmquist and Beissel developed the socalled JHB
model [12]. This model is an improved version of the JH1 model. Material strength
and damage are smooth analytical functions of pressure, whereas JH1 uses a piecewise
approximation. In both models the material strength does not decrease until complete

5.3. MICROPHYSICAL MODELS

23

Strength of
intact material (D<1)

P2,2

P1,1

Strength of
failed material (D=1)

Pressure

(a) JH1 model

(b) JH2 model

Strength of
intact material (D<1)

P ,

Strength of
failed material (D=1)

P ,
f

Pressure

(c) JHB model

Figure 5.2: JohnsonHolmquist ceramic models for high pressure, high strain and high
strain rate conditions (at dimensionless strain rate 1.0).
damage (D = 1) has occurred. This in contrast to the JH2 model which gradually
softens the material strength (from intact to failed) as damage accumulates.
The JohnsonHolmquist models have been used to compute plate impact simulations
and long rod penetrations on silicon carbide and aluminium nitride [15, 16, 17]. All
these simulations generated acceptable results.

5.3

Microphysical models

The previously mentioned ceramic material models describe the material response from
a phenomenological point of view. Intact and failed strength, as well as damage accumulation are functions of pressure. However as damage originates from an existing
distribution of micro-cracks, Rajendran and Groove [18] and Addesio and Johnson [19]
developed microphysical models where damage is described by a crack density parameter. Crack nucleation and growth are based on a generalised Griffith criterion. Due to
this microcracking, a shear modulus and bulk modulus reduction was modelled. These

24

CHAPTER 5. MATERIAL MODELLING

models were used to successfully reproduce velocity histories under one dimensional
strain conditions. Modelling steel projectile impact on ceramic targets resulted in less
acceptable results, especially for unconfined experiments [18].

5.4

Material model selection

For many ceramic materials, a change in physical (especially mechanical) properties


can be observed with different physical environments. Compressive strengths and
ductility may be enhanced under an increased pressure, due to the suppressing or
favouring of certain slip systems [20]. Besides that, strength increases with increasing
strain rate.
Microphysical modelling, which is based on evolving microcracks, is one way of
modelling ceramic material. But since this theory is only based on a single crack
propagation and no crack interaction is considered, this way of describing the ceramic
material response remains to be more phenomenological than properly micromechanical based. Therefore, this type of constitutive modelling will not be adopted for
computations of ceramic impact simulations.
When analysing phenomenological constitutive models, they appear to be in
essence a MohrCoulomb type of model [11]. This trend is observed by several
authors. In [21] for example the JH1 model was simplified and approached by a
MohrCoulomb model. Results of depth of penetration simulations were in reasonable
agreement with the experimental data. The ceramic model of Simha [11] appears to
be very appropriate. Computational results are within a few percent of experimental
results and it also generated the interface defeat feature (i.e. no residual penetration
into steel backing in a depth of penetration experiment). One part however of the
model is not consistent with any other constitutive model. In the Simha model it
is assumed that failed material in compression is stronger than intact material. The
reason given for this is that failed material consists of more (but smaller) particles
with therefore more surface area that interacts during deformation [9]. This increases
the internal friction, which results is higher strength. This is contradicting with what
is to be expected, because it also can be stated that failed material has almost infinite
slip possibilities, resulting is less strength than intact material, which has a very
limited number of slip systems. Because of this inconsistency, this model will not be
selected.

5.5. JOHNSONHOLMQUISTBEISSEL MODEL

25

The JohnsonHolmquist models are frequently used in ballistic research of brittle materials because of their relatively easy implementation. In CenturyDynamics
AUTODYN [22] the first two JohnsonHolmquist models (i.e. JH1 and JH2) are
implemented. This software is especially used for simulating impact loading on various
types of structures, like spacecrafts, armour material or brick walls. In LSDyna
[23] these same model are also available to model brittle material behaviour. In
both program codes the JohnsonHolmquist models generate agreeable results which
therefore can provide good insight in the ceramic material response. The latest model
however, the JohnsonHolmquistBeissel model (JHB), has not yet been implemented
in any of these codes since it is rather new. This model is, as already mentioned, very
similar to the JH1 model, both having a discrete damage model. The JH2 model
is somewhat different since it has a continuous damage model. To model the correct
material behaviour, the intact strength of the JH2 model is always higher than that
of the JH1 or JHB model, because the damage reduces the strength as plastic strain
accumulates. This eventually results in similar material strengths for all three models.
However determination of the JH2 parameters is more difficult, since the damage
evolution has to be taken into account. Therefore the JHB model will be selected to be
implemented into MSC.Dytran to model the ceramic material behaviour. This model
is more favourable than the JH1 model, since it has smooth analytical functions.

5.5

JohnsonHolmquistBeissel model

In section 5.4 the JohnsonHolmquistBeissel model (JHB) was found to be the most
suitable for describing the response of ceramic material. This model describes the intact
and failed material strength by two separate analytic expressions (see figure 5.2(c)),
both being strain rate dependent. A damage parameter distinguishes between the intact
and failed response. Furthermore, a specific model is used for the hydrostatic response
[16]. A Von Mises yield criterion is used to decide at which stress a purely elastic
deformation will be followed by elastoplastic deformation. To determine the amount of
plastic strain, an associated flow rule is used.
Intact material strength
The strength of the intact material is given by
h

i
intact = i + max
i [1.0 exp (i (p pi ))] ,

(5.2)

26

CHAPTER 5. MATERIAL MODELLING

imax

Pi,i

Pf,f

fmax

Pressure

Pressure

(a) JHB model intact material strength

(b) JHB model failed material strength

Figure 5.3: Intact and failed material strength for the JHB model
where
i =

i
(max

i
.
i )(pi + T )

(5.3)

i
Here i , max
and Pi are material parameters, where the subscript i indicates intact

material. Pressure, p, is defined as p = 31 tr(). T is the maximum tensile stress of the


ceramic. Equation (5.2) holds for a pressure greater than pi . For smaller pressures, the
strength is a linear function from (p = T ) = 0 to (p = pi ) = i . In figure 5.3(a)
the intact material strength is shown with its characteristic points marked.
Failed material strength
The strength of failed material is represented in a similar way as intact material. For
pressures greater than pf , failed material strength is defined by
h

f
f ailed = f + max
f [1.0 exp (f (p pf ))] ,

with
f =

f
f
pf (max

f )

(5.4)

(5.5)

where f , max and pf are material parameters and the subscript f denotes the failed
material. For pressures smaller than pf the failed strength is a linear function from
(p = 0) = 0 to (p = pf ) = f . In figure 5.3(b) the failed material strength is shown
with the characteristic points marked.
Strain rate dependence
Equations (5.2) and (5.4) hold for a dimensionless strain rate of = 1.0, where
= /0 , with 0 the reference strain rate and the effective strain rate. To include

Failure strain

5.5. JOHNSONHOLMQUISTBEISSEL MODEL

27

* n

p=D (P +T )
1

T 0

Pressure

Figure 5.4: Plastic strain to failure in JHB model.

strain rate dependence, the strength at other effective strain rates is defined as
y = (1.0 + C ln ),

(5.6)

where C is the dimensionless strain rate constant and is the material strength with
= intact, f ailed, denoting intact (eq. 5.2) and failed (eq. 5.4) material, respectively.

Damage
The JHB model has a pressure dependent damage model included. Damage is accumulated in a similar manner as in the JohnsonCook failure model (which will be
discussed in section 5.6) and is defined as:
D=

X p

fp

(5.7)

where p is the current increment in equivalent plastic strain. The plastic strain to
failure, fp , is defined as
fp = D1 (P + T )n ,

(5.8)

i
i
are
and P = P/max
where D1 and n are dimensionless constants. T = T /max

dimensionless pressures. The material fails if damage, D, equals unity or if the tensile
hydrostatic pressure is greater than the maximum tensile pressure, T . Once the material
has failed, no damage will be accumulated anymore and damage is set to D = 1. In
figure 5.4 the plastic strain to failure is shown.

28

CHAPTER 5. MATERIAL MODELLING

Pressure
In compression the hydrostatic pressure-volume response before failure is a polynomial
function of the volumetric compression, :
P = K1 + K2 2 + K3 3 ,

(5.9)

where K1 is the bulk modulus, K2 and K3 are material fitting parameters. The volumetric compression, , is defined as
=

V0

1=
1,
V
0

(5.10)

with V0 and V the initial and current volume respectively and 0 and the initial and
current density. After complete failure has occurred the dilatancy occurs because of a
pressure and/or volume increase [24]. Therefore an additional pressure, P , is added
to equation (5.9). This pressure increase is determined by energy considerations. The
loss of internal elastic energy, U , is converted into potential hydrostatic energy. The
derivation of this is given in Appendix B.
The pressure increase is defined as
q

P = K1 f +

(K1 f )2 + 2K1 U ,

(5.11)

with the fraction of converted internal energy loss (0 1) and f the volumetric
compression at failure.
In tension, the pressure is represented by a linear dependence on the volumetric
strain:
P = K1 .

(5.12)

If the material fails under tension no bulking pressure is computed.

5.6

Material models for metals

SteinbergGuinan
The SteinbergGuinan model [25] is applicable for metals subjected to high strain
rates. In the work of B. Adams [1] this model was used for copper in bullet impact
simulations. The material strength, y , increases with increasing pressure and decrease

5.6. MATERIAL MODELS FOR METALS

29

with increasing temperature. The flow stress is written as:

y = [A + B (p + 0 ) ] 1 + H1

P
1

J3

H2 (T Tr ) ,

where A is the initial yield stress, B and n workhardening parameters, p the equivalent plastic strain, 0 the initial equivalent plastic strain, T the temperature and Tr
the room temperature. H1 and H2 are material parameters.

JohnsonCook model
A model frequently used in ballistic simulations is the JohnsonCook model [26]. It describes the strength of metals at large strains, high strain rates and high temperatures.
An independent term is created for each of these phenomena, which makes characterisation relatively easy. For this reason this model is frequently used in ballistic impact
simulations. The material strength is given by
h

y = A +

B np

"

p
1 + C ln
0

!#

T Tr
Tm Tr

(5.13)

with A the initial yield stress, B the strain hardening coefficient and n the strain
hardening exponent. p is the effective plastic strain rate, 0 the reference strain rate,
C the strain rate coefficient and m the temperature softening exponent. T , Tr and Tm
are temperature, room temperature and melting temperature, respectively.
The JohnsonCook model also has a failure model incorporated, which is based
on Von Mises stress,
, the average of the three normal stresses, m , effective plastic
strain rate, p and temperature, T . Damage, D, is defined as [27]
D=

p
,
f

(5.14)

where

m
f = D1 + D2 exp D3

"

p
1 + D4 ln
0

!#

1 + D5

T Tr
Tm Tr

(5.15)

where D1 ,. . . ,D5 are material parameters. The damage can have a continuous or a
discrete effect on the yield stress. When modelled continuously, the material strength
is weakened during damage accumulation by multiplying equation (5.13) with a factor
(1D). When D equals unity, material strength is zero and the material fails. The same

30

CHAPTER 5. MATERIAL MODELLING

holds for a discrete damage effect, but than material fails instantly when D reaches
unity.

Chapter 6

MSC.Dytran code
For calculating the mechanical response of material during impact simulations, within
PDE Automotive the finite element code MSC.Dytran is used. This program uses an
explicit method to solve the equation of motion, given in equation 6.1 [28]
t + C u t + Kut = F t ,
Mu
e

(6.1)

where M is the mass matrix, C the damping matrix and K the stiffness matrix. The
columns u
t , u t , ut and F t are the nodal accelerations, velocities, displacements and
e

forces, respectively. The equation of motion at the current time t is used to predict a
solution of the acceleration at time t + t. The equilibrium determines the acceleration
at the beginning of the increment and it is assumed to be constant over the time step,
t. Therefore equation 6.1 can be rewritten as [1]:
u
t+t = M 1 F res
t .
e

(6.2)

To advance in time, nodal velocities and displacements are calculated by [1]:


u t+t = u t + t
ut+t

(6.3)

ut+t = ut + tu t+t .

(6.4)

and
e

In figure 6.1 the calculation loop is shown, carried out by MSC.Dytran each time
step, to calculate the material response. Because the accelerations are assumed to be
constant over the time step, the time step has to be chosen small. For stability reasons
it must be chosen smaller than the smallest natural period in the mesh. This means,
31

32

CHAPTER 6. MSC.DYTRAN CODE

Figure 6.1: Diagram of loop carried out by MSC.Dytran each time step [28]

the time step has to be smaller than the time for a stress wave to travel through the
smallest element.
Within the MSC.Dytran code, two solving techniques are available, Langrian
and Eulerian (see figure 6.2). The code can use either one, or both, and can couple
the two types to define an interaction. In the Lagrangian approach, nodes correspond
to material points. During deformation, the grid points move, distorting the element.
In the Eulerian approach, the mesh has a fix reference frame, which means that grid
point are fixed in space and material flows through the elements. The mesh itself is
defined the same as for the Lagrangian approach, only the use is different because the
mesh for the Eulerian approach should be large enough to contain the material after
deformation. Unless specified, the material is not allowed to cross the boundary of the
mesh, which therefore act as a boundary condition. In general, a Lagrangian approach
is used for engineering applications. The Eulerian solver is used for fluids or materials
that undergo very large deformations [28]. Because bullet impact experiments shown
very large deformations, an Eulerian approach will be used to simulate these kind
of experiments. Therefore an explanation will be given of some important numerical
aspects when working within an Eulerian domain.
Following the scheme presented in figure 6.1, at the beginning of an increment

33

(a) Lagrange

(b) Euler

Figure 6.2: Lagrange and Euler approach [28].


the grid point velocities are known. From this, the transported volume V can be
determined by
V = v A t,

(6.5)

where v is the speed of the moving material in normal direction of the boundary
surface,A, the material passes and t is the time step. Now the transported volume
from donor to acceptor element is known, it can be filled up with mass of the donor
element by using the donor material density. When more than one material is present
in an element, the transported volume can be filled up with more than one material.
In order to maintain an interface between different materials, this is done in a
specific order. If the donor and acceptor element have common material, this will be
transported first. If there still remains volume to be filled up, the remaining material
in the donor element will be transported according to the largest material fraction in
this element.
To calculate elemental stresses and strains, different subroutines are used. The
pressure (or hydrostatic stress), is determined by an equation of state subroutine. In
the yield stress subroutine, the yield stress is calculated by selecting one the available
yield models. To initiate damage and/or finally failure, a failure routine has to be used.
When a specific pressure, yield and/or failure model is not present in the MSC.Dytran
library, users can program their own subroutine by socalled user subroutines.
Since the JohnsonHolmquistBeissel model was not included in the MSC.Dytran
material models library, it therefore had to be programmed. Although the user is
free of programming within the user subroutine, some special features, needed for

34

CHAPTER 6. MSC.DYTRAN CODE

incorporating the JHB model, were not available. In cooperation with MSC.Software
the user subroutine for damage was adjusted so that a material failure would not lead
to zero strength, as would happen in the standard code. Also the user subroutine
for the equation of state was modified to be able to account for the additional pressure.
In MSC.Dytran, no conventional axisymmetric elements are available. Although
this shortcoming, axisymmetric can be performed by using a 5wedge. On the
centerline, 6node pentahedral elements are used. For the remaining part of the mesh,
8node hexahedral elements are used. In tangential direction there is only 1 element.
Since the edge length in tangential direction of the pentahedral elements is small, the
time step also becomes very small. Because there is no material flow in this direction,
the time step is much smaller than needed for still being able of generating stable
results. To increase the time step and reduce calculation time, an axial symmetry
option is used. The time step is then determined only by element sizes in radial and
axial direction. Also, this option aligns all the element face normals in tangential
direction, so that all element faces are in one plane. Without the axial symmetry
option, this may not be the case because of roundoff differences.

Chapter 7

Model validation
To validate the implementation of the JohnsonHolmquistBeissel model (JHB), in
MSC.Dytran various simulations have been conducted. These applications differ from
the onedimensional strain response to complex threedimensional behaviour. The simulations are compared to experimental results and to computations performed by other
authors as well. In all examples, silicon carbide (SiC) is considered, for which the constants for the JHB model are shown in table 7.1
Table 7.1: Constants for the JHB model for Silicon Carbide [16].
parameter
Density
Shear modulus
Hydrostatic tensile strength
Intact strength constant
Intact strength constant
Maximum intact strength
Strain rate constant
Failure strength constant
Failure strength constant
Maximum failure strength
Bulk modulus
Pressure coefficient
Pressure coefficient
Damage coefficient
Damage exponent
Energy conversion factor

35

Value
0 = 3215 kg/m3
G = 193 GPa
T = 0.75 GPa
i = 4.92 GPa
Pi = 1.5 GPa
max = 12.2 GPa
C = 0.009
f = 0.10 GPa
Pf = 0.25 GPa
f
= 0.20 GPa
max
K1 = 220 GPa
K2 = 361 GPa
K3 = 0 GPa
D1 = 0.16
n = 1.0
= 1.0

36

7.1

CHAPTER 7. MODEL VALIDATION

Plate impact

Plate impact experiments, as mentioned in chapter 3, are frequently used for model
validation purposes. Grady and Moody performed extensive research on the impact
response of several ceramic materials [29]. Johnson and Holmquist [16] simulated two
silicon carbide plate impact experiments from Grady and Moody [29],using the JHB
model within a Lagrangian finite element approach. The same simulations have been
performed using MSC.Dytran with the implemented JHB model. The thickness of the
silicon carbide target is 9 mm and is backed by a lithium fluoride (LiF) window with a
thickness of 25.4 mm. The silicon carbide projectile with a thickness of 4.5 mm has a
polymethyl methacrylate (PMMA) backing with a thickness of 8.0 mm. Both simulation have the same configurations but differ in speed. One impact is at 2259 m/s, the
other at 1566 m/s.
Table 7.2: Material constants for LiF [30] and PMMA [9].
Material
LiF

0
2638 kg/m3
n
0

G
49 GPa
H1
-3.028 MPa

K
63 GPa
cp
129 J/kg

Material
PMMA

0
1180 kg/m3
K1
6 GPa

G
2.3 GPa
K2
29 GPa

y
190 MPa
K3
13 GPa

A
360 MPa
Tm
1207

B
0

During the experiments Grady and Moody measured the particle velocity of the
samplewindow interface and plotted this against time. In order to compare the computed results with experimental data, the same information needed to be obtained from
the simulation in MSC.Dytran. Since the JHB model is only available in an Eulerian
approach, tracking a single material point is not possible (i.e. material travels from
element to element). Therefore a LagrangeEuler interface has been modelled, with
the window material defined in a Lagrangian mesh. In figure 7.1, the simulation setup
used in MSC.Dytran is shown. To assure axial symmetry, the bottom notes of the Lagrangian mesh are constrained in radial direction. The right bottom node is constrained
in all directions. The free surface at the back of the Lagrange mesh is constrained not
to move in axial direction. The remaining nodes are not allowed to move in tangential
direction. Although this is a simplification of the actual experimental setup, it is however permitted since the recording time is of such short period that this constraining

7.1. PLATE IMPACT

37

Figure 7.1: Plate impact setup with Eulerian and Lagrangian elements
Experiments
JohnsonHolmquist
MSC.Dytran

1000
800
600
400
200
0
0

0.5

1.5

time [s]

(a) 1566 m/s

2.5
6

x 10

Experiments
JohnsonHolmquist
MSC.Dytran

1500

Particle speed [m/s]

Particle speed [m/s]

1200

1000

500

0
0

0.5

1.5

time [s]

2.5
6

x 10

(b) 2259 m/s

Figure 7.2: Silicon carbide plate impact data at two different impact speeds.

does not influence the results. The Eulerian mesh consisted of 22500 elements with an
element size of 0.1x0.2 mm. The Langragian mesh consisted of 29680 elements with
an element size of 0.1x0.25 mm. A SteinbergGuinan model is used to model the LiF
window. To model the PMMA backing an elastic perfectly plastic model is used. In
table 7.2 the material parameters for LiF and PPMA are shown. In figure 7.2, data is
plotted of results obtained from MSC.Dytran using the implemented JHB model, the
experimental data of Grady and Moody and results of Johnson and Holmquist also
using the JHB model. It can be seen that there is good agreement between the individual results. Although the release waves of both simulated results show some differences
with the experimental results, they show very similar behaviour.

38

CHAPTER 7. MODEL VALIDATION

(a) Target configuration


duced from [32]

repro- (b) Wedge


shape
configuration

simulation

Figure 7.3: Thick target geometry.

7.2

Thick target impact

Lundberg et al [31] performed experiments on thick confined silicon carbide targets


impacted by long rods. In figure 7.3, the target geometry is presented. The Eulerian
mesh contains 69080 elements with an element size of 0.1 mm. The projectile is a
tungsten rod of 80 mm in length and 2 mm in diameter. The silicon carbide target is
confined by a high strength steel tube (S7 steel) with a steel top and bottom plug
(4340 steel). The target dimensions are 20 mm in diameter and 20 mm in height.
The steel tube is 28 mm in diameter with a height of 36 mm. The plugs are both
20 mm in diameter and 8 mm in height. The JohnsonCook strength and fracture
model constants to model the materials of the confinement are given in table 7.3. In
figure 7.4 computed results are presented together the experimental data. The three
impact velocities are 1410, 1645 and 2175 m/s and are identical to the velocities used
in the experiments. Also plotted in this figure are results from [32] where the same
experiments were simulated using the JH1 model in a Lagrangian mesh and meshless
particles. The mesh initial only held Lagrangian elements, which were automatically
converted into meshless particles during the course of computations. The simulations
performed with MSC.Dytran and from [32] show similar behaviour. Both generate a

7.2. THICK TARGET IMPACT

39

Table 7.3: JC material constants for 4340 steel, S7 steel and tungsten [26, 27, 17, 32].
Material
o (kg/m3 )
G (GPa)
K1 (GPa)
K2 (GPa)
K3 (GPa)
Tm ()
cp (J/kg)
A (MPa)
B (MPa)
n
C
m
D1
D2
D3
D4
D5

4340 steel
7850
78
164
294
500
1520
477
600
510
0.26
0.014
1.03
0.05
3.44
-2.12
0.0020
0.61

S7 steel
7750
78
164
0
0
1490
475
2000
477
0.18
0.012
1.0
-0.80
2.10
-0.50
0.0020
0.61

Tungsten
17600
124
285
470
335
1768
250
1300
141
0.18
0.016
1.0
0
0.33
-1.50
0
0

slightly to high penetration depth for 1410 m/s and 2175 m/s with respect to the experimental results. The evolution of penetration in MSC.Dytran at 1645 m/s however
differs from the experimental results as well as from the computational results from [32].
Materials defined in an Eulerian based approach that have coinciding surfaces
are considered to be bonded together. At the contact surface there is no friction
defined, only the shear modulus of both materials has influence on the resistance
against movement. Separating the materials is only possible when internal forces are
in opposite direction. This is also the case when a real interface (without any bonding)
is present, but the force needed in simulations is higher, because the material are
considered to be bonded together. When the tungsten cylinder arrives at the interface
of the steel plug and the ceramic block, dwelling occurs (i.e. during a short period
of time the impactor continuous to deform without any penetration into the ceramic
material). During the dwelling, the tungsten therefore has to flow radially between
the contacting surfaces of the ceramic block and the steel plug by separating these
materials from another. In the experimental configuration, the plug and the ceramic
block are placed against each other without any bonding forces. In the Eulerian
simulations there is a bonding though, thereby restricting the radial movement of the

40

CHAPTER 7. MODEL VALIDATION

Ceramic penetration [mm]

25
MSC.Dytran

20
JohnsonHolmquist

15

V = 2175 m/s
Experiment

10
V = 1645 m/s

V = 1410 m/s
0
0

time [s]

4
5

x 10

Figure 7.4: Thick Target impact results at various speeds.

tungsten at the steel plug/ceramic interface. In figure 7.5 is shown what takes place
during the simulation at the interface. Instead of flowing radially the tungsten material
starts to swirl. This has an effect on the penetration progress. During the 1645 m/s
impact simulation the swirling of the tungsten at first delays the penetration into the
ceramic. Once less projectile material flows into the swirl during the movement of the
projectile, the penetration is increased because more projectile material is now used
to penetrate the target material. The swirl creates an open space and therefore more
material is flowing into this area resulting in less material penetrating the ceramic
block. This space would not be present in this amount when the tungsten material flow
was more radial. The open area created would be filled sooner, forcing the projectile
to penetrate into the ceramic, leading to a more continuous penetration.
For the tungsten not being able to have enough radial flow at the steel plug/ceramic
interface also results in the larger penetration results of the 1410 m/s impact simulation. Since the material cannot move enough in radial direction, it will flow into the
ceramic block, resulting in some penetration.
Figure 7.6 shows the damage contour at several simulation times. Very clearly
can be observed the forming of radial cracks as well as the typical creation of the
damage conoid in front of the projectile. Because of wave reflection at the ceramic/steel

7.2. THICK TARGET IMPACT

41

plug interface at the bottom, a fracture conoid is also formed from this surface upwards
to the projectile. This type of behaviour is typically what is to be expected as ceramic
material response to impact loading conditions [16].

(a) 12 s

(b) 18 s

(c) 24 s

(d) 30 s

Figure 7.5: Material movement during a thick target impact simulation at v=1645 m/s.

(a) 8 s

(b) 12 s

(c) 14 s

(d) 16 s

Figure 7.6: Damage contour of thick target impact simulation at v=2175 m/s.

42

CHAPTER 7. MODEL VALIDATION

Chapter 8

Numerical issues in MSC.Dytran


The validity of some of the results generated during the performed simulations were
questioned because of unexpected material responses. Before performing actual impact
simulations which are to be used as a replacement for experiments, it is necessary
to find the origin of these inconsistencies and, if possible, solve the problems. In the
work of B. Adams [1], the steel parameters used were obtained by fitting numerical
simulations to bullet impact experiments. The steel used there was designated as
BSEC180. The steel used in the experiments of chapter 4, designated as BSEC510,
differed from the BSEC180 steel (i.e. the BSEC510 material had less ballistic protection capabilities). For this reason, an adjustment of the material parameters obtained
by B. Adams was required.
Various bullet impact experiments on BSEC510 plates were conducted by Benteler
Automotive. The results from these experiments were used to determine the BSEC510
material parameters. From the experimental data, two sets of bullet impacts were
used which are given in table 8.1. All bullets were launched at the same steel plate,

Table 8.1: Bullet impact data used for BSEC510 material fitting.
Bullet type
SS109

M80

Impact velocity
(m/s)
550
560
870
880

43

Result
Stop, small bulge
Full penetration (bullet
diameter size hole)
Stop, heavy bulge
Full penetration (bullet
diameter size hole)

44

CHAPTER 8. NUMERICAL ISSUES IN MSC.DYTRAN

(a) SS109

(b) M80

Figure 8.1: Initial configuration for simulations of bullet impacts on a 6.3 mm steel
plate.
which had a thickness of 6.3 mm. A numerical simulation of these experiments was
used to obtain the constitutive parameters of the target material. In figure 8.1 the
simulation configuration is shown.

8.1

Voids

The axial symmetry option, mentioned in chapter 6, reduces calculation time significantly by increasing the time step of the simulation. However when using this option,
computed results differ from those obtained with a simulation performed without this
option set. In figure 8.2 two M80 bullet impacts at 870 m/s on a 6.3 mm steel plate
are shown with and without the axial symmetry option being activated. Without the
axial symmetry option, the bullet is completely stopped by the steel plate, whereas
in case the axial symmetry option is used, the bullet penetrates the plate. When the
axial symmetry option is activated, in every simulation, regardless of the material
parameters, a cracks develops at the back of the steel plate (see figure 8.3). In order to
be able to stop the bullet at this speed, the material should bulge significantly. This
however is not possible since the material cracks when the material starts to bulge.
This cracking is caused by the development of voids at the back of the plate.
A void is an element where the stress and the pressure are set to zero. It contains a certain amount of material mass (or no mass at all) that causes the pressure
to be less than the limited minimum pressure (pmin ). When there is a very small

8.1. VOIDS

45

(a) Without option activated

(b) With option activated

Figure 8.2: Significant difference in results when using the axial symmetry option.

Figure 8.3: Loss in material when using the axial symmetry option.
amount of mass present in an element, the density (which is defined by the mass inside
the element and the element volume) becomes very small with respect to the initial
material density. This causes the volumetric strain, defined in equation (5.10), to
approximately equal 1. Calculating in that case the pressure would give a value close
to K, with K the bulk modulus (assuming a linear pressure model). To avoid an
element to have a pressure of the physical impossible value of K, the limit pressure
is set slightly above K.
To calculate the void fraction in an element the following equation is used

FV =

Vm
Vm

m
0

(8.1)

46

CHAPTER 8. NUMERICAL ISSUES IN MSC.DYTRAN

where F V is the void fraction, Vm , the current volume occupied by the mass present
in the
element, m, the mass in the element and 0 the reference density. The term

m
0

represents the volume that the mass would occupy if it had the reference density.

An element is considered to be void if the void fraction is greater than a defined


tolerance fraction, F Vtol . Note that equation 8.1 is only used in case the pressure is
less the pmin . This means that only if the calculated element pressure is less than pmin
a void fraction, calculated by equation 8.1), that is larger than F Vtol leads to a void
element. If an element already had a void fraction the previous increment, negative
pressures are not allowed. Therefore a void can only generate stress and pressure
again, if enough mass has entered the element to generate a positive pressure. In that
case the void fraction will be less than F Vtol . In that case the void fraction of the
element will be set to zero.

Figure 8.4: Computed and manually calculated values of 4 elements near and at the
vicinity of the crack initiation.
Looking closely at what happens at the position where the crack starts to grow it becomes clear that the void fraction is not calculated correctly. In figure 8.4 the element
output of 4 elements near and at the vicinity of the crack initiation is given together
with manually calculated values, calculated from the current element data. Element
1, 2 and 4 are correctly calculated by MSC.Dytran. Note that manually calculated
values can show a small difference from computed values, because MSC.Dytran
performs calculations with double precision. The manually calculated values are
calculated with single precision. The void fractions of elements 2 and 4 are greater

8.2. BULLET IMPACT FITTING

47

than the tolerance void fraction, which is set to be 6.104 . This however, does not lead
to a void fraction, because the element pressure is greater than the limit pressure, pmin .
Element 3 however, had a void fraction after the previous increment. To be
able to contain nonzero pressures, the element first should have enough mass to
generate a positive pressure (which is not the case). Although a negative pressure
is being generated during the increment, a void fraction of F V = 0 is computed by
MSC.Dytran. The manually calculated value however indicates a void fraction larger
than the tolerance value of F Vtol = 6.104 , which should lead to a void element.
As a consequence, element 3 is capable of having tensile stresses, while element 1 is
not. This leads to material separation at this point which is manifested by a crack
initiation. The steel plate therefore is not able of bulging anymore but starts to
fracture and hence the bullet is not stopped. Performing calculations with the axial
symmetry option activated with a time step close to the time step used by the code
when axial symmetry is not active, leads to results similar to those computed when
the axial symmetry options is not activated. Therefore the larger time step source of
the inaccurate void fraction calculations.

8.2

Bullet impact fitting

Because of the inconsistencies discussed in the previous section the axial symmetry
option will not be used for bullet impact simulations. In table 8.1, an overview is
given of the data which will be used for determining the values of the BSEC510 steel
parameters values for the JohnsonCook model. For each bullet type the ballistic
limit velocity, the velocity at which 50% of the bullets fired will be stopped, was
taken within 10 m/s to guarantee an accurate material fitting. Using the material
parameters obtained by Adams [1] as a first approximation, it was expected that
the parameters should not be altered extensively. During the simulations, however, a
discrepancy occurred (see figure 8.5).
Parameter set 1 represents a material with a higher strength than parameter
set 2. Using the first set an M80 bullet impacting at 870 m/s is stopped as well
as an SS109 bullet impacting at 960 m/s. According to table 8.1, the M80 bullet
should indeed be stopped but the SS109 bullet should penetrate the steel plate. To
generate this result, the material parameters should be changed such that bullet
penetration and bullet stops would occur at the proper velocities, described in 8.1.

48

CHAPTER 8. NUMERICAL ISSUES IN MSC.DYTRAN

(a) SS109 bullet at 960 m/s against a (b) M80 bullet at 870 m/s against a
6.3 mm BSEC510 steel plate with
6.3 mm BSEC510 steel platewith
parameter set 1
parameter set 1

(c) SS109 bullet at 960 m/s against (d) M80 bullet at 870 m/s against a
a 6.3 mm BSEC510 steel platewith
6.3 mm BSEC510 steel platewith
parameter set 2
parameter set 2

Figure 8.5: Discrepancy in BSEC510 material fitting.


The second set has a reduced strength but looking at figure 8.5(c) and 8.5(d) the steel
plate still stops the SS109 bullet but is penetrated by the M80 bullet. This results
in an impossible situation because the material should have a higher strength when
impacted by a M80 bullet and a reduced strength when impacted by a SS109. After
performing calculations with a large number of other parameter sets, the discrepancy
still occurred. Therefore it must be concluded that the material parameters for the
BSEC510 steel can not be uniquely determined.

8.3

Depth of penetration

Depth of penetration tests (DOP), mentioned in section 3.3, are frequently used to
investigate the ballistic performance of ceramic material. The depth of penetration into

8.3. DEPTH OF PENETRATION

(a) Simulation configuration

49

(b) Computational result at v=1770 m/s

Figure 8.6: Depth of penetration experiment setup and computational result.


a steel block with a ceramic tile in front of it, is compared to the depth of penetration
the same projectile has into the bare steel, impacted with the same velocity (see
figure 8.6(a)). In [33], five ceramics were impacted by a tungsten projectile and backed
by a thick steel block. This data could be used to further validate the JHB model
implementation and to determine JHB material parameters of other types of ceramics,
like aluminium oxide.
Before fitting the ceramic material parameters, the impact into the bare steel block
is modelled to determine the accuracy of the JohnsonCook model in MSC.Dytran.
In [17], the experiments have also been simulated. The material parameters used
in these simulations are given in table 8.2. Using the these parameters, the same
simulations were performed in MSC.Dytran. In figure 8.6(b), the computed result is
given of a tungsten projectile with an impact velocity of 1770 m/s into the bare steel
block. The final depth of penetration is 30 mm, but according to the experimental
data, this should be 36 mm [33]. When comparing the results from MSC.Dytran
with results from [17], which are very close to the experimental data, there is a
significant difference. However, the only difference between both simulations is that
the simulations in [17] are performed with a Lagrangian approach. When looking
in detail at how the JohnsonCook model is programmed in MSC.Dytran, a few
inconsistencies can be observed:
The material parameter D3 in the JohnsonCook damage model (see equation

(5.15)) determines the influence of the mean stress to equivalent stress ratio,

50

CHAPTER 8. NUMERICAL ISSUES IN MSC.DYTRAN

Table 8.2: JC material constants for 4340 steel and tungsten[17].


Material
o (kg/m3 )
G (GPa)
K1 (GPa)
K2 (GPa)
K3 (GPa)
Tm ()
cp (J/kg)
A (MPa)
B (MPa)
n
C
m
D1
D2
D3
D4
D5

4343 steel
7850
78
164
294
500
1520
477
910
586
0.26
0.014
1.03
-0.08
2.10
-0.50
0.002
0.61

Tungsten
18360
124
302
470
335
1768
250
770
177
0.12
0.016
1.0
0
0.33
-1.50
0
0

on
the plastic strain to failure, which is implemented in the MSC.Dytran code as

p

as

. In the reference manual of MSC.Dytran [34], however, this is documented

. According to this manual the sign of all parameter values of D3 mentioned

in literature should be changed. It is clear this change of sign is not correct and
generates incorrect results.
In equation (5.13) the strain rate influence on the material strength is determined
by the effective plastic strain rate, p . In MSC.Dytran this effective plastic strain

rate is replaced by the effective total strain rate, . This should be allowed under the assumption that the effective plastic strain rate equals the effective total
strain rate. For high strain rates, the difference between both strain rate definitions is small and the assumption is valid. However, the initial yield stress is
increased, because the effective total strain rate is nonzero in case of elastic
deformation. The strain rate therefore has a continuous influence on the material strength. This is not correct, since the effective plastic strain rate is zero for
elastic deformation and therefore has no influence on the initial yield strength.
As a solution for the strain rate implementation problem, the influence of the strain
rate the material strength was activated only if the yield strength was exceeded.

8.3. DEPTH OF PENETRATION

51

Changing the sign of the material parameter D3 and the use of the adjusted
strain rate influence, however, did not generate depth of penetration results close
enough to the experimental data. The results shown in figure 8.6(b) are computed
with the adjusted material parameter and strain rate influence.
In the work of Adams [1] all D3 material parameters used were changed in sign. The
obtained parameters for the BSEC180 steel are therefore not completely correct.
Material parameters for the BSEC510 steel therefore could differ more from the
BSEC180 material parameters than just a few percentages.
With this knowledge, new simulations of bullet impacts on BSEC510 steel plates
were performed but still did not generate unique material parameters, because the
inconsistency mentioned in section 8.2 still occurred.

52

CHAPTER 8. NUMERICAL ISSUES IN MSC.DYTRAN

Chapter 9

Conclusion and recommendations


9.1

Conclusion

To model ceramic material subjected to high strain rates and high pressures,
various constitutive models have been reviewed. From the models presented, the
JohnsonHolmquistBeissel model was determined to be the most appropriate.
Because it was not available in the FEcode MSC.Dytran it was implemented through
usersubroutines. Various features needed for a correct model implementation were
not available in these subroutines and were therefore adaptations were made by
MSC.Software.
For model validation purposes, plate impact experiments on silicon carbide were
performed and compared to experimental data. The characteristic material loading
and unloading was successfully reproduced.
To further investigate the validity of the model implementation, a tungsten projectile
impacting a thick ceramic block was simulated. The overall results of the depth
of penetration and the damage propagation were satisfying but the simulations
also revealed limitations when using an Eulerian based approach. These limitations
however are no restriction for being able to simulate bullet impacts on ceramic
material in MSC.Dytran since the limiting features are not present in this type of
application.
The successful simulations show a correct model implementation of both the strength
model and the damage model.
Obtaining new material parameters for the BSEC510 steel revealed a discrep53

54

CHAPTER 9. CONCLUSION AND RECOMMENDATIONS

ancy. Simulations of the impact of a SS109 bullet above the ballistic limit velocity,
resulted in bullet stops, whereas simulations with a M80 bullet below the ballistic limit
velocity resulted in full penetration, which should not occur according to experimental
results. It appeared not to be possible to define a unique set of parameters to model
the armour steel material response for both bullet impacts.
The depth of penetration experiments did not generate the required penetration
distance. Although correct material parameters were used, a significant error occurred.
After analysing the implementation of the JohnsonCook model in MSC.Dytran
two inconsistencies were found. The actual implementation differed from what was
documented, which initially resulted in incorrect use of material parameters obtained
from literature. Also the strain rate behaviour was not correctly modelled. Performing
depth of penetration simulations with a modified implementation still did not result
in an accurate penetration distance.
The axial symmetry option which can be used in MSC.Dytran to significantly
reduce calculation time, did not generate the same results as when this option
was not used. Using the option resulted in material cracking leading to a bullet
penetration. Performing the simulation without this option resulted in a complete
bullet stop. Detailed analysis of what initiates the incorrect material cracking revealed
an inconsistency in the void fraction calculation procedure.

9.2

Recommendations

Having a ceramic material model available in MSC.Dytran, it is possible to simulate


bullet impacts on ceramic/steel targets. Before reliable results can be obtained the
following issues need to be taken care of first:
The incorrect calculation of void fraction should be thoroughly investigated. Al-

though only the void fraction algorithm influences the void fraction, another yet
unknown procedure also adjusts or influences the void fraction. A possible cause
could be the void fraction tolerance which could be altered during calculations
and therefore sets an element incorrectly to a zero void fraction. Apparently this
only takes place when the axial symmetry option is used, but at this point it
cannot be excluded that without this option activated this does not happen as
well.

9.2. RECOMMENDATIONS

55

The JohnsonCook model has to generate correct results when proven material

parameters are used.


If these software problems are solved, the following steps towards the development of
ceramicbased armouring solutions need to be taken by PDE Automotive:
Define the correct material parameters for the BSEC510 steel.
Define material parameters for aluminium oxide by fitting plate impact experi-

ment data. As a parameter validation the performed bullet impact experiments


on ceramic/steel targets can be simulated.
Investigate the ballistic performance of confined ceramic tiles, or ceramic tiles

with tape or a metal plate in front of it [35].


Simulate oblique impacts on ceramic material and compare the generated results

to experimental results to extend the validation and reliability of ballistic impact


simulations.
Find an optimum thickness ratio of a ceramic/steel target that is able of stopping

bullets from the FB7 class in order to reduce the weight of the armouring system.

56

CHAPTER 9. CONCLUSION AND RECOMMENDATIONS

Bibliography
[1] B. Adams. Simluation of ballistic impacts on armored civil vehicles. Masters
thesis, Eindhoven University of Technology, Eindhoven, The Netherlands, 2006.
[2] W. D. Callister Jr. Materials Science and Engineering an Introduction. John
Wiley and Sons, Inc., 5th edition, 1999. ISBN 0-471-32013-7.
[3] J.A. Zukas, T. Nicholas, H.F. Swift, L.B. Greszczuk, and D.R. Curran. Impact
Dynamics. John Wiley and Sons, Inc., 1982. ISBN 0-471-08677-0.
[4] E.H. Lutz and H. Hoppert. High-Tech Ceramics in Ballistic Protection. Catalogue,
2001.
[5] Declaration (iv,3) concerning expanding bullets. International Committee of the
Red Cross, 1899. http://www.icrc.org.
[6] Beschussamt Ulm. PM2000, Richtlinie zur Pr
ufung und Zertifizierung Durchschusshemmende plattenartige Materialien. Vereinigung der Pr
ufstellen fur angriffhemmende Materialien und Konstruktionen, 2002.
[7] Metallwerk Elisenh
utte GmbH Nassau. www.men-nassau.de.
[8] D.E. Grady. Dynamic Properties of Ceramic Materials. Sandia National Laboratories, Albuquerque, Feb 1995. Experimental Impact Physics Department 1433.
[9] C.H.M. Simha. High rate loading of a high purity ceramicone dimensional stress
experiments and constitutive modeling. PhD thesis, University of Texas at Austin,
1998.
[10] Morgan Advanced Ceramics. www.morganadvancedceramics.com.
[11] C.H.M. Simha, S.J. Bless, and A. Bedford. A constitutive model for high strain
rate response of a high purity ceramic. U.S. Army Research Laboratory, 2003.
DAAA2193C101.
57

58

BIBLIOGRAPHY

[12] G.R. Johnson, T.J. Holmquist, and S.R. Beissel. Response of aluminum nitride
(including phase change) to large strains, high strain rates and high pressures. J.
Appl. Phys., 94(3):16391646, 2003.
[13] G.R. Johnson and T.J. Holmquist. A computational constitutive model for brittle
materials subjected to large strains, high strain rates, and high pressures. Proceedings of Explomet Conference, pages 10751081, 1990.
[14] C.E. Anderson Jr., G.R. Johnson, and T.J. Holmquist. Ballistic experiments and
computations of confined 99.5 Al2 O3 . 15th Int. Symp. on Ballistics, pages 6572,
1995. Jerusalem, Israel.
[15] T.J. Holmquist and G.R. Johnson. Response of silicon carbide to high velocity
impact. J. Appl. Phys., 91(9):58585866, 2002.
[16] T.J. Holmquist and G.R. Johnson. Characterisation and evaluation of silicon
carbide for high-velocity impact. J. Appl. Phys., 97:112, 2005.
[17] T.J. Holmquist, D.W. Templeton, and K.D. Bishnoi. Constitutive modeling of
aluminum nitride for large strain, highstrain rate and highpressure applications.
Int. J. Impact Eng., 25:211231, 2001.
[18] A.M. Rajendran. Modeling the impact behavior of ad85 ceramic under multiaxial
loading. Int. J. Impact Eng., 15(6):749768, 1994.
[19] F.L. Addessio and J.N. Johnson. A constitutive model for the dynamic response
of brittle materials. J. Appl. Phys., 67:32753286, 1990.
[20] C.F. Cline H.C. Heard and. Mechanical behaviour of polycrystalline beo, Al2 O3
and aln at high pressure. Journal of Materials Science, 15:18891897, 1980.
[21] E. Dekel Z. Rosenberg and, V. Hohler, A.J. Stilp, and K. Weber. Hypervelocity
penetration of tungsten alloy rods into ceramic tiles: experiments and 2d simulations. Int. J. Impact Eng., 20:675683, 1997.
[22] Century Dynamics. Autodyn. www.century-dynamics.com.
[23] J.O. Hallquist. LSDyna Theoretical manual. Livermore Software Technology
Corporation, 2876 Waverly Way, Livermore, California, 94550-1740, 1998.
[24] W.F. Brace. Dilatancy in the fracture of crystalline rocks. Journal of Geophysical
Research, 71(16):39393953, 1966.

BIBLIOGRAPHY

59

[25] D.J. Steinberg, S.G. Cochran, and M.W. Guinan. A constitutive model for metals
applicable at high-strain rate. J. Appl. Phys., 51:14981503, 1980.
[26] G.R. Johnson and W.H. Cook. A constitutive model and data for metals subjected to large strains, high strain rates and high temperatures. Proceedings of the
Seventh International Symposium on Ballistics, pages 541647, 1983. The Hague,
The Netherlands.
[27] G.R. Johnson and W.H. Cook. Fracture characteristics of three metals subjected
to various strain, strain rates, temperatures and pressures. Eng. Fract. Mech,
21:3148, 1985.
[28] MSC.Software Corporation, Santa Ana. MSC.Dytran theory manual, 2005 r3 edition.
[29] D.E. Grady and R.L. Moody. Shock compression profiles in ceramics. Sandia
Report SAND96-0551, 1996. Sandia National Laboratories Albuquerque.
[30] X. Quan, R.A. Clegg, M.S. Cowler, N.K. Birnbaum, and C.J. Hayhurst. Numerical
simulation of long rods impacting silicon carbide targets using JH1 model. Int.
J. Impact Eng., 33:634644, 2006.
[31] P. Lundberg, R.Renstrom, and B. Lundberg. Impact of metallic projectiles on
ceramic targets: transition between interface defeat and penetration. Int. J. Impact
Eng., 24:259275, 2000.
[32] T.J. Holmquist and G.R. Johnson. Modeling prestressed ceramic and its effect on
ballistic performance. Int. J. Impact Eng., 31:113127, 2005.
[33] J.E. Reaugh, A.C. Holt, M.L. Wilkins, B.J. Cunningham, B.L. Hord, and A.S.
Kusubov. Impact studies of five ceramic materials and pyrex. Int. J. Impact Eng.,
23:771782, 1999.
[34] MSC.Software Corporation, Santa Ana. MSC.Dytran Reference manual, 2005 r3
edition.
[35] S. Sarva, S. Nemat-Nasser, J. Mcgee, and J. Isaacs. The effect of thin membrane
restraint on the ballistic performance of armor grade ceramic tiles. Int. J. Impact
Eng., 34(2):277302, 2007.
[36] M.G.D. Geers. Applied Elasticity in Engineering. Technische Universiteit Eindhoven, lecture notescourse 4a450 edition, 2002.

60

BIBLIOGRAPHY

Acknowledgement
First of all, I would like to thank my supervisor at PDE Automotive, Freddie Huizinga,
for giving me the opportunity to take part in this interesting project. I also thank
him for his help and guidance during this project. Especially, I would like to thank
my academic coach, Hans van Dommelen, for his critical reviews and remarks. They
significantly helped me to stay focused on the project objective and the way to proceed.
I thank MSC Software for their help during this project, since the material model
implementation was not always straightforward. Finally, I would like to thank my
colleagues at PDE Automotive for the pleasant and stimulating working atmosphere
and for their interest and useful feedback.

61

62

BIBLIOGRAPHY

Appendix A

Experimental results
CONFIDENTIAL

63

64

APPENDIX A. EXPERIMENTAL RESULTS

Appendix B

Additional pressure
In figure 5.2(c) is shown that a decrease in material strength occurs when the material
changes from an intact (D < 1) state to a failed state (D = 1). This represents a
decrease of the internal elastic energy corresponding to the deviatoric stress. The total
energy present,Et , for a linearly elastic material is: [36]

Et =
=
=

1
:
2
1
: 4S :
2

1 h 4
: S : h + h : 4S : d + d : 4S : h + d : 4S : d ,
2

(B.1)

where is the stress tensor, the strain tensor, 4 S the compliance tensor, h the
hydrostatic stress tensor and d the deviatoric stress tensor.
This can be split into a hydrostatic and a deviatoric part, U . The deviatoric part is
given by
U

= d : 4S : d
1
=
J2 ( d )
2G
1
( d : d )
=
4G

2
=
,
6G

(B.2)

where J2 ( d ) is the second invariant of the deviatoric stress,


is the equivalent Von
Mises stress and G is the shear modulus. The decrease in internal elastic energy now
65

66

APPENDIX B. ADDITIONAL PRESSURE

is expressed as
U = Ui Uf ,

(B.3)

where Ui is the internal deviatoric elastic energy of the material before failure (D < 1)
and Uf is the internal deviatoric elastic energy after failure (D = 1), both calculated
at the volumetric strain at time of failure, f . The loss of internal energy is partly or
completely converted into potential hydrostatic pressure, which is determined by the
parameter .
The potential energy of the material at a certain volumetric strain is complete
the marked area in figure B.1. The potential energy increase, Up , when the pressure
is increased by P , is approximately
Up = P f + (P 2 /2K1 ),

(B.4)

where f is the volumetric strain at failure. Here, the first expression is the area of the
marked parallelogram in figure B.1. The second expression if the area of the marked
triangle in figure B.1. Converting the internal elastic energy partly or completely to
potential hydrostatic energy leads to
U = Up .

(B.5)

Substituting herein equation B.4 and solving for P gives


q

P = K1 f +

(K1 f )2 + 2K1 U .

Pressure

P+ P

P
0

Volumetric strain

Figure B.1: Potential energy increase at failure.

(B.6)

Anda mungkin juga menyukai