Professional Documents
Culture Documents
2019 Chambe Jean-Emmanuel D
2019 Chambe Jean-Emmanuel D
&OWVFEFMPCUFOUJPOEV
%0$503"5%&-6/*7&34*5²%&506-064&
%ÏMJWSÏQBS
Institut Supérieur de l’Aéronautique et de l’Espace
1SÏTFOUÏFFUTPVUFOVFQBS
Jean-Emmanuel CHAMBE
le mercredi 10 juillet 2019
5JUSF
6OJUÏEFSFDIFSDIF
Institut Clément Ader
%JSFDUFVS T
EFʾÒTF
M. Christophe BOUVET (directeur de thèse)
M. Jean-François FERRERO (co-directeur de thèse)
Jury :
With the perspective of the design and development of an aircraft seat and in order to respect
the safety regulations in effect, the structure of the developed seat must allow for a swift
dissipation of the energy received in the event of an aircraft crash (Fig. 1) so as to protect the
passengers. The majority of systems integrated into the seats structure and allowing energy
dissipation (Fig. 2) consists of metal components that sustain plastic deformation to dissipate
the energy induced by the crash. Currently, industry and research sectors are turning their
focus towards composite materials to substitute such systems. However, the behavior of these
materials during severe mechanical stress is strongly different from metallic materials, in
particular due to the fact that damage mechanisms are very distinct.
The purpose of this study on composite tubular structures is to evaluate their ability to
dissipate the energy. To this end, different laminate structures were tested in compression
(Fig. 3 and 4) in order to identify their behavior, compare their properties and calculate their
SEA value (Specific Energy Absorption, in kJ.kg-1) used to evaluate their capacity to dissipate
the energy generated during a crash. Those are resulting from the load-displacement curves
obtained during the crushing tests (Fig. 5). The various compression tests were instrumented
and monitored by means of rapid imaging cameras and post-crushing tomographic imaging
has been realized in order to understand the damage mechanisms involved (Fig. 4 and 6).
Testing has been carried out under quasi-static and dynamic loading and using several
boundary conditions. The different results of compression and crushing behavior are also used
in order to build and improve a finite element calculation model (Fig. 7 and 8) allowing to
simulate the response of composite structures of different natures subjected to crash by
integrating the geometry and the composition of the structure (Fig. 8).
The objective of this research work is thus to evaluate the energy that can be dissipated by
composite tubular structures, to compare the absorption values induced by composite
structures of different compositions, and/or bi-materials, and, finally, to provide a finite
element model representing the behavior of composite structure submitted to compression
until damage and fracture of the structure.
It has consequently been established that in static loading, a unidirectional laminate oriented
at 0° and stabilized by woven plies strongly meets the expectations in terms of energy
dissipation, but that is not the case in dynamic loading. In this case, a 90° stratification seems
more adequate. Incidentally, an inner constrained containment is more effective in most cases,
reducing the initial peak load without drastically reducing the SEA value.
Le but de cette étude portant sur des structures tubulaires composites est d’évaluer leur
capacité à dissiper l’énergie. A cette fin, différentes stratifications ont été testées en
compression (Fig. 3 et 4) dans le but de déterminer leur comportement, comparer leurs
propriétés et calculer leurs valeurs de SEA (absorption d'énergie spécifique, en kJ.kg-1)
servant à évaluer leur aptitude à dissiper l’énergie engendrée en cas de crash. Ces dernières
sont issues des courbes effort-déplacement obtenues lors des essais d’écrasement (Fig. 5). Les
différents essais de compression ont été instrumentés et suivis au moyen de caméras rapides et
des images post-essais ont été réalisées par tomographie pour comprendre les mécanismes
d’endommagement mis en jeu (Fig. 4 et 6). Ces essais ont été réalisés à vitesse de chargement
quasi-statique puis dynamique et selon diverses conditions limites. Les différents résultats de
comportement en compression sont également utilisés dans le but de construire et enrichir un
modèle de calcul par éléments finis (Fig. 7 et 8) permettant de simuler la réponse de structures
composites de différentes natures soumises au crash en intégrant la géométrie et la
composition de la structure (Fig. 8).
L’objectif de ce travail de recherche est ainsi d’évaluer l’énergie pouvant être dissipée par des
structures tubulaires composites, de comparer les absorptions induites par des structures
composites de compositions différentes, et/ou bi-matériaux, et enfin de fournir un modèle
éléments finis représentant le comportement de structures composites en compression jusqu’à
l’endommagement et la ruine de la structure.
En particulier, remercier l’Institut Clément Ader, site de Toulouse et toute son équipe de techniciens,
ingénieurs, chercheurs et enseignants-chercheurs pour leur présence au quotidien et l’ISAE-SUPAERO
qui m’a permis de m’épanouir dans un cadre agréable et adapté à une poursuite d’études, ainsi que
de fournir les infrastructures et la logistique pour la bonne réalisation de ces travaux de recherche.
Le CRITT Mécanique & Composites de Toulouse, notamment son directeur technique et l’équipe
d’Ingénieurs Etudes et Développement pour leur assistance lors de réalisations techniques.
La société ARTEC Aerospace pour s’être tournée vers l’Institut Clément Ader de Toulouse et avoir
procuré le besoin ayant servi de base à cette étude.
Et enfin remercier l’Université Fédérale Toulouse Midi-Pyrénées, ainsi que la Région Midi-Pyrénées,
devenue Occitanie – Pyrénées-Méditerranée dans l’intervalle de cette thèse, et le FEDER, Fonds
Européen de Développement Régional, pour avoir contribué au financement de ce projet.
Jean-Emmanuel Chambe
pour les apports nutritifs et vertueux tout comme le bien moral qu’ils ont procuré au cours
de ces années.
L’objet de la présente étude comporte deux volets : le premier porte sur les thématiques de
dimensionnement de structure et de sécurité, réalisé à l’Institut Clément Ader (ICA) de Toulouse,
avec le support de l’ISAE-SUPAERO, et le second porte sur les aspects ergonomie et confort, réalisé
au laboratoire de Physiologie de la Posture et du Mouvement (POM) d’Albi, sous la tutelle de
l’Institut National Universitaire Jean-François Champollion.
L'objectif global de ce projet est de développer des connaissances sur l'ergonomie et sur la sécurité
du siège passager d'avion afin de proposer une structure remplissant pleinement ces conditions
d’utilisation. Seul le volet lié à la conception mécanique et la sécurité du passager via la dissipation
d’énergie par la structure du siège aéronautique est traité et présenté dans ce travail de thèse.
La structure du siège développé doit permettre une dissipation rapide de l’énergie perçue en cas de
crash aérien, ou d’atterrissage dur, dans le but de protéger les passagers. Il s'agit donc dans un
premier temps de se focaliser sur la réglementation et d’en maîtriser les spécificités. Dans un second
temps, il s’agit de comprendre l’effet d’une accélération subie en direction opposée au déplacement
pour un objet en mouvement dans le but de le freiner et de simuler une situation de crash. Enfin il
sera nécessaire de parvenir à modéliser les divers phénomènes entrant en jeu dans le cas d’un crash
pour vérifier le respect des critères de sécurité au regard de la géométrie et des matériaux choisis.
L’objet scientifique de cette étude porte sur l’évaluation et la quantification de structures composites
à dissiper l’énergie engendrée en cas de crash. A cette fin, différentes structures tubulaires en
matériaux composites de compositions différentes sont testées en compression dans le but de
déterminer leur comportement, comparer leurs propriétés et évaluer leur capacité à dissiper de
l’énergie en se basant sur leur valeur de SEA (Specific Energy Absorption, en anglais, ou absorption
d'énergie spécifique).
L’objectif du travail de recherche présenté est ainsi d’évaluer l’énergie pouvant être dissipée par des
structures tubulaires composites, de comparer les absorptions effectuées par des structures
composites de compositions différentes et enfin de fournir un modèle éléments finis représentant le
comportement de structures composites en compression jusqu’à l’endommagement et la ruine de la
structure lors d’un crash.
Lors d’un crash aérien, les éléments constituant la structure de l’avion doivent garantir au maximum
l’intégrité physique des passagers en minimisant les sollicitations mécaniques et physiologiques
perçues par les passagers. Ceci est d’autant plus valable pour les sièges d’avion, s’agissant du dernier
élément structurel pouvant protéger les passagers.
Actuellement, la majorité des sièges d’avion étant principalement réalisés en alliage d’aluminium, le
phénomène de déformation plastique du métal va atténuer les sollicitations perçues par les
passagers et notamment dissiper une partie importante de l’énergie cinétique associée à l’ensemble
siège-passager. Dans le cas d’un siège dont la structure est réalisée en matériaux composites,
l’endommagement seul des éléments composites devra dissiper l’énergie due au crash et atténuer
ainsi les efforts perçus et les accélérations vues par le passager.
Dans les faits, les autorités de certification aéronautique imposent lors d’une sollicitation de type
crash un seuil d’accélération maximal perçu par le cerveau du passager, ainsi que des efforts
maximum en compression dans les lombaires et les fémurs des passagers. Afin de simuler le crash
dans son cas le plus critique, le niveau d’accélération perçu par le siège au niveau du plancher doit
atteindre au minimum 16g (i.e. 16 fois l’accélération de la pesanteur terrestre) en moins de 0,09 s
(selon les spécifications de la norme CS25 - Certification Specifications for Large Aeroplanes,
European Aviation Safety Agency).
La présente étude porte sur le développement de pièces composites devant dissiper de l’énergie lors
de leur endommagement. Dans cette optique, la réalisation d’une étude numérique en dynamique
sur une structure de siège d’avion est menée afin de dimensionner les pièces composites
susceptibles de s’endommager et quantifier l’énergie ainsi dissipée.
Pour ce faire, une modélisation numérique de la structure de sièges aéronautiques ainsi que de ses
passagers est réalisée dans le but de déterminer les efforts transmis à travers la structure nécessaires
à son maintien et garantissant son intégrité, ainsi que la loi d’endommagement et l’absorption
d’énergie nécessaire pour garantir la sécurité des passagers en cas de crash selon les spécifications
de la norme.
La structure de siège envisagée ainsi que les parties permettant la dissipation d’énergie ont pour but
d’être réalisées en tubes composites s’endommageant par écrasement (ou crushing). Ces structures
tubulaires sont également testées expérimentalement pour valider le modèle numérique et le
compléter, notamment afin de s’affranchir des cas de flambage. Plusieurs études ont porté sur le cas
d’endommagement par crushing de structures composites et le lien qu’il présente avec l’énergie
absorbée.
Dans une première partie, constituée de deux volets (Etat de l’art et Pré-dimensionnement d’un
système absorbeur d’énergie) sont exposés les aboutissements de la veille technologique et
réglementaire menée sur les aspects normatifs en matière de sécurité, et plus particulièrement dans
le domaine aéronautique, ainsi que sur les principaux systèmes de dissipation d’énergie, passés et
actuels, d’une part et les moyens mis en œuvre pour mener au dimensionnement d’une structure
absorbeur d’énergie d’autre part. La constitution et les résultats d’un modèle numérique simplifié,
basé sur les spécificités de la norme aéronautique relative à la sécurité des passagers se rapportant
aux aéroplanes de grande envergure et développé dans le but de reproduire une situation de crash
aérien afin de quantifier la dissipation d’énergie nécessaire à la protection des passagers, sont
présentés. Ces premiers résultats servent au dimensionnent et à la confection de structures
tubulaires composites, dont la capacité à dissiper de l'énergie est testée par la suite. Cette partie
relate le contexte global du projet et fixe le cadre de l’étude scientifique.
Une dernière partie regroupe les principales conclusions de cette étude et propose plusieurs
perspectives expérimentales et numériques afin de la parachever.
Un premier constat est l’existence d’un oligopole des fabricants de sièges d’avions (Figure [1]).
Figure [1] : Oligopole des fabricants de sièges aéronautiques, avec en haut les principaux et en
dessous les fabricants secondaires.
1. Aspect et matériaux
A ce jour, seul Expliseat, entreprise française se distinguant pour ses sièges d'avions ultra-légers, a
réussi à passer et obtenir les certifications nécessaires pour proposer un siège en matériaux
composites, sous le nom commercial TiSeat. Le siège composite L3 issu de la collaboration entre
Zodiac Aerospace et Hexcel, respectivement leader français en équipements aéronautiques type
sièges d'avions et fabriquant de matériaux composites en fibre de carbone pour l'aérospatiale, n’a
pas été certifié pour l’instant.
On peut relever deux formes de piètement de sièges majoritairement présents sur les avions
commerciaux actuellement :
Figure [2] et [3] : Photos de sièges avec un dispositif type « vérin » (à gauche) (aviationexplorer.com)
et un dispositif type « boomerang » dépliable (à droite) (aviointeriors.it)
En dessous, schémas représentatifs simplifiés du « vérin » (gauche) et du « boomerang » (droite).
Ces dispositifs assurent le maintien du siège ou de la banquette de sièges d’une part, mais agissent
aussi en tant que systèmes de dissipation de l'énergie en cas de crash.
Les contraintes liées à la sécurité des passagers (et personnels navigants) sont données dans la
norme européenne EASA CS25 - Certification Specifications for Large Aeroplanes, European Aviation
Safety Agency - et détaillées dans la partie EMERGENCY LANDING CONDITIONS aux paragraphes CS
25.561 General et CS 25.562 Emergency landing dynamic conditions [EASA 2009].
Les spécifications données par cette norme en matière de crash (conditions d’atterrissage d’urgence)
sont d’ordre mécanique (structure) et biomécanique (passager). La partie concernant les critères
biomécaniques est principalement définie au moyen de valeurs numériques et précise des seuils
limites de tolérances (cf. Tableau [1]).
La partie concernant la structure du siège laisse une marge plus importante à interprétation. Ainsi la
structure du siège peut se déformer durant le crash et revenir ou non à sa forme et dans sa position
initiale. Sa déformation à l’état final est permise si celle-ci est contenue de manière à ne pas gêner ou
entraver l’évacuation rapide des passagers (cf. Tableau [2]). La seule contrainte majeure imposée se
situe au niveau des points d’attaches aux rails du plancher avion, qui doivent rester ancrés en appuis
permanent et en maintien.
pelvis and lumbar 680 kg (1500 lb) 6 670 N Maximum compressive load
head 1000 units Maximum HIC (Head Injury Criterion)
femur (each) 1021 kg (2250 lb) 10 016 N Maximum axially compressive load
Tableau [1] : Synthèse des spécifications de la norme CS25 concernant les critères sièges-passagers.
Le critère de HIC (Head Injury Criterion) est défini comme suit dans le paragraphe CS 25.562 (c) (5) de
la norme EASA CS25 :
t2 2.5
1
HIC = {(t 2 − t1 ) [ ∫ a(t)dt] } Équation (1)
(t 2 − t1 ) t1
𝑚𝑎𝑥
où :
t1 est le temps initial d’intégration, en s
t2 est le temps final d’intégration, en s
a(t) est la résultante de l’accélération subie par la tête, en g.
Dans la majorité des cas, l’intervalle de temps t2-t1 est choisi égal à 36 ms et le HIC noté alors HIC36.
De plus amples détails relatifs au calcul du HIC et au comportement crânien en cas de choc sont
donnés dans le travail de thèse [MARJOUX 2006] et dans diverses études telles que [HUTCHINSON 1998]
[VAN BEEK 1998] [MARJOUX 2008].
Fx 20 300
Fy 12 000
Fz 26 300
Tableau [3] : Chargement permis pour un double ou triple banc de sièges (spécification Airbus)
[OLSCHINKA 2006]
L’essentiel du travail de recherche porte sur la partie arrière du siège, absorbant l’énergie (en rouge,
représentée par un vérin sur la figure [4]). Cette partie doit être susceptible de se déformer (en
s’allongeant par exemple) de manière à dissiper de l’énergie lors du crash et garantir le respect des
critères biomécaniques imposés [LAANANEN 1982].
Les systèmes existant jusqu’alors étaient principalement réalisés à base de matériaux métalliques et
dissipaient ainsi l’énergie par déformation plastique. Des exemples de tels vérins sont présentés par
[CANNON 1985] et [DESJARDINS 2006] (bien que ce dernier expose des systèmes préférablement adaptés
aux sièges d’hélicoptères) et se trouvent dans diverses études ou communications à ce sujet.
Une illustration d’un tel système est donnée par la figure [5], montrant un « vérin » dans deux
positions, déplié et sain (à gauche) et compacté et endommagé par plastification (à droite), et la
courbe effort-déplacement résultant de ce changement d’état. On peut noter que dans l’exemple
donné, elle correspond à une courbe de "plasticité parfaite". L’énergie dissipée par la déformation de
la structure correspond à l’aire sous la courbe présentée. Plus le niveau du plateau d’effort est élevé,
plus il y a d’énergie dissipée, à déplacement égal.
La figure [6] présente le comportement de ces deux types de systèmes dissipateurs d’énergie. On
note une courbe effort-déplacement semblable aux absorbeurs métalliques dans la forme globale,
mais beaucoup plus chaotique et saccadée. En effet, pour ces matériaux, les mécanismes
d’endommagement sont très différents de la plasticité observée dans les structures métalliques (cf.
Chapitre I, Energy Absorption Capacity of Composite Thin-Wall Circular Tubes under Axial Crushing
with Different Trigger Initiations).
En cas de crash, le comportement de ce type de pièce dans une structure influe sur le comportement
et la tenue globale de la structure, et dans le cas d’un siège de passager, sur l’occupant du siège lui-
même.
Des études expérimentales et numériques ont été menées avec des sièges d’avion complets et des
mannequins réalistes entiers [OLSCHINKA 2006] [SENATORE 2016] [CHEN 2018] pour en percevoir l’effet.
Pour des modèles complets de sièges et passagers, le temps de calcul n’est pas négligeable et peut
s’avérer important [SENATORE 2016]. Certaines études [LANKARANI 2002] [LANKARANI 2007] [SENATORE 2016]
[CHEN 2018] comparent les résultats de simulations numériques avec des résultats d’essais
expérimentaux avec des sièges et des mannequins réels et se basent sur ces tests pour construire et
alimenter leur modèle. La figure [7] présente les résultats de tests expérimentaux et de simulations
numériques de l’une de ces études faite au moyen d’un mannequin (dummy, en anglais).
La présence d’une rangée de sièges supplémentaire devant celle testée semble également avoir une
influence, de même que l’écart entre ces deux rangées (pitch, en anglais) qui est primordial, comme
le montre certaines études, notamment [CHEN 2018] qui teste deux cas, l’un avec un pitch (éclatement
entre les rangées) standard de 29 inch (74 cm) et l’autre de 37 inch (94 cm). Les résultats montrent
une augmentation du HIC de 33% pour un pitch réduit. Ceci n’est cependant observable que dans le
cas de la simulation et non pour les essais, d’où l’influence forte du paramétrage de la rangée avant
si elle est présente. L’impact de la tête du passager sur un obstacle frontal avant est aussi considéré
dans certaines études [LANKARANI 2002] [DeWEESE 2004] [LANKARANI 2007]. Le système de fixation au sol
ou plancher avion est également examiné et pris en compte dans des simulations [OLSCHINKA 2006].
Dans certains cas, des simulations de certifications ont également été réalisées. En effet, dû aux
coûts dispendieux, au temps nécessaire et aux moyens d’essais requis pour tester des prototypes de
sièges d’avions en conditions réelles, de nombreux tests sont réalisés par simulations. Ainsi les sièges
sont d’abord certifiés numériquement, avant de passer une certification d’essai réelle et onéreuse. Il
existe de ce fait des mannequins numériques certifiés dans divers codes éléments finis pour réaliser
de tels essais valables.
Conclusion partielle
La partie suivante est consacrée à la présentation d’un modèle éléments finis basé sur les spécificités
normatives afin de simuler numériquement un crash et ainsi pré-dimensionner la structure du siège.
Cette partie détaille les différents composants utilisés pour parvenir à cette modélisation.
1. Le siège
Une structure de siège passager (piètement et assise) a été modélisée (Figure [8]) au moyen
d’éléments finis de type poutre, de rigidité élevée pour la partie avant et l’assise (en bleu) mais
contraints en plasticité avec une loi variable pour la partie arrière (en rouge), de manière à forcer la
déformation du siège selon certaines directions, principalement en se déformant vers l’avant. Des
modélisations similaires, basées sur un principe de structure simplifiée par des barres ou des
poutres, ont été réalisées lors de précédentes études [WARRICK 1979] [LAANANEN 1990] [LAANANEN 1993].
Le cas d’un banc de 3 sièges (3 assises) supporté par 2 piètements est considéré comme le cas le plus
critique vu le ratio passager par pied de siège. Ce cas a donc été celui retenu pour l’étude. La
structure a été définie de manière à avoir une masse totale de 36 kg, soit une masse moyenne de
siège de 12 kg. Cette valeur est basée sur la masse moyenne de sièges issue de la veille
technologique menée sur ce sujet, malgré le fait que peu de données précises soient disponibles à ce
sujet.
Figure [8] : Modèle 3D d’une structure de piètement de siège simplifiée, en éléments poutres.
La partie arrière (présentée en rouge sur le schéma en figure [8]) est prévue pour se déformer en
allongement « par plastification » et ainsi dissiper de l’énergie. Des éléments poutres avec une loi de
déformation linéaire élastique ont également été rajoutés de manière à modéliser sommairement les
ceintures, retenant les passagers en cas de choc. Certains travaux présents dans la littérature ont
déjà également inclus et tenu compte des ceintures dans leurs modélisations [LAANANEN 1993]
[OLSCHINKA 2006] et certains montrent leur importance capitale [SHANAHAN 2004].
Un tel mannequin a été modélisé au moyen d’éléments poutres via le logiciel de calcul Abaqus FEA,
en se basant sur un modèle tridimensionnel à 12 segments et 29 degrés de liberté développé et
présenté par [LAANANEN 1983], avec une attention particulière portée aux dimensions et masses des
différents membres le constituant (Figure [9]).
Figure [9] : Modèle 3D d’un mannequin biomécanique (ATD) d’après [LAANANEN 1983],
où sont identifiées en rouge les zones d’intérêts au regard des critères de la norme CS25.
Ce mannequin positionné sur une structure de siège simplifiée modélisée en éléments poutres à
propriétés matériaux variables a permis de mesurer les différentes valeurs liées aux critères
biomécaniques selon différents cas de chargements (en vitesse et accélération) et selon différentes
propriétés matériaux testées pour les éléments structuraux du siège.
Différents agencements et différentes géométries de sièges ont également été testés. Une loi
linéaire élastique de comportement de ceinture a également été ajoutée au modèle afin de le
compléter et le rendre plus réaliste. Cette loi est présentée au moyen des figures [11] et [12] ci-
dessous.
Le modèle de mannequin construit a été combiné et placé sur la banquette de 3 sièges modélisée
(Figure [10]). L’ensemble a été soumis à un évènement de type crash aérien. Les données
mécaniques des éléments du modèle sont précisées au moyen de la figure [11]. On notera sur cette
figure que les sections des poutres ont été forfaitairement prises égales à 1 mm2, ce qui revient à
identifier les modules d’Young en MPa aux raideurs en N/mm.
Figure [10] : Assemblage des 2 constituants du modèle. Les points d’intérêts biomécaniques (d’après
la norme CS25) sont signalés en rouge sur le modèle de mannequin. Les éléments d’intérêt majeur
sur le modèle de siège (« vérins » arrière) sont également signalés en rouge.
Figure [11] : Données mécaniques pour le comportement des constituants de la figure [10].
Les données de raideur et de plasticité de la partie « vérin » arrière ont été progressivement
incrémentées pour arriver à son dimensionnement optimal. La loi régissant le comportement du
« vérin » arrière (rear absorber law) est une loi en plasticité parfaite, donnée en figure [12].
Figure [12] : Modélisation d’une situation de crash pour un ensemble {sièges + passagers} selon les
spécificités de la norme CS25.
Les valeurs de vitesse initiale (en rouge et dirigée vers la droite) et d’accélération (en orange et
dirigée vers la gauche) sont issues de la norme CS25, la loi de comportement « ceinture », linéaire
élastique, (en bleu clair, Figure [12]) est approximée et tirée de spécifications constructeurs et
empiriques. Peu de données sont disponibles à ce sujet. Une ceinture de siège avion a par ailleurs été
testée en traction-compression afin de vérifier l’ordre de grandeur de la raideur utilisée dans cette
loi.
Le système total {sièges + passagers} est soumis à une vitesse initiale Vo et une accélération a(t) est
imposée en sens inverse (décélération) pour simuler l’évènement de crash. Les valeurs d’accélération
au niveau de la tête des passagers, de vitesse et déplacement, ainsi que des efforts dans les fémurs
et les lombaires sont relevées, constituant la base des critères biomécaniques (en rouge, Figure [13]).
D’un point de vue structure et mécanique, le déplacement de l’avant du siège, ainsi que les valeurs
de déplacement, contrainte et déformation de la partie arrière sont relevés (en vert, Figure [13]).
4. Le logiciel de calculs EF
Deux logiciels de calcul par éléments finis ont été utilisés pour la réalisation de cette étude :
- Abaqus de Dassault Systèmes, utilisé au sein de l’ISAE-SUPAERO et de l’ICA,
- Dytran de MSC Software Corporation, utilisé au sein d’ARTEC AEROSPACE.
Les calculs de comportement du siège et du passager ont été réalisés principalement via le logiciel
Abaqus de manière à pouvoir itérer rapidement. Une comparaison a été effectuée subséquemment
entre les deux logiciels de calcul afin de vérifier le comportement du mannequin modélisé.
La comparaison des modèles Abaqus ICA et Dytran ARTEC est présentée ci-après dans une seconde
partie, notamment au moyen des figures [16] et [17] et des courbes [18] à [21].
Il a tout d’abord été vérifié que les conditions limites et sollicitations imposées au système ont bien
été respectées, notamment aux points de fixation « plancher », et que les profils de vitesse et
accélération étaient concordants avec les données de la norme (Figure [14]).
Une comparaison a ensuite été effectuée entre les résultats obtenus via le logiciel de calcul Abaqus
utilisé à l’ICA et le logiciel Dytran utilisé par ARTEC AEROSPACE. L’intérêt de l’utilisation du logiciel
Dytran au sein de la société ARTEC réside essentiellement dans l’accès à un mannequin ATB
(Articulated Total Body) complet et certifié pour les calculs de crash.
Le graphique ci-dessous (Figure [15]) reprend le profil d’accélération imposée en indiquant différents
marqueurs servant de points de références temporelles pour effectuer la comparaison entre les deux
modèles.
Figure [16] : Captures d’écran à divers instants de l’animation résultant de la modélisation du siège
simplifié avec le mannequin ATD modélisé via Abaqus positionné dessus.
Figure [17] : Captures d’écran à divers instants de l’animation résultant de la modélisation du siège
simplifié avec le mannequin ATB Dytran positionné dessus.
Les similitudes rencontrées dans les résultats ont permis de conclure à une assez bonne corrélation
entre les deux modèles, malgré des divergences sur des comportements extrêmes, notamment au
niveau des avant-bras et des jambes.
Pour compléter et confirmer ces observations, une comparaison plus précise et qualitative des
paramètres biomécaniques relevés aux principaux points d’intérêts du mannequin a également été
effectuée.
Les courbes présentées à la figure [18] ont pour objectif de permettre la comparaison entre les
efforts perçus par les lombaires dans le cas des deux modèles. Malgré la présence d’oscillations plus
marquées dans le cas de la simulation Dytran, les allures des deux courbes sont similaires et l’ordre
de grandeur reste le même.
Figure [19] : Comparaison des trajectoires de la tête du passager, issues des simulations numériques,
entre les deux modèles.
Figure [20] : Comparaison des courbes d’accélération perçue par la tête du passager, issues des
simulations numériques, entre les deux modèles.
Des valeurs similaires ont été obtenues pour l’accélération de la tête [LANKARANI 2002] [LANKARANI 2007]
[CHEN 2018] et son déplacement [DeWEESE 2004] [OLSCHINKA 2006] dans de précédentes études,
confirmant ainsi la validité des résultats issues des simulations dans la présente étude.
Bien que les différences d’accélération se retrouvent dans les valeurs de HIC (critère dépendant
directement de l’accélération de la tête du passager), on observe la même tendance entre les deux
courbes sur le graphe présentant l’évolution du HIC au cours du temps (Figure [21]), malgré un retard
pour le modèle construit à l’ICA.
Les valeurs de HIC obtenues par simulations (Figure [21]) sont du même ordre de grandeur que celles
de la littérature issues de tests (972-720) ou de simulations (842-837, ou 1559-1126 selon la
configuration) [CHEN 2018]. Ceci permet de confirmer encore une fois la concordance avec le modèle
construit.
La plus grande valeur de HIC obtenue pour la simulation avec le modèle Dytran (Figure [21])
s’explique par la valeur plus importante (pics plus importants, plus larges et plus hauts) de la courbe
de l’accélération de la tête du passager (Figure [20]) pour ce modèle, en particulier autour de 0.14 s.
Le HIC étant en première approximation une intégrale mobile de l’accélération de la tête du passager
(à l’exposant 2.5 près), sa valeur en est donc nécessairement augmentée dans ce cas.
Bien que présentant des différences, notamment en termes d’oscillations numériques, l’ensemble
des courbes présentées montre une grande similarité dans le comportement des mannequins entre
le modèle développé sur Abaqus avec un mannequin simplifié et celui développé sous Dytran avec
un mannequin certifié issu d’une acquisition commerciale.
Les différences observées entre les deux modèles sur les deux codes de calculs différents restant
limitées, le modèle développé pour le logiciel Abaqus a été jugé suffisant pour réaliser l’étude et
accomplir les itérations de calculs.
La comparaison ainsi effectuée valide la construction du modèle de mannequin utilisé pour cette
étude, le but étant d’avoir un outil rapide de calcul de pré-dimensionnement, estimé suffisant et
satisfaisant pour l’application souhaitée.
Plusieurs itérations de calcul ont été faites en utilisant ce modèle. A chaque fois, le seuil de plasticité
de la poutre en partie arrière du siège (en rouge Figure [8]) simulant le « vérin » a progressivement
été modifié de manière à quantifier son influence sur le système total et ainsi observer l’évolution
des différents paramètres retenus (déformation siège, HIC, effort fémur et effort lombaires).
Les résultats donnés dans cette section sont tous présentés en fonction de la variation du seuil de
plasticité du « vérin » arrière (en rouge figures [8] et [12], avec la loi plastique donnée figure [12]).
On notera que toutes les données « passager » présentées dans cette partie sont issues d’un seul
mannequin, correspondant au passager central sur la banquette de 3 sièges, configuration la plus
critique, selon le ratio nombre de passagers par nombre de piétements de sièges (Figure [22]).
Figure [22] : Répartition du nombre de passagers par piétements de sièges et cas le plus critique.
Figure [23] : Déplacement horizontal (dx) de l’avant du siège en fonction de l’effort plastique du
« vérin » arrière.
Figure [24] : Valeur de HIC36 calculée pour un passager en fonction de l’effort plastique du « vérin »
arrière.
La figure [25] donnée ci-dessous combine les deux précédents graphes et positionne les seuils à
respecter pour les différents critères discriminants. Elle permet ainsi de mettre en regard l’évolution
du déplacement horizontal du siège face à l’évolution du HIC passager en fonction de l’effort
plastique du « vérin ».
Bien qu’une zone ultime de recouvrement correspondant au respect de ces deux critères n’ait pas
été atteinte lors des diverses itérations, les calculs ont tout de même donné des résultats conduisant
à une convergence autour de 20 kN pour l’effort plastique du « vérin » arrière.
Concernant les efforts transmis dans le fémur et les lombaires, les résultats (basés sur la même
démarche que pour le HIC) sont présentés en graphe [26] et [27]. Pour rappel, les efforts perçus par
ces deux composants du corps humain ne doivent pas dépasser 10000 N et 6670 N respectivement.
On observe sur le graphe [26] ci-dessous que pour toutes les valeurs d’effort plastique testées, la
condition restrictive sur l’effort maximal transmis au fémur lors du crash est toujours respectée (avec
la modélisation réalisée). Ceci est donc vrai pour la valeur du déplacement permanent fixée (0.08 m).
En ce qui concerne les lombaires, on observe également au moyen du graphe [27] ci-dessous que
pour toutes les valeurs d’effort plastique testées, la condition restrictive sur l’effort maximal transmis
lors du crash est toujours respectée. Ceci est donc encore une fois vrai pour la valeur du
déplacement permanent fixée (0.08 m).
Les critères sur le chargement perçu par le fémur et les lombaires étant toujours respectés, celui
imposé sur l’accélération de la tête (via le critère de HIC) est le plus drastique et le plus compliqué à
respecter. De précédentes études sont déjà arrivées à la même conclusion [LAANANEN 1990].
En se basant sur ce constat et les précédents résultats, il a donc été décidé de se référer uniquement
à la double barrière en valeur limite de HIC et de déplacement permanent maximal du siège, tendant
à converger autour de 20 kN pour l’effort plastique du « vérin » arrière.
Les valeurs déterminées par ces études paramétriques ont permis de pré-dimensionner les
structures tubulaires composites qui seront ensuite utilisées lors de cette étude et notamment
testées lors de campagnes d’essais quasi-statiques et dynamiques.
Le choix pour la fabrication des tubes en composite s’est porté sur une confection au moyen d’un
moule rigide formé de deux demi-coques en aluminium (en bleu sur la figure suivante) se refermant
sur une vessie gonflable en silicone (en vert sur la même figure) pour former le pourtour du tube (en
gris sur la figure) (cf. Figure [28]).
Une gaine de fibres brutes en tissu de verre (115 gr/ml) ou de carbone (HR 12K 136 gr/ml) imprégnée
de résine époxy (Epoxy-Araldite® LY 5052) est emmanchée sur un mandrin central, formé par la
vessie gonflable (en vert, Figure [28]). Le diamètre extérieur des tubes fabriqués a été fixé à 50 mm.
Figure [28] : Photos (en haut à gauche) et schématisation du principe de mise en œuvre (en haut au
centre) du moule et de la vessie gonflable pour la fabrication de tubes composites,
et vues 3D et éclatées du montage de fabrication (en bas) et données théoriques des épaisseurs de
plis permises en fonction du nombre de plis (en haut à droite).
Figure [29] : Images des tubes confectionnés, Carbone (au-dessus) et Verre (en-dessous).
Malheureusement les tubes issus de ce procédé de fabrication présentaient une porosité trop
importante pour pouvoir correctement tenir une sollicitation en compression et rendre un niveau de
SEA acceptable. Ceci a été confirmé par quelques essais préliminaires en compression sur ces tubes,
renvoyant une valeur moyenne de SEA de 30 kJ.kg-1, comparativement aux 120 kJ.kg-1 donnés par des
échantillons produits industriellement et aux valeurs moyennes issues de la littérature, de l’ordre de
50 à 80 kJ.kg-1 selon les configurations (cf. Chapitre suivant, détaillant ce sujet).
La forte baisse des performances de ces tubes en compression (ou crushing) s’explique par la forte
porosité présente dans ces tubes, telle qu’observée sur les images microscopique de la figure [30].
Figure [30] : Images issues d’observations microscopiques réalisées au moyen d’un système Alicona
Infinite Focus SL avec un grossissement x10 à x50,
avec à gauche, pour un tube tissé Carbone et à droite pour un tube tissé Verre.
Il était initialement prévu de faire varier le drapage (orientation des fibres et empilement), les
matériaux, et les types de tissage (sergé, satin, taffetas,…) pour tenter de séparer et évaluer
l'influence de ces différents paramètres sur la valeur de SEA obtenue. Cependant les problèmes et
délais de fabrications rencontrés ont conduit à s’orienter vers les tubes présentés dans la partie
suivante.
Conclusion partielle
L’étude numérique préalablement réalisée au moyen d’une structure de siège simplifiée et d’un
mannequin simplifié a permis de déterminer le chargement devant être supporté par la structure du
siège. Ce chargement a été optimisé de manière à garantir une déformation minimale du siège en cas
de crash, tout en dissipant suffisamment d’énergie pour respecter les critères biomécaniques de
sécurité envers le passager imposés par la norme.
Cette étude a également conduit au dimensionnement global de la structure composite qui sera
endommagée par crushing dans le but de dissiper de l’énergie.
La fabrication de tubes échantillons a été menée conjointement avec le CRITT Mécanique &
Composites de Toulouse. Par ailleurs des tubes composites de différentes compositions et issus
d’une fabrication industrielle ont été acquis en complément.
En effet, lors des essais préliminaires de compression quasi-statiques sur les tubes fabriqués au
moyen du moule demi-coques, sont apparues des propriétés mécaniques globalement insuffisantes,
notamment liées principalement à de fortes porosités, nuisibles pour la suite de l’étude, ce qui a
conduit à se tourner vers des tubes fabriqués industriellement.
Les tubes ainsi acquis ont été testés en écrasement quasi-statique et dynamique et ils ont de ce fait
servi de base à cette étude. Les résultats expérimentaux sont présentés en chapitre I et II.
th
[CHEN 2018] CHEN, E.-J. Aircraft Seat Row-to-row Head Injury Criteria (HIC) Simulation Using LS-DYNA®. In: 15
International LS-DYNA® Users Conference, 10-12 June 2018, Detroit, Michigan, USA.
[DESJARDINS 2006] DESJARDINS, S.P. The Evolution of Energy Absorption Systems for Crashworthy Helicopter Seats, Journal
of the American Helicopter Society, Volume 51, No. 2, 1 April 2006, Pages 150-163, DOI 10.4050/JAHS.51.
[DeWEESE 2004] DeWEESE, R.L., MOORCROFT, D.M. Evaluation of a Head Injury Criteria Component Test Device. Report,
DOT/FAA/AM-04/18, Office of Aerospace Medicine, Federal Aviation Administration, U.S. Department of Transportation,
Washington, DC, USA, November 2004.
[EASA 2009] EASA, European Aviation Safety Agency, Certification Specifications for Large Aeroplanes - CS25, Amendment
8, December 2009.
[EICHELBERGER 1985] EICHELBERGER, C.P., ALFARO-BOU, E., FASANELLA, E.L. Development of an energy-absorbing
passenger seat for a transport aircraft. In: NASA Conference Publication 2371, Proceedings of the 19th Aerospace
Mechanisms Symposium, SEE N85-33512 22-39, 1-3 May 1985, NASA Ames Research Center, Moffett Field, California, USA,
Pages 39-58, Unclassified, NASA, Scientific and Technical Information Branch, Langley Research Center, Hampton, VA, USA,
August 1985.
[EXPLISEAT 2011] Expliseat, The world’s lightest aircraft seat, site internet disponible à l’adresse: http://expliseat.com.
[HEIMBS 2014] HEIMBS, S., BERGMANN, T. Bearing Mode Absorber – On the Energy Absorption Capability of Pulling a Bolt
through a Composite or Sandwich Plate, International Symposium on Dynamic Response and Failure of Composite
Materials, DRaF2014, Procedia Engineering, Volume 88, 2014, Pages 149-156, ISSN 1877-7058, DOI
10.1016/j.proeng.2014.11.138.
[HUTCHINSON 1998] HUTCHINSON, J., KAISER, M.J., LANKARANI, H.M. The Head Injury Criterion (HIC) functional, Applied
Mathematics and Computation, Volume 96, Issue 1, 1998, Pages 1-16, ISSN 0096-3003, DOI 10.1016/S0096-3003(97)10106-
0.
[LANNISTER 1980] LAANANEN, D.H. Human body simulations for analysis of aircraft crash survivability. In: Proceedings of
the 1980 International IRCOBI Conference on the Biomechanics of Impact, Pages 95-106, 9-11 September 1980,
Birmingham, England.
[LAANANEN 1982] LAANANEN, D.H. Biomechanically based performance criteria for crashworthy aircraft seats. In:
Proceedings of the 1982 International IRCOBI Conference on the Biomechanics of Impact, Pages 207-219, 8-10 September
1982, Cologne, Germany.
[LAANANEN 1983] LAANANEN, D.H., BOLUKBASI, A.O., COLTMAN, J.W. Computer Simulation of an Aircraft Seat and
Occupant in a Crash Environment Volume I - Technical Report. FAA Technical Report, DOT/FAA/CT-82/33-1, Simula Inc.,
Tempe, AZ, USA - Federal Aviation Administration, Technical Center, Atlantic City Airport, NJ, USA, March 1983.
[LAANANEN 1990] LAANANEN, D.H. Crash survivability analysis of commuter aircraft seats and restraint systems. In:
Proceedings of the 1990 International IRCOBI Conference on the Biomechanics of Impact, Pages 143-154, 12-14 September
1990, Lyon, France.
[LAANANEN 1993] LAANANEN, D.H. Crashworthiness Analysis of Commuter Aircraft Seats. Report, DOT/FAA/CT-TN 91/28,
Federal Aviation Administration, U.S. Department of Transportation, Atlantic City Airport, NJ, USA, November 1993.
[LANKARANI 2007] LANKARANI, H.M. Development of a Component Head Injury Criteria (HIC) Tester for Aircraft Seat
Certification—Phase 2. Report, DOT/FAA/AR-06/47, Federal Aviation Administration, U.S. Department of Transportation,
Washington, DC, USA, January 2007.
[MARJOUX 2006] MARJOUX, D. Mécanismes de lésion de la tête humaine en situation de choc. Thèse de Doctorat en
Mécanique spécialité Biomécanique, Université Strasbourg I – Louis Pasteur, Strasbourg, France, décembre 2006.
[MARJOUX 2008] MARJOUX, D., BAUMGARTNER, D., DECK, C., WILLINGER, R. Head injury prediction capability of the HIC,
HIP, SIMon and ULP criteria, Accident Analysis & Prevention, Volume 40, Issue 3, 2008, Pages 1135-1148, ISSN 0001-4575,
DOI 10.1016/j.aap.2007.12.006.
[OLSCHINKA 2006] OLSCHINKA, C., SCHUMACHER, A., RIEDEL, D., Dynamic Simulation of Flight Passenger Seats. In:
Proceedings 5. LS-DYNA Anwenderforum, 12-13 October 2006, Ulm, Germany.
[SENATORE 2016] SENATORE, F., VITOLO, B., GUIDA, M., MARULO, F., CAPUTO, F. Numerical-experimental investigation on
st
the biomechanical performances of an aeronautical seat. In: Proceedings of the 1 International Conference on Impact
Loading of Structures and Materials (ICILSM), 22-26 May 2016, Turin, Italy.
[SHANAHAN 2004] SHANAHAN, D.F. Human Tolerance and Crash Survivability. In: Proceedings of the RTO HFM Lecture
Series on “Pathological Aspects and Associated Biodynamics in Aircraft Accident Investigation”, 28-29 October 2004,
Madrid, Spain.
[VAN BEEK 1998] VAN BEEK, J.A.B. Dynamic Seat Testing and Head Injury Criteria - An Aircraft Manufacturer’s Experience
and a View on the Future, In: Proceedings of the 21st Congress of International Council of the Aeronautical Sciences, 13-18
September 1998, Melbourne, Australia.
[WARRICK 1979] WARRICK, J.C., DESJARDINS, S.P. Development of Two Underseat Energy Absorbers for Application to
Crashworthy Passenger Seats for General Aviation Aircraft. Technical Report, NASA-CR-158927, 19840024297, Simula Inc.,
Tempe, AZ, USA - Langley Research Center, Hampton, VA, USA, September 1979.
Keywords: Crushing, Composite tubes, Energy dissipation, Specific Energy Absorption, Aeronautics
Abstract:
The purpose of this study is to evaluate and compare the ability of various composite structures to
dissipate the energy generated during a crash. To this end, circular composite tubes were tested in
compression in order to identify their behavior and determine their absorbing capabilities using the
Specific Energy Absorption (SEA) (energy absorbed per unit weight). Several composite tubular
structures with different materials and architectures were tested, including hybrid composition of
carbon-aramid and hybrid configuration of 0/90° UD with woven or braided fabric. Several inventive
and experimental trigger systems have been tested to try and enhance the absorption capabilities of
the tested structures. SEA values up to 140 kJ.kg-1 were obtained, achieving better than most
instances from the literature, reaching around 80 kJ.kg-1. Specimens with 0°-oriented fibers
coincidental with the direction of compression reached the highest SEA values while those with no
fiber oriented in this direction performed poorly. Moreover, it has consequently been established
that in quasi-static loading, a unidirectional laminate oriented at 0° and stabilized by woven plies
strongly meets the expectations in terms of energy dissipation. Incidentally, an inner constrained
containment is more effective in most cases, reducing the initial peak load without drastically
reducing the SEA value.
I.1. Introduction:
CFRP are known to be effective energy absorbing structures, due to the highly dissipative damage
mechanisms involved during severe solicitations [1,2,3,4]. For that reason, in addition to their
lightness combined with interesting mechanical properties, they are present in various domains:
automotive [5,6,7], railway [8,9], aeronautics aircrafts [10,11], and helicopters [12,13,14], all of which
may be subjected to brutal impact resulting in a violent crash.
Damage mechanisms resulting from the crushing of composite structures have been identified on
CFRP plates [15,16,17] and tubes [2,3,10,12,18,19,20,21] structures.
Failure mechanisms that contribute the most to the energy dissipation are depending on various
factors that include delamination, bending, kinking and fracture of the fibers as well as fracture of
the matrix [5].
Figure [1]: Catastrophic (a) and progressive (b) failure [THORNTON 1979][18], [THORNTON 1990][2],
progressive folding (c) and crushing (d) [HULL 1991][19].
In progressive crushing mode, damage mechanisms at the structural scale may be summed up into
three types: (i) splaying, (ii) fragmentation and (iii) debris creation and accumulation [17,19,21] (Figure
[2]).
Figure [2]: Major damage mechanisms occurring at the crush zone (adapted from [HULL 1991][19]).
Fragmentation might occur at two levels: under the tip of the plies (due to micro-buckling of fibers
for 0° plies and to multiple shear micro-cracks for 90° plies) or within the plies as intra-laminar
failure, fiber breakage and matrix cracks (due to a combination of compression, bending and shear)
[17,19] (Figure [2]).
In addition, [BRIGHTON 2009][21] links the cracks density – and therefore the energy absorbed – to the
radius of curvature of the splayed parts. A large radius of curvature leads to a lower number of
cracks, hence less energy absorbed, whereas a small radius of curvature leads to high crack density
and higher energy absorbed.
Figure [3]: Four identified failure modes for splayed fronds (adapted from [HULL 1991][19]).
The successive stages of composite tube progressive crushing have been well identified [19,22,23] as
displayed in Figure [4].
The localized fragmentation at the tip of the plies is pointed as the mechanism leading to the
definition of the ply Mean Crushing Stress (MCS) [16,24], corresponding to a stable crushing level
(Figure [4](f)), occurring after the yielding peak (Figure [4](b,c)) and stabilization phase (Figure [4](e)).
Recent studies take their interest in the Mean Crushing Stress characterization [16,17] on standalone
plies and within laminates as well as the influence of fibers orientation or braided structure effect
and the use of mix or hybrid materials on the crushing level.
Most research works focus on means to increase the Energy Absorbed while reducing the initial peak
load to enhance crushing performances. To that end, a multitude of factors have been studied.
Tube scaling [5,37], and especially the effect of thin-walled tubes and the influence of wall thickness
[14,21,25,27,30,46,51,52,54,56] have been considered, generally highlighting an energy absorption
increase trend with the increase of the wall thickness.
Investigations have also been carried out on materials types used to manufacture the structures,
mainly including Carbon, Glass and Aramid fibers, with a various range of polymeric resin:
Carbon [7,10,53] Carbon/Epoxy [2,5,8-14,16-18,20,21,32,33,35,36,38,39,41,46,50,57], Carbon/PEEK
[3,35,37,58], Carbon/Vinylester [39,45,51,52], Carbon/Polyamide [43], Glass/Epoxy [10,12,18,30,33,40,47],
E-glass/Polyester [5,49,42,28,29,26], Glass/Vinylester [42], Glass/Polypropylene [43,44], Kevlar/Epoxy
[1,5,8,9,18], hybrid materials [1,5], composite metal-fibers [25,54], hybrid Aramid/Carbon/Epoxy [13] or
even woven silk/Epoxy [55].
[CARRUTHERS 1998][59] proposes a review of the energy absorption capability and the crashworthiness
of composite material and metal structures, giving values recorded for axially compressed FRP and
metal tubes clearly highlighting the superiority of FRP structures over steel structures.
A comparison with steel and aluminum, with SEA values of 15 and 30 kJ.kg-1 respectively [21], places
FRP energy absorption capacity significantly above. [HAMADA 1992][35] reports a value of 53 kJ.kg-1 for
45° oriented Carbon fibers/Epoxy tubes while values range from 50 to 80 kJ.kg-1 for a variety of
Glass fibers reinforced thermosetting resin composites [35]. He also presents a value of 110 kJ.kg-1
obtained for Carbon/Epoxy tubes (from [19] and [Grundy 1983] (unpublished)). This in accordance
with [RAMAKRISHNA 1993][36] reporting values of 85 to 120 kJ.kg-1 for Carbon/Epoxy tubes. Those
values are significantly below the 127 kJ.kg-1 obtained for 30° oriented carbon fibers/PEEK tubes
and the 180 kJ.kg-1 obtained in the 0° carbon fibers/PEEK tubes [35].
The crushing behavior of composite tubes is found to be dependent on the fiber content and the
fiber architecture [35,36]. Tubes with lower fiber contents crushed irregularly whereas tubes with
fiber contents above 15% crushed progressively [36].
The relative amounts of 0° and 90° fibers as well as their position in the stratification of the tube’s
wall is a major factor that determine the geometry of the crush zone and therefore the specific
energy absorption [35].
Congruently, [CHIU 1998][38] finds that when varying the braided angle of composite laminate
structures for composite tubes, the smallest braiding angles produce the highest specific energy
absorption, up to 89 kJ.kg-1 for a 20° angle, (and 100 kJ.kg-1 for 15° [12]) and reports a clear decline of
the SEA value as the braiding angle increases, down by almost 50% to 45 kJ.kg-1.
[RAMAKRISHNA 1993][36] concludes that the specific energy absorption capability increases with an
appropriate fiber content and that the insertion of inlay fibers into the knitted fabric is an effective
method of improving the energy absorption capability of fabric composite tubes.
Improving the crushing initiation with specific trigger geometries or profiles [49], tulip shape [55],
notched outlines [1,29,42], different tapered angles [47], SMA trigger (shape memory alloy wires) [41],
inserting lateral circular cutouts [56] or by chamfering or beveling the edge [1,3,7,14,
16-21,26,28,29,32,34-38,42,51] or using a double-chamfer trigger (chamfered at both ends) [40] has been
attempted. The use of plug initiators [21,39,45,46,51-53] is also a recurring attempt to initiate and
enhance the crushing.
Finally specific boundary conditions with chamfer external triggers [46] or semi-circular cavity
external trigger [46] have recently been tested. [TONG 2018][46] states that the energy absorption is
improved by 53% by replacing a chamfer trigger with innovative trigger, with SEA values ranging from
45 up to 102 kJ.kg-1 for 2D-braided Carbon/Epoxy tubes.
To jump over the main points, two significantly exploitable axes of interest are standing out from the
literature: materials and fibers orientation choice on one hand and trigger initiation and boundary
conditions optimization on the other. These are therefore the two aspects presented in this study
(Part A and Part B).
The work presented in this study therefore aims at comparing the results in Energy Absorption
capability of hybrid composite circular structures with different fibers orientations and of different
natures (UD, woven) using specific boundary conditions and trigger initiation systems.
The present study will relate to the experimental testing results of several circular composite tubes
of different composition and stratification. More specifically various combinations of unidirectional
and woven structures as well as hybrid Carbon-Aramid reinforcement fibers are being tested.
A first part will focus on the effect of different stratification and materials in a simple free-face
crushing configuration between two compressive planes (Part A).
A second part will present the effect of different trigger initiation systems on the crushing
performances using mainly one singular sample, before highlighting specific features for each tube
sample in some of the configurations (Part B).
A variety of fiber/Epoxy tubes were acquired for testing, with different fibers orientations and fibers
types. These fiber-reinforced tubular structures were studied in axial compressive crushing. In total,
five specimens with different structures and different compositions were tested in various crushing
configurations. Structural and material basis for the specimens include 12K HR carbon fibers and
polymer Epoxy resin. Fibers orientation and laminate stratification differ from one specimen to
another as shown in Figures [5] and [6] and summarized in Table 1. Provided tubular structures were
machined and shaped in tubes of 100 mm length as pictured in Figure [5]. Medium diameter was set
at an average of 50 mm, with inner diameter varying from one sample to another due to stacking
differences. Stratification layout and tube wall thickness are summarized in Table 1 for each sample.
The Carbon/Epoxy combination was selected with Epoxy resin as a matrix because of its low density
and for its high strength and good mechanical properties with reliable chemical stability, as well as its
worthy performance regarding energy absorption based on the literature review and due to the
aeronautical context.
In order to verify and establish the composition and stratification of the composite tubes specimens,
samples were polished and observed using a high resolution optical microscope. Measurements and
images acquisitions were performed using an Alicona Infinite Focus SL microscope system with a x10
to x50 magnification. Resulting observations are displayed in Figure [6], along with a lay-up
schematization.
For the laminate lay-up schematization presented in Figure [6], the 0° direction of the fibers was
chosen to coincide with the longitudinal axis of the tube and subsequently with the axial crushing
direction. Plies dimensions are given as an averaged best approximation since plies thickness is not
even and regular. That is supposedly the results of the fabrication process.
Table 1 reports the structural specificities and geometrical properties for the 5 tube specimens.
C: Carbon A: Aramid
Table 1: Tubes specimens stratifications and properties.
When looking at the density values from Table 1, it can be pointed out that they are relatively low for
some samples (sample 1 especially, and to a lesser extent, sample 3). This is allegedly strongly related
to the high porosity observed in the samples (Figure [6]) and also lower fiber density (or fraction
volume vf) in some areas of the samples.
Table 2 intends a comparison in Stiffness and Compressive strength failure between experimentally
and theoretically obtained values for all five samples. Both the experimental and theoretical methods
used to achieve those results are presented below. The magnitude referred to as Stiffness (in MPa)
relates to the elastic compressive modulus (Young’s modulus).
Experimental Theoretical
Stiffness Compressive Stiffness Compressive
(MPa) strength (MPa) (MPa) strength (MPa)
+ 2800 + 100
Tube 1 43 700 - 5200 -350 - 139 54 200 -650
+ 4900 + 65
Tube 2 54 300 - 3100 -340 -102 52 100 -652
+ 8700 + 47
Tube 3 30 200 - 8600 -170 - 92 22 800 -265
+ 6800 + 44
Tube 4 27 300 - 6000 -180 - 59 24 500 -259
+ 3600 + 29
Tube 5 24 700 - 5400 -200 - 41 23 600 -250
For the five tubular specimens, materials and fibers properties are not well known and identified nor
completely mastered. Compressive experimental testing shows that fibers mechanical properties are
less resistant than usually encountered in current modern composite materials.
Table 3: Mechanical properties used for the Classical Lamination Theory calculations.
When comparing values from the theoretical calculations to the experimental data (Table 2), the
latter reflects lower values in stress failure, although it can be noted that they remain in the same
order of magnitude. Besides, the experimental compressive strength values that are reported in
Table 2 are rather related to a failure in crushing mode than pure compression.
Figure [7] presents the experimental mechanical stiffness modulus obtained from quasi-static axial
compression testing of samples 1 to 5 positioned under 2 crushing plane, used as a free-crushing face
for the bottom plane and being slightly encased on top for the upper plane, to avoid drifting. Actual
testing begins with the tube structure in contact at both ends, and being very slightly stress-loaded.
Figure [7]: Compression stiffness from experimental testing of axial crushing for tubes 1 to 5.
Results reported in Table 2 show that the compression stiffness is lower than anticipated for 3 of the
5 CFRP specimens, especially for specimens 1 and 3. Compression failure is also lower than
estimated, especially for specimens 1 and 2 that incidentally mainly present 0°-oriented fibers.
This supports the hypothesis of poor fibers’ properties used to manufacture the tubular specimens,
and it can also be explained by the high porosity inherent to many samples, as observed on the
microscopic images (Figure [6]). This is also most obvious for tube specimen 1, for which braided
thread remnants from the manufacturing process are visible too. In addition, experimental
compressive failure values resulting from the performed compression tests could rather be affected
by a bearing phenomenon under the tip of the plies and resulting from the crushing nature of the
solicitation than related to a pure compressive mode, hence the observed discrepancy in
compression strength failure, which thereafter seems more rational.
Quasi-static crushing tests were carried out using a 250 kN Schenck hydraulic testing machine in
compression testing mode, through a constant loading speed of 0.2 mm.s-1 (Figure [8]). To account
for repeatability, tube specimens were tested at least 3 to 5 times on average (and up to 10 times)
for each sample and each configuration.
Contact and alignment conditions have been verified. Flatness of the top and bottom support faces
of the tubes were verified and quantified between 0.02 and 0.05 mm, depending on the sample. Top
and bottom pans of the testing machine were supposed parallel, with a slight parallelism default
being observed and estimated at around 0.01-0.02 mm.
This first part presents the experimental results for the free face axial compressive crushing of the 5
CFRP circular tubes mentioned above between two plane surfaces.
Figure [9] represents the Stress-Displacement curve resulting from such trials, for tube specimen 1.
The stress thereby represented corresponds to an average value from multiple trials for the same
specimen and the same testing configuration, with the dispersion range indicated on both sides of
the curve, corresponding to the minimum and maximum stress value for all trials for each
displacement value.
Figure [9]: Stress-Displacement curve and dispersion for tube 1, resulting from 10 trials.
Generally it can be noted a relatively good repeatability in the trials, especially for composite
material, with a dispersion of -12% and +32% on average, compared to the medium value, as
illustrated in Figure [9]. Overall, less than 2% of the total of tested samples has been discarded for
being deemed aberrant.
A crushing curve such as the ones displayed in Figures [9] and [12] can be divided into three main
parts: a loading phase ending by the main rupture of the structure and leading to a peak, a
stabilization phase and a stable crushing phase [19,23]. Sometimes, when crushing is extended long
enough, a last phase known as compaction of debris or densification may occur, matching an
increase of the end of the curve.
Once reached, the level of the crushing threshold or crushing plateau, defining the Mean Crushing
Stress (MCS) [16,24], is very steady and regular, as well as constant from one trial to another. The
compression stiffness is also mainly identical from one trial to another. However, it can be observed
a large dispersion on the initial peak (Figure [9]) with random values being reached before structural
yielding and failure. A dispersion of more than 200% is observed for the initial peak (Figure [9]).
Figure [10]: Crushed CFRP tubes specimen 1 to 5, from the top (above) and from the side (below)
Apart from the multitude of resin and fibers debris generated during crushing, splaying (both inward
and outward) and large bands of material resulting from intra-laminar shear are a consequence of
composite laminate crushing, as seen on picture in Figure [10].
A difference can be observed between tubes samples 1 and 2 presenting both inner and outer
spreading of splayed parts and tubes samples 3, 4 and 5 mainly presenting outer spreading. This can
be explained mostly by the core structure of the tubular samples, with oriented-fibers pattern
(specimens 1 and 2) and unoriented-fibers pattern (specimens 3, 4 and 5).
Additionally, a specificity can be mentioned for samples with an aramid overlayer (tube specimens 4
and 5) as this latter acts as a girdle, drawing the shattered composite parts, resulting in a closer
folding and wrapping around itself.
Figure [11] gives the load over the axial displacement curves for the 5 tube specimens in axial
crushing. It can be noted that after the initial load peak (corresponding to the structure yield) has
passed and once the crushing phase has established and become constant, the range of crushing
load varies from 15 kN to 40 kN.
Figure [12] presents the crushing stress over the axial displacement for the 5 tube specimens in axial
crushing.
Two groups of CFRP tube specimens can be made from the Stress-Displacement curves resulting
from axial compression, with tubes 1 and 2 averaging a higher value of nearly 140 MPa for the MCS
plateau and tubes 3 to 5, lowering at 75 MPa. This is directly resulting from the structure and
stratification of the tubes specimens, which can consequently be separated into two groups: group 1:
oriented-fibers tubes (tube specimens 1 and 2) and group2: unoriented-fibers tubes (tube specimens
3, 4 and 5).
Finally, it can be noted that the length of the stabilization phase is directly linked to the wall
thickness of the tube specimens; it corresponds to approximately twice the thickness (Figure [12],
Table 1).
Figure [13]: Stress/density over Displacement curves of experimental crushing for tubes 1 to 5.
One means to characterize and compare the absorbing capability of materials is through the Specific
Energy Absorption (SEA), also referred to as Specific Sustained Crush Stress. The SEA value is given by
the following equation (Eq. 1):
EA 1 u
SEA = = ∫ (u∗ )du∗ Eq. (1)
m . u 0
where EA is the Energy Absorbed (given by the area under the force-displacement curve), divided by
the mass of the crushed mater m. is the compression stress, u the crushing distance, and the
density of the material.
It can be established that the SEA value can be very closely approximated as an instantaneous value
using the crushing stress cr divided by the density of the crushed material (Eq. 2).
𝑐𝑟
SEA → Eq. (2)
→ 𝑐𝑜𝑛𝑠𝑡
Similarity in behavior and energy absorbing capability for specimens 3, 4 and 5 are not incoherent
when referring to Table 1, which shows the same basis structure for those specimens. Outer aramid
covering does not improve or worsen the general behavior or energy absorbing capability.
Once the transition phase has passed, the SEA value tends toward a constant value, as displayed in
Figure [14]. The influence of the peak can be noticed at the very beginning of the curve and a
displacement of at least 10 mm is needed to erase the effect of this peak.
Figure [14]: SEA evolution of the 5 CFRP tube specimens submitted to axial crushing.
The histogram chart shown in Figure [15] displays the average SEA values for each tube sample in
free axial crushing. They are presented side-by-side with a factor, referred to as overshoot, and
defined by the crushing stress initial peak maximum value divided by the density (Eq. 3). This
indicator was chosen to represent and compare the overflow of energy for each sample.
𝑚𝑎𝑥
Overshoot = Eq. (3)
As the overshoot is uniform to the SEA (kJ.kg-1) a direct comparison between these two values is
possible and it can be revealed that for each specimen, the overshoot is at least twice as high as the
SEA value (x2 for tube specimen 3, x2.5 for tube specimen 1 and up to x2.8 for tube specimen 5).
Some instances in the literature refer to the Load Uniformity or Trigger Ratio defined as Fmax / Fmean
or Crushing Load Efficiency defined as Fmean / Fmax [48]. The overshoot parameter was deemed more
suited to represent the overflow of energy than the other parameters or ratios from the literature.
Furthermore, as it is homogenous to the SEA, the comparison between the two magnitudes is made
easier.
In order to limit the peak and therefore reduce the overshoot, several boundary conditions and
crushing initiations have been tested.
Intermediate conclusion
To summarize the first part of this experimental work, the following observations can be made.
Failure mechanisms remained similar for all tube samples, with moderate to significant splitting,
delamination and multiple brittle crack paths, leading to large debris creation.
Tubular specimens with 0° fibers oriented in the loading axial direction perform better in
compressive strength and therefore return higher SEA values.
Woven fabric plies help containing and guiding the unidirectional plies, restraining them from
splaying too easily.
Aramid fibers bring no additional rigidity or energy dispersion capability but aramid covered tube
samples contain the fragmented carbon/matrix wreckages better, avoiding large spillage.
Several configurations were tested for crushing initiation improvement and hopeful energy
absorption enhancement. Tubular structures were clamped and encased at one end at the top, while
several options were tested for the other end at the bottom. They were (a) let free on a plane
surface, (b) also encased, (c) encased while guided through a conic shape, (d) positioned on a conic
plug initiator, and (e) submitted to pure flaring, where the structure gradually becomes wider from
one end to the other, as a conic part passed through. Figure [16] presents the five configurations
tested. The general motivation for each testing configuration was (a) to use a reference case, similar
to what is found in the literature, (b) try and enhance the SEA value, (c) try and reduce the
overshoot, (d) inversely to (c) test an outer dispersion, that could be compared to (a) and (e) with the
later aiming at testing and evaluating the benefits of the crimping property alone.
For the two conic-shaped configurations ((c) and (d)), specific conic-shape parts were designed and
machined for each specimens according to the specificity of each tube’s internal diameters (Table 4),
with a goal of expressing a hoop strain of 20000 µ and –15000 µ respectively via the gradual slope.
Those values are usual approximations of tension and compression stain ruptures for composite
materials.
Figure [17]: Outer (left) and Inner (right) conic-shaped parts used to change the boundary conditions.
Table 4: Conic-shaped parts specifications and dimensions for inner and outer cones.
For the outer cone (forcing a hoop strain of 0.020), the diameter of the base of the conic part (D int)
was made to match the Interior diameter (Int. diam.) of the tube and for the inner cone (forcing a
compressive hoop strain of – 0.015), the diameter of the base of the conic-shaped part (d ext) was
made to match the Exterior Diameter (Ext. Diam.) of the tube. Specifications are reported in Table 4.
Figure [18]: Crushed CFRP tubes (1) showing outer spreading (left) and inner folding (right)
based on boundary condition (a) free Crushing and (b) Inner Crushing, respectively.
As an illustration, the pictures in Figure [18] present the difference in crushing behavior for the same
CFRP tube specimen (tube 1) submitted to axial crushing under the first two configurations ((a) and
(b)) presented in Figure [16]. Both underwent progressive crushing. The first (left) corresponding to
configuration type (a) presents both inner and outer spreading of splayed parts, whereas the second
(right) corresponding to configuration (b), reveals the whole bundle of splayed parts folding towards
the inside of the tube.
Configuration type (d), with a cone-shaped plug initiator, does not differ much from configuration
type (a) apart from the fact that the entire splayed bundle spreads on the outside and configuration
type (c), with an inner-conic-shaped part, does not differ from configuration (b). This can also be
correlated by the tomographic images.
Figure [19] displays the Stress/density over Displacement results for tube specimen 1, for the 5
described configurations. In case of inner-crushing confinement (configuration (b)), a slight but still
significant increase of the curve can be noted towards the end, starting at 40 mm (), corresponding
to the densification of debris and shattered parts. This rise should even be starting sooner, at about
23 mm (half the tube’s interior diameter)(), when the wall’s inferior end meets at the center of the
tube (Figure [20]).
Figure [20]: Schematic representation of the tube’s wall convergence and collision at the center of
the tube, for the Inner Crushing configuration (with tube specimen 1 dimensions).
Surprisingly, this outcome is not seen for configuration (c) (Inner Conic Crushing) where the end of
the tube’s wall should also meet at the center. This might be explained by the fact that the tube’s
wall’s inner plies are more damaged and fractured by the friction and the progressive confinement
induced by the conic sloped part or that the outer woven ply is less damaged by the slow progressive
sliding, which would help keeping the UD plies in place, softly folding and guiding them inwards.
This increase of the stress at the end could be beneficial and valuable for a surge in the SEA value:
the densification of partially crushed material inside the tubular structure stabilizes the crushing
process, resulting in an increase of the mean crushing stress and therefore the SEA.
Figure [21]: Outer and Inner conic parts and initial tube position.
The very beginning of the Stress-Displacement curve for those two configurations matches the last
configuration (e), where the conic-shaped plug widens the extremity of the tubular structure. This
setback is incidental to the height of the conic shape () (Figures [17], [21]).
Pure Flaring (configuration (e)) was tested to try and take advantage of the crimping property of
woven structures in braided composite tubes (Figure [22]). Furthermore, the idea was to compare
configuration (d) with configurations (a) and (e) and evaluate if the total energy dissipated by (d) was
the summation of (a) and (e). Needless to say when referring to Figure [19], that this is hardy the
case, with configuration (e) only dissipating 1 or 2% of the energy compared to pure axial crushing
(configuration (a)).
In 2D-braided or woven patterns, the crimping can be characterized by the ratio of the real length of
the fully deployed fiber (B) divided by the actual length of this fiber within the woven pattern (A)
(Figure [22], Eq. 4):
B𝑙𝑒𝑛𝑔ℎ𝑡 × 100
Crimping = Eq. (4)
A𝑙𝑒𝑛𝑔ℎ𝑡
Overall, although trying to make the most of the crimping configuration in woven patterns seemed
promising and worth investigating, this attempt unfortunately proved to be unpractical and non-
optimal in terms of energy dissipation afterwards.
Figure [23]: SEA and overshoot values for tube specimen 1 for the 5 configurations.
For each axial compression test, a stiff peak can be observed on the Stress/density-Displacement
curve as showed in Figure [19], when the structure yields before it starts collapsing by progressive
crushing at a stable and constant stress, as reflected by the continuous plateau level. Ideally the gap
between the peak and plateau value has to be reduced to a minimum, as a small gap and constant
plateau level means an optimized dissipation of energy for a fixed and given load value.
In case of Outer Crushing, the use of a conic plug initiator (d) to introduce a radial flaring of the tube
structure before it is submitted to crushing does not improve the energy absorbing capability,
compared to pure and plain crushing (a), as shown in Figures [19] and [23]. Configuration (d) does
not reduce the overshoot either (Figure [23]). Moreover, this configuration worsens the energy
absorbing capability at it significantly lowers the SEA value (Figure [19] and [23]). At last, the use of
the conic plug alone, passing through the tube along its whole length and inducing an axial flaring of
the structure (configuration (e)) hardly dissipate any energy, leading to the conclusion that the
expansion of the crimping fibers is not a primordial mechanism in composite absorption capability.
For this reason, results from configuration (e) will mostly be disregarded in the following discussions.
Post-testing X-Ray micro-measurement observations were conducted to observe and determine the
damage mechanisms involved during crushing on the inside of the tubes’walls thickness. X-Ray 3D-
micro-computed tomography images were performed using a Micro-Tomography EasyTom 130
machine, manufactured by RX Solutions, France. The tubular specimens were placed at a distance of
91.3 mm from the source. The source has a voxel size of 18 μm. Each specimen was scanned through
a 360° rotation using a Varian PaxScan 1313DX imager to capture layer-by-layer 2D X-ray images used
for full-scale 3D reconstruction. RX Solutions X-Act 2.0 software was used for 3D reconstruction and
post-processing. Due to the samples size and dimensions the maximum possible and workable
resolution was 18 μm. The source-object distances (sod) and source-detector distances (sdd) were
91.3 mm and 643.9 mm respectively, which determine the magnification (sdd/sod) at 7.05. The X-ray
voltage and current were set to 60 kV and 133 mA respectively. Each sample was scanned for 160
min with 0.6 s per projection.
Figure [24]: Tomographic images and schematizations of crushed CFRP tubes showing major damage
and plies dispersion for specimen 1 in four configurations ((a), (b), (c) and (d)).
Figure [24] highlights major occurring damage mechanisms resulting from progressive crushing
instigated by quasi-static axial compression load for tube specimen 1 through the means of
tomographic imaging for four different crushing modes ((a), (b), (c) and (d) when referring to Figure
[16]). Inner (a), (b), (c) and outer (a), (d) splaying of fragmented or un-fragmented parts are clearly
visible, as well as debris accumulation.
The crushing plane surface is schematized by a discontinued line on all pictures in Figure [24]. It is
strongly suspected that for each tube samples, folded plies at the end of the tube walls between the
sane part of the tubes and the crushing surface moved back downward when the crushing force was
unloaded, due to a spring-back effect. Hence an estimated positioning of the crushing surface
appearing to be situated within the tube and entering inside the tube structure. Similarly, positions
of the boundary parts and conic-shaped parts were added.
For the first configuration (a), a pyramidal-shaped debris accumulation can be observed at the center
under the tube wall, between the tube section and the crushing surface, where the laminate plies
spread towards the inside or the outside of the tubular structure. This debris accumulation forms
from the void created by the plies splaying from the center under the tube wall then subsequently
helps further and heighten this splaying.
At the contact surface between the tube circular section and the plane surface, micro-bucking occurs
progressively and successively, resulting in matrix and/or fibers fragmentation for the laminate plies
that undergo such splaying. This damage mode increases the debris formation. As observed and
mentioned in previous study, the formation and the evacuation or accumulation of debris remains
rather random.
For the second test configuration (b), damage mechanisms are similar in type and classification, but
are all shifted toward the inside of the tube, since the outer wall is encased, leaving no leeway for
splayed plies to spread that way. Incidentally, although major occurring damage are the same, their
origin varies somewhat. Bending and folding resulting from the compression generate the splaying of
the composite plies, and therefore the fracture and fragmentation for those sustaining a higher
stress. Fragmentation still generates an important number of debris that varies in forms and shape,
but those are freely evacuated from the crushing zone under the tube section to the inside of the
tube. They do not form a tip that parts the laminate wall and split it, as observed with the pyramidal
debris accumulation on the first configuration.
At the extremity of the tube wall, where the folding appears, outer plies seem to be more submitted
to bending whereas inner plies undergo plain compression and break through intra-laminar fracture.
According to the tomographic imaging, configuration (c) is almost identical to configuration (b) but
appears to be less densified at the center, on the inside of the tube’s wall. This might explain why no
increase of the crushing stress is observed on the crush-displacement curves for that configuration.
The inner slope inclination favored the “folding” of the inner fabric ply by guiding it, contrary to
configuration (b) where it was more abruptly fractured.
For the last test configuration (d), outer splaying is mainly predominating, since the conic shape
blocked the inner splaying and spreading towards the interior of the tube. On the interior side of the
tube, splaying is almost nonexistent. However fragmentation is intensified, with large fractures, of
both the fibers and the matrix. Additionally, the tube wall went back to its initial position, once the
metallic conic part was withdrawn, due to a vertical spring back effect and the wall’s rigidity.
Those inner plies are the most impacted and the most solicited plies during crushing, creating a
localized fragmentation that leads to a localized crushing which characterizes the Mean Crushing
Stress defined by [ISRAR AHMAD 2013][16,24]. This would explain the overall crushing stress (crushing)
always averaging around 150 MPa for every configuration tested (except configuration (e), which
underwent no crushing at all) as exhibited in Figure [25]. It can be inferred from that observation that
this definite number of fragmented plies is optimum in order to optimize the SEA value.
When related to the 250 MPa constant crushing stress value highlighted by [ISRAR AHMAD 2013][16]
on standalone plies independently of the orientation (0 or 90°), it can be inferred that crushing can
still be optimized for this structure (tube 1).
Figure [25]: Mean Crushing Stress values for tube specimen 1 for the 5 configurations.
Overall, and notwithstanding the current results, it is essential to keep in mind that configurations (d)
and (e) are highly dependent on the dimensions of the conic plug initiator, which may have been
improperly chosen to achieve the goal of reducing the overshoot. A more complete and focused
study on more adequate dimensions of the conic slope may be needed, with a series of tests
imposing a gradual strain deformation (0.005, 0.010, 0.015, 0.020, 0.025…) for instance.
Some additional and nonetheless notable observations were incidentally made regarding other tube
specimens in specific crushing configurations, as pointed out in the following section.
For the inner-crushing configuration (b), tubes specimens 3, 4 and 5 displayed two major modes of
failure. Figure [26] presents that distinction in failure behavior mode observed for some samples in
inner-crushing configuration (b) using tube specimen 3 as an example. Those two modes, namely
catastrophic failure and progressive crushing, were reported to occur by [THORNTON 1979][18],
[THORNTON 1979][2], and previously illustrated in general crushing case by Figure [1](a) and (b).
Figure [26]: Stress-Displacement curve and dispersion for tube specimen 3 in Inner Crushing
configuration (b).
Although in most instances tube specimens 3, 4 and 5 behaved according to a usual progressive
crushing mode, in some cases, they underwent catastrophic failure from a middle point of the
tubular structure and without undergoing progressive crushing from one end. Visual illustrations of
the catastrophic failure mode demonstrated by those samples are given with pictures in Figure [27].
This unwanted failure mode drastically lowers the SEA value for these samples in this particular
configuration to 13.1 kJ.kg-1 for specimen 3 and 26.5 and 27.4 kJ.kg-1 for specimen 4 and 5, while they
stand at 52.8, 44.7 and 52.8 kJ.kg-1 respectively when progressive crushing occurs in this
configuration.
The ratio of catastrophic failure stands at 25%, 33% and 50% for tube specimens 3, 4 and 5
respectively, in inner-crushing configuration (b), when these tubes displayed a 99.8% progressive
crushing successful behavior in all other configurations.
In practical terms, inner-crushing configurations reduced the production and scattering of debris, as
the undamaged part of the tube act as a sheath as can be seen on pictures in Figure [18]. To that
extend, tube specimens 4 and 5 (the two specimens that have an aramid cover on top of the carbon
structure) do not present any interest regarding the SEA value, but the aramid cover acts as a girdle,
helping in folding and containing the fragmented parts, both in case of inner and outer crushing. This
comes at the expense of a lower compression stiffness and a lower tubular section to conserve a
similar density, when comparing with tube sample 3 and referring to Table 1.
Figure [28] summarizes the Specific Energy Absorption average values obtained for all tube samples
(1 to 5) in all configurations ((a) to (e)), with dispersion marks for each testing.
Figure [28]: SEA average values for the 5 CFRP tube specimens submitted to axial crushing
in 5 configurations.
For tube specimen 1, inner-crushing (b) seems to enhance the SEA while inner-conic crushing (c)
lowers the value (for the chosen conic dimensions). A complementary study on the conic slope
inclination might be needed to test different angle dimensions and confirm that result.
For tube specimen 2, inner-crushing (b) enhances the SEA and inner-conic crushing (c) seems to
improve it a little more (with the margin of dispersion taken into account). This improvement might
come from the 90°-oriented plies that are more solicited in confined compression. As a result, the
overall crushing stress increases and therefore so does the SEA.
Tube specimen 3 performs the most irregularly. In addition to the catastrophic failure mode
observed for configuration (b) – which has been discarded for the SEA average calculation in Figure
[28] – tube specimen 3 shows the biggest dispersion, especially in configuration (b) (Inner Crushing).
The lowest SEA values obtained for this sample may be the result of some crushing instabilities,
which might be the consequence of the small thickness of the tube’s wall, as that specimen displays
the thinness wall thickness when referring to Table 1. Too thin a wall thickness may lead to some
unstable crushing phenomena and most notably affect the buckling modes.
Inner Conic Crushing (configuration (c)) seems to work better for specimens with 90°-oriented fibers
(specimens 2, 4 and 5, except specimen 3) for the same reasons mentioned above (i.e. confined
compression and adequate tube wall thickness), while it can also be noted that they all contain some
plies with a fiber orientation at 0° or close to 0° (specimen 2 has 0°-oriented fibers, specimens 4 and
5 have 20°-oriented fibers).
For all samples (1 to 5), Outer Conic Crushing (configuration (d)) lowers the SEA value and degrades
the structure absorbing capacity. Failure phenomena involved are primarily the same as described by
[HULL 1991][19] and presented in Figure [3]. For this configuration, while fiber rupture in traction is
energetically very dissipative, it is also very localized, whereas crushing, which is slightly less
dissipative, could repeatedly happen on a whole area.
The difference in results between inner and outer crushing (both in plain and conical configurations)
is fully in accordance with the statement formulated by [BRIGHTON 2009][21] on the relation between
cracks density and the radius of curvature of the splayed parts.
I.4. Conclusions:
Quasi-static axial crushing tests were performed on composite circular tubes with 5 different
structural compositions and using 5 different trigger initiation configurations.
SEA values up to 140 kJ.kg-1 were obtained, achieving better than most instances from the literature,
averaging around 50 kJ.kg-1 and reaching up to around 100 kJ.kg-1 for braided Carbon/Epoxy
structures.
Specimens with 0°-oriented fibers in the axial loading direction achieved better in energy absorption
than specimens with no fibers in that direction. It has consequently been established that in static
loading, a unidirectional laminate oriented at 0° and stabilized by woven plies strongly meets the
expectations in terms of energy dissipation. Incidentally, an inner constrained containment is more
effective in most cases, reducing the initial peak load without drastically reducing the SEA value.
Woven reinforcement on the inner and outer wall structure proved more effective than solid
strengthening supports such as metal constraints or poured resin in providing stability to the
structure. Additionally, the woven plies help containing the 0°-oriented fibers from splaying and
flaring too easily. Moreover, as such woven plies are structurally required for stability reasons, it is
opportune to try and put them in beneficial use, hence the inner oriented crushing concepts, which
load and stress these fibers in crushing.
However, the additional aramid draping was proven needless in term of energy absorbing capacity.
Yet this overlapping covering may be valuable in acting as a net to refrain outer spreading by
directing splayed chunks and debris towards the inside and keep brittle parts within the inside of the
tubular structure, avoiding expelled debris, as could be required in an aeronautical context.
In order to complete this study and provide complementary understanding regarding the crushing of
fiber-reinforced composite tubes and SEA enhancement, considerations may be given to the
following actions:
- varying the strain rate and conducting a study on dynamic crushing,
- varying the slope inclination and the conic dimensions, especially for the inner-conic concept,
- chamfering the samples and combining that trigger initiation technique with the presented
boundary conditions,
- using mechanically known fibers and materials, that is to say constituents for which properties are
independently tested and identified to permit the best selection.
Acknowledgments:
Special thanks are addressed to the CRITT Mécanique & Composites of Toulouse for assisting in
performing the tomographic imaging used in this article. Financial support for this study was
acquired thanks to FEDER funds through the joint project SKYSEAT co-funded by ARTEC Aerospace
Company, the Occitanie – Pyrénées-Méditerranée Region (ex-Midi-Pyrénées Region) and the ISAE-
SUPAERO.
References:
[1] [FARLEY 1983] FARLEY, G.L. Energy Absorption of Composite Materials, Journal of Composite Materials, Volume 17,
No. 3, May 1983, Pages 267-279, ISSN 0021-9983, DOI 10.1177/002199838301700307.
[2] [THORNTON 1990] THORNTON, P.H. The Crush of Fiber-Reinforced Plastics. In: CHEREMISINOFF, N.P. (Ed.), Handbook
of Ceramics and Composites Volume 1: Synthesis and Properties. Marcel Dekker Inc., New York, New York, 1990,
Pages 307-337, ISBN: 0-8247-8005-1.
[3] [HAMADA 1995] HAMADA, H., RAMAKRISHNA, S., SATOH, H. Crushing mechanism of carbon fibre/PEEK composite
tubes, Composites, Volume 26, Issue 11, 1995, Pages 749-755, ISSN 0010-4361, DOI 10.1016/0010-4361(95)98195-Q.
[4] [MAMALIS 1997] MAMALIS, A.G., ROBINSON, M., MANOLAKOS, D.E., DEMOSTHENOUS, G.A., IOANNIDIS, M.B.,
CARRUTHERS, J. Crashworthy capability of composite material structures, Composite Structures, Volume 37, Issue 2,
February 1997, Pages 109-134, ISSN 0263-8223, DOI 10.1016/S0263-8223(97)80005-0.
[5] [THORNTON 1985] THORNTON, P.H., HARWOOD, J.J., BEARDMORE, P. Fiber-reinforced plastic composites for energy
absorption purposes*, Composites Science and Technology, Volume 24, Issue 4, 1985, Pages 275-298, ISSN 0266-
3538, DOI 10.1016/0266-3538(85)90026-0.
[6] [JACOB 2002] JACOB, G.C., FELLERS, J.F., SIMUNOVIC, S., STARBUCK, J.M. Energy Absorption in Polymer Composites
for Automotive Crashworthiness, Journal of Composite Materials, Volume 36, No. 7, April 2002, Pages 813-850, ISSN
0021-9983, DOI 10.1177/0021998302036007164.
[7] [BISAGNI 2005] BISAGNI, C., DI PIETRO, G., FRASCHINI, L., TERLETTI D. Progressive crushing of fiber-reinforced
composite structural components of a Formula One racing car, Composite Structures, Volume 68, Issue 4, May 2005,
Pages 491-503, ISSN 0263-8223, DOI 10.1016/j.compstruct.2004.04.015.
[8] [KIM 2009] KIM, J.S., YOON, H.J., LEE, H.S., KWON, T.S. Energy Absorption Characterization of Composite Tubes for
Railway Application. In: Proceedings of the 17th International Conference on Composite Materials (ICCM), 27-31 July
2009, Edinburgh, Scotland, UK.
[9] [KIM 2011] KIM, J.-S., YOON, H.-J., SHIN, K.-B. A study on crushing behaviors of composite circular tubes with different
reinforcing fibers, International Journal of Impact Engineering, Volume 38, Issue 4, April 2011, Pages 198-207, ISSN
0734-743X, DOI 10.1016/j.ijimpeng.2010.11.007.
Keywords: Dynamic crushing, Composite tubes, Energy dissipation, SEA, Crash, Aeronautics
Abstract:
The purpose of this study is to complete a previous study on the quasi-static crushing of composite
tubes and evaluate the influence of the dynamic effect. The aim is to compare the ability of diverse
composite tubular structures to dissipate the energy generated during a crash. To this end,
composite tubes were tested in dynamic crushing through the mean of a drop test in order to
identify their behavior and determine their absorbing capabilities using the Specific Energy
Absorption (SEA) (energy absorbed per unit weight). Various composite tubular structures with
different materials and architectures were tested, including hybrid composition of carbon-aramid
and hybrid configuration of 0/90° UD with woven or braided fabric. Several inventive and
experimental trigger systems have been tested to try and enhance the absorption capabilities of the
tested structures. While it was established that in static loading, a unidirectional laminate oriented at
0° and stabilized by woven plies strongly meets the expectations in terms of energy dissipation, in
dynamic loading, it has been observed that a stratification with 90°-oriented fibers stabilized by
woven plies performed better and best meets the expectations in terms of energy dissipation.
Furthermore, unidirectional laminate oriented at 0° performed lesser by 50% between quasi-static
and dynamic rates. Incidentally, an inner constrained containment is more effective in most cases,
reducing the initial peak load without drastically reducing the SEA value.
II.1. Introduction:
Many studies regarding composite structures crashworthiness and inherent energy absorption
capabilities are carried out under quasi-static loading conditions. However, to fully apprehend the
structural behavior and assimilate the crash aspect, this topic requires an approach that must
consider the dynamic aspect. Furthermore, automobile impact and aircraft crash are generally
categorized as dynamic events, thus involving high velocity [1]. Passenger safety is consequently a
paramount issue [1,2,3], as human tolerance and crash survivability are utterly dependent on the
energy absorbing systems or damageable structural parts positioned within the transportation
structure. The purpose of those parts, sometimes referred to as “shock absorbers”, is to reduce the
loads or accelerations induced by a violent crash and acting on vulnerable components or
passengers. In such systems, energy dissipation is usually achieved by controlled deformation or
destruction of an energy absorbing structure.
Successive stages of composite tube progressive crushing have been well identified [8,9,10] in quasi-
static loading, as displayed in Figure [1].
During a progressive crushing mode (Figures [1] and [2](a)), damage mechanisms at the structural
scale may be summed up into three types: (i) splaying, (ii) fragmentation and (iii) debris creation and
accumulation [8,11,12] (Figure [2](b)).
Figure [2]: Major damage mechanisms occurring at the crush zone (adapted from [HULL 1991][8]).
Figure [3]: Schematic representation of the ‘shear’ kink formation chronology (a),
schematic of the transition zone between fully-kinked and unkinked zone (b) [EVANS 1978][13],
imaging of a typical kink-band geometry (c) [GUTKIN 2011][14].
Figure [3] displays a schematization of the kink-band phenomenon (a) and (b) and imaging
illustration (c). The kink-band formation is described as a band of matter with its edges lying at angles
+ 0 and . It is induced by the shear stress A due to combination of (11, 12) in a coordinate
system aligned with the rotated fibers. The angle + 0 corresponds to the angle of rotation of the
fibers within the kink-band (with 0 the initial misalignment and the additional rotation introduced
by the compressive loading) and corresponds to the kink-band propagation [14].
Figure [4]: Schematic representation of damage mechanisms in 0° (left) and 90° (right) plies
(from [CHIU 2015][15]).
Most research works that focus on crashworthiness and energy absorption improvement tend to
take advantage of those rupture modes as a mean to increase the Energy Absorbed while reducing
the initial peak load to enhance crushing performances.
The influence of many parameters on the energy absorption capabilities of composite structures
under quasi-static and dynamic loading at various test speeds (speed rates) has widely been studied.
Many studies have evaluated and compared the energy absorption ability of composite structures
under different testing conditions and have tried to enhance that capacity by interacting with several
structural parameters or properties.
Investigations have been carried out on the choice of materials, such as aluminum [1,16,17] or steel
[18], and mainly FRP [19,20,21], with a various range of fibers and polymeric resins types
(Carbon/Epoxy [1,11,12,15,17,21-39], Carbon/PEEK [25], Carbon/Vinylester [40,41], Carbon/Polyamide
[42], Glass/Epoxy [1,22,24,27,34,37,43-46], Glass/Polypropylene [11,42], Glass/Polyester [5,47,48],
Glass/Vinylester [5,40,49], Aramid (Kevlar®) [17,22,23,27,29,31,35,37], and hybrid materials [18,34], or
Carbon-Kevlar/Epoxy [25]). [BROWNE 2007][17] reports a 20 to 62% increase using unidirectional carbon
fibers tow with adjacent oriented fabric plies compared to aluminum construction, with an average
SEA value of 32.5 kJ.kg-1. Carbon/PEEK and Carbon/Epoxy are reported as the most performant
materials combination in term of Energy Absorption by most of the studies.
Improving the crushing initiation with specific trigger systems or techniques has been widely
attempted, using different trigger geometries or profiles [39,47], such as notched outlines (sometimes
referred to as tulip forms) [5,15,47] or with tapered angle [46], via chamfered or beveled edges
[5,11,12,17,23,24,28,29,33,36-38,40,45,48], double-chamfered trigger (tubes chamfered at both ends) [43],
and also with the use of a plug initiator [11,27,29,32,39,41] to initiate and enhance the crushing.
Chamfered-edge tubes produced the best results in term of initiation and energy absorbing capacity
according to [BRIGHTON 2009][11], but the use of a plug initiator tremendously reduces the SEA [11,41].
Furthermore, regarding the size, tallest plug most reduced the SEA capacity, still according to
[BRIGHTON 2009][11]. Moreover, concerning such plug initiator triggers, it has been reported that the
plug radius has a strong effect, with an energy absorption capability decreasing all the more that the
radius is important [11,29] compared to none or 0 radius plug.
Regarding the dynamic influence, for [BRIGHTON 2009][11] SEA values of Carbon/Epoxy tubes decrease
for increased testing speed, specifying that the 45° chamfer Carbon/Epoxy tests showed the highest
SEA and the least dependence on loading speed [11]. A notable decrease is also observed by
[McGREGOR 2016][41] in dynamic loading when using a plug initiation form.
Focusing on the material type, [SCHMUESER 1987][22] reports a decrease in energy absorption of
dynamic impacted Carbon/Epoxy, Glass/Epoxy and Kevlar/Epoxy tubes as compared to quasi-static
loading. With the ply orientation taken into account, [FARLEY 1991][23] finds no difference between
low and high velocity, reporting that specific energy absorption is independent of crushing speed for
0°+oriented structures (labelled [0°]). However, he remarks an increase in energy-absorption
capability for pure [] structure, reporting the magnitude change to be a function of ply orientation
(45° and 75° in his study). [DAVID 2013][33] finds similar results for pure [] structure (at 45°)
hence joining [FARLEY 1991][23] on crushing speed and orientation dependence.
[SCHULTZ 2001][26] however, despite reporting a clear energy absorption drop between dynamic and
static testing, reports only a very slight but not significant nor constant lay-up and fibers orientation
influence. The effect of fibers orientation and the presence of 0° unidirectional plies in the lay-up
structure appear to have a limited influence since reported performances are not repeatable from
one sample type to another and the effect of unidirectional plies is not observable on some samples.
Figure [5]: Fiber orientation effect on the SEA under quasi-static and impact crush condition
[WANG 2016][36].
Finally, while acknowledging a velocity effect by observing a decrease of energy absorption with the
increase of impact velocity, [MURALIKANNAN 2010][44] reports that the basic modes of deformation for
tubular structures are the same for both quasi static and dynamic loading.
Table 1 reports synthetized outcomes resulting from some previous studies regarding the influence
of the testing condition from Quasi-Static to Dynamic loading on the SEA evolution.
Table 1: Comparison in SEA percentages drops from Quasi-Static to Dynamic for several studies.
Table 2 reports synthetized results from previous studies regarding the influence of the plug size on
the SEA evolution in Dynamic loading condition.
In Dynamic loading
[BRIGHTON 2009] [11] plug size (height) up to 55% i
[McGREGOR 2016] [41] plug initiation from 23% i to 32% i
Table 2: Comparison and highlights in SEA percentages drops in Dynamic crushing for different
conditions for several studies.
Table 3: Comparison in SEA percentages variations from Quasi-Static to Dynamic for different fibers
orientations for several studies.
Overall, almost every instance from the literature reports a strong to moderate decrease in Energy
Absorption capacity from quasi-static to dynamic loading for fiber-reinforced composite materials
(Table 1).
Concluding results from that variety of studies, although not always consistent in terms of SEA levels
being reached, all concur in finding a strong plug-initiator influence on the crushing performances
that usually translates into a SEA decrease in presence of a plug initiator. Moreover, the SEA level has
been found to drastically decrease with the increase of the plug size (Table 2).
The effect of fibers orientation and stratification lay-up dependence between dynamic and quasi-
static loading are not always consistent in previously conducted studies (Table 3).
Generally, findings in the literature review on the difference between quasi-static and dynamic
loading conditions are not congruent and not always concordant. Incidentally, most studies highlight
the advantages of circular round tubes over other cross sectional shapes as well as the significant
effect of trigger initiation systems and testing configurations (plugs, chamfers, and testing speeds).
One of the main disputed issues lies in the laminate stacking and stratification lay-up.
The work presented in this study therefore aims at comparing the results in Energy Absorption
capability of hybrid composite circular structures with different fibers orientations and of different
natures (UD, woven) using specific boundary conditions and trigger initiation systems in both quasi-
static and dynamic crushing. Only one impact velocity is being tested in the present work.
A variety of fiber/Epoxy tubes were acquired for testing, with different fibers orientations and fibers
types. These fiber-reinforced tubular structures were studied in axial compressive crushing, at two
different rates, namely quasi-static and dynamic. In total, five specimens with different structures
and different compositions were tested in various crushing configurations. Structural and material
basis for the specimens include 12K HR carbon fibers and polymer Epoxy resin. Fibers orientation and
laminate stratification differ from one specimen to another as shown in Figures [6] and [7] and
summarized in Table 4. Provided tubular structures were machined and shaped in tubes of 100 mm
length as pictured in Figure [6]. Medium diameter was set at an average of 50 mm, with inner
diameter varying from one sample to another due to stacking differences. Stratification layout and
tube wall thickness are summarized in Table 4 for each sample.
The Carbon/Epoxy combination was selected with Epoxy resin as a matrix because of its low density
and for its high strength and good mechanical properties with reliable chemical stability, as well as its
worthy performance regarding energy absorption based on the literature review and due to the
aeronautical context.
In order to verify and establish the composition and stratification of the composite tubes specimens,
samples were polished and observed using a high resolution optical microscope. Measurements and
images acquisitions were performed using an Alicona Infinite Focus SL microscope system with a x10
to x50 magnification. Resulting observations are displayed in Figure [7], along with a lay-up
schematization.
For the laminate lay-up schematization presented in Figure [7], the 0° direction of the fibers was
chosen to coincide with the longitudinal axis of the tube and subsequently with the axial crushing
direction. Plies dimensions are given as an averaged best approximation since plies thickness is not
even and regular. That is supposedly the results of the fabrication process.
Table 4 reports the structural specificities and geometrical properties for the five tube specimens.
C: Carbon A: Aramid
Table 4: Tubes specimens stratifications and properties.
Table 4 presents the stratification structure in number of plies. The “individual” thickness of the plies
(or group of plies) is given in Figure [7].
For the five tubular specimens, materials and fibers properties are not well known and identified nor
completely mastered. Preliminary compressive experimental testing shows that fibers mechanical
properties are lower and less resistant than usually encountered in recent composite materials.
In order to better estimate the mechanical properties of the materials used to manufacture these
samples, and to correlate the theoretical and experimental stiffness, Classical Lamination Theory was
used. Taking a failure criterion expressed in fibers compression strain, chosen at failure = -0.0125 for
all the plies, an estimated stress failure value was calculated for each sample, for the first ply
reaching failure. For the woven plies of the structures, mechanical properties were also calculated
using the Classical Lamination Theory but the related plies were approximated as two superposed
oriented unidirectional plies of half the thickness. The mechanical stiffness properties used for the
theoretical calculations with the Classical Lamination Theory are reported in Table 5, where El is the
longitudinal modulus, Et the transverse modulus, Glt the shear modulus and lt Poisson’s ratio.
Calculated theoretical values are reported in Table 6.
Table 5: Mechanical properties used for the Classical Lamination Theory calculations.
Experimental Theoretical
Stiffness Compressive Stiffness Compressive
(MPa) strength (MPa) (MPa) strength (MPa)
+ 2800 + 100
Tube 1 43 700 - 5200 -350 - 139 54 200 -650
+ 4900 + 65
Tube 2 54 300 - 3100 -340 -102 52 100 -652
+ 8700 + 47
Tube 3 30 200 - 8600 -170 - 92 22 800 -265
+ 6800 + 44
Tube 4 27 300 - 6000 -180 - 59 24 500 -259
+ 3600 + 29
Tube 5 24 700 - 5400 -200 - 41 23 600 -250
When comparing values from the theoretical calculations to the experimental data (Table 6), the
latter reflects lower values in stress failure, although it can be noted that they remain in the same
order of magnitude. Moreover, the experimental compressive strength values that are reported in
Table 6 are rather related to a failure in crushing mode than pure compression, which can further
explain the observed differences.
Results reported in Table 6 show that the compression stiffness is lower than anticipated for 3 of the
5 CFRP specimens, especially for specimens 1 and 3. Compression failure is also lower than
estimated, especially for specimens 1 and 2 that incidentally mainly present 0°-oriented fibers.
This supports the hypothesis of poor fibers’ properties used to manufacture the tubular specimens,
and it can also be explained by the high porosity inherent to many samples, as observed on the
microscopic images (Figure [7]). This is also most obvious for tube specimen 1, for which braided
thread remnants from the manufacturing process are visible too. In addition, experimental
compressive failure values resulting from the performed compression tests could rather be affected
by a bearing phenomenon under the tip of the plies and resulting from the crushing nature of the
solicitation than related to a pure compressive mode, hence the observed discrepancy in
compression strength failure, which thereafter seems more rational.
Finally, when looking at the density values from Table 4, it can be pointed out that they are relatively
low for some samples (sample 1 especially, and to a lesser extent, sample 3). This is allegedly strongly
related to the high porosity observed in the samples (Figure [7]) and also lower fiber density (or
fraction volume vf) in some areas of the samples.
With the objective of establishing a comparison, both quasi-static and dynamic testing were carried
out for the purpose of this study.
Quasi-static crushing tests were carried out using a 250 kN Schenck hydraulic testing machine in
compression testing mode, through a constant loading speed of 0.2 mm.s-1 (Figure [8]). To account
for repeatability, tube specimens were tested at least 3 to 5 times on average for each sample and
each configuration. More details on experimental testing conditions and results are given in the first
part (cf. Chapter I, Energy Absorption Capacity of Composite Thin-Wall Circular Tubes under Axial
Crushing with Different Trigger Initiations or the corresponding article [50]).
Dynamic shock testing trials were carried out using a drop tower (Picture in Figure [9], Figure [10]).
Load and displacement were acquired during testing and trials were monitored using rapid imaging
cameras. To monitor the trials and verify that the velocity did not intensely decrease during mass
dropping due to friction or out of alignment fall, a high-speed camera FastCam Photron SA5 was
employed to record trials pictures at high speed with 20 000 FPS.
Figure [10]: Dynamic shock testing system configuration (a), (b), (c).
In this experimental set-up, a mobile mass is dropped from a predefined height on a stacking of
several components. From top to bottom, they consist of the studied composite tube sample,
encased on the top end and positioned on one of either three different parts (Picture [9], Figures
[10], [11]), setting the boundary conditions for the trials. Three load sensors (HBM C6A - 200 kN) are
then equidistantly positioned, in a 120° triangle shape between two steel reinforced plates. The
sensors are fixed between the plates using through screws and are slightly prestressed. Finally a
Nomex® / or aluminum honeycomb absorber brick is placed underneath the assembly in order to
dissipate any residual energy that would not have been absorbed by the composite tubular structure
during crushing. Adjustment wedges were used as drop-mass stops to prevent tubes specimens from
totally collapsing and preserve the integrity of the assembly underneath.
Tubes specimens 1 to 5 were tested in 3 main configurations, (a) (b) and (c) in Figure [11] with a
focus on tube 1 and 2 in configuration (b) and (c), which produced the best results in quasi-static
crushing. Tube specimen 3 was only tested in configuration (a) and (b) and tubes specimens 4 and 5
were only tested in configuration (a), as a consequence of the available quantities of materials. To
account for the repeatability, tests were performed 1 to 3 times depending on the configuration and
materials (number of tube specimens) availability.
For the Inner Conic configuration (c), forcing a compressive hoop strain of – 0.015, the diameter of
the base of the conic-shaped part (d ext) was made to match the Exterior Diameter (Ext. Diam.) of the
tube. Specifications are reported in Table 7. The same conic part was used for tube specimens 1 and
2, as they have similar diametric dimensions (especially the exterior diameter).
Table 7: Inner conic parts specifications and dimensions for configuration (c).
As a reminder, tube specimens 3 to 5 have not been tested in this particular configuration.
The beginning of the test is defined at t0 = timpact, at which time the mobile mass reaches the sample
and impacts the upper part of the experimental configuration clamping the tubes.
Raw data acquired from the 3 HBM C6A (200 kN) sensors is summed and converted to kN and then
filtered in order to be able to compare dynamic results with quasi-static results. As an example,
Figure [12] presents the stress and stress-filtered (*) data for tube specimen 1 in Inner Crushing
configuration (b).
In order to have a better reading of the signal, the force and stress signals resulting from dynamic
trials were filtered using a time-dependent filter as described below (Eq. 1):
∗
𝑥 𝑛+1 = 𝑥 𝑛+1 + (1 − ) 𝑥 𝑛∗ Eq. (1)
and = 2 . t . f0 where t is the time interval between 𝑥 𝑛+1 and 𝑥 𝑛 and f0 a cutoff frequency,
taken at 2 kHz.
This cutoff value was chosen in adequacy with the resonance frequency of the sensors (4.5 kHz) and
in order to obtain trend curves more easily comparable to quasi-static curves results. Besides
providing a clearer signal and offering a better interpretation, dynamic signals filtering helps with the
comparison between quasi-static and dynamic testing.
One drawback from filtering, however, is that the filtered signal is heavily dependent on the filtering
parameters. The overshoot factor, which may be defined by the crushing stress initial peak maximum
value divided by the density of the crushed material (Eq. 2), is a useful indicator chosen to represent
and compare the overflow of energy for each crushed sample.
𝑚𝑎𝑥
Overshoot = Eq. (2)
As the overshoot factor is uniform to the SEA (kJ.kg-1) a direct comparison between these two values
is usually possible during a quasi-static approach. In dynamics however, signal filtering strongly
interferes with the initial peak load, and therefore any attempt to properly evaluate the overshoot
value may be improper and inappropriate. The filtering influence on the overshoot is clearly visible
when referring to Figure [12], where both filtered and unfiltered signals are overlaid.
One means to characterize and compare the absorbing capability of materials is through the Specific
Energy Absorption (SEA), also referred to as Specific Sustained Crush Stress. The SEA value is given by
the following equation (Eq. 3):
EA 1 U
SEA = = ∫ (uz )duz Eq. (3)
m . u 0
where:
EA is the energy absorbed,
m is the mass of the crushed mater,
is the density of the material,
u is the crushing distance,
is the compression stress.
A very reliable approximation of the SEA value can be obtained "instantaneously" using the stabilized
crushing stress cr (in MPa) divided by the density of the crushed material (Eq. 4).
𝑐𝑟
SEA → Eq. (4)
→ 𝑐𝑜𝑛𝑠𝑡
The SEA consideration as a mean to evaluate the energy dissipated presents the tremendous
advantage of easily and directly comparing different materials’ energy absorption capacity as it is
directly linked to their crushing performance and takes into account their inherent density.
During such dynamic crushing tests, i.e. using a mobile drop mass, the speed of that mobile mass
decreases as it crushes the impacted sample, from contact at timpact to the end of the test, as can be
observed in Figure [13], displaying the crush load (blue) registered by the sensors during the crushing
of tube specimen 1 in configuration (b), with the corresponding speed (red) and displacement (grey).
The load value (blue) is resulting from the summation of the data acquired by the three load sensors.
The speed of the drop mass (red) decreases linearly as the displacement (length of the tube being
crushed) increases progressively (grey). They both end their progression when the mobile drop mass
is stopped by the lateral wedges. The speed presented in Figure [13] results from a first integration of
the motion equation of the drop mass, and the displacement from a second integration.
Figure [13]: Speed, Displacement and Load evolution over time for tube specimen 1 under crushing.
Similar to what is observed in quasi-static crushing tests, the load (or stress) signal in dynamic
crushing also decomposes in 3 stages: a loading phase ending by the main rupture of the structure
and leading to an initial peak load, a stabilization phase and a stable crushing phase, as previously
reported in many studies [8,10] and observed in Figures [12], [13] or [15]. If crushing is prolonged
long enough, a last phase known as densification or compaction of debris may occur, matching an
increase of the load (or stress) at the end of the curve. More oscillations are observed in dynamic
loading than in quasi-static’s however (Figures [15], [16] and [18]) because of a shock wave effect,
generating rebounds in the impacted structure as well as back and forth waves within the drop test
system. Also the sensors’ position, slightly distant from the impact area, marginally affects the signal.
Pictures shown in Figure [14] display snapshots of crushed CFRP tube samples 1 to 4 in free Crushing
configuration (configuration (a)) obtained via rapid-imaging cameras. Brittle fracture and debris
generation are clearly visible for tube samples 1, 2 and 3, whose reinforcement is constituted of pure
carbon fibers, unlike what can be observed for tube samples 4 (and 5), where the aramid cover acts
as a girdle, drawing the shattered composite fronts, resulting in a close folding and wrapping around
itself. This aramid net therefore helps with debris containment and refrains large outer spreading.
Tube sample 5, although not displayed in this figure, behaves very similarly to tube sample 4.
Figure [15] depicts a typical Stress-Displacement curve for a dynamic crushing test, using tube
specimen 1 in configuration (b) as an example. The stress value results from the summation of the
data acquired by the three load sensors evenly positioned beneath the tested structure (Figure [10]).
Generally, a relatively good repeatability can be noted in dynamic testing from one trial to another,
especially considering the dynamic and composite material factors, as illustrated in Figure [15], with
a dispersion range of -24% and +35% compared to an averaged value. The dispersion area on this
graph is obtained using a superimposition of the lowest and highest curves results for this trial.
Overall, less than 1% of the total of tested samples has been discarded by being deemed aberrant.
Figure [15]: Stress-Displacement curve and dispersion for tube specimen 1 under crushing.
This section discusses the stress over density (/) curves for different tubes samples and compares
different testing configurations in quasi-static and dynamic loading.
Composite crushing in dynamic loading significantly reduced the stress generated and thus the SEA
values, compared to quasi-static crushing of the same materials. Examples of dynamic and quasi-
static crushing differences are displayed in Figures [16] and [18] for tubes samples 1 and 2
respectively, in configuration (b) and (c).
Figure [16]: Stress/density over Displacement curves for tube sample 1 in 2 configurations for Quasi-
Static (QS) and Dynamic (DYN) testing.
In case of quasi-static inner-crushing confinement (b), a slight but still significant increase of the
curve can be noted towards the end, starting at 40 mm. Allegedly, this rise should even be starting
sooner, at about 23 mm, distance corresponding to half the tube’s interior diameter, when the wall’s
bottom end meets at the center of the tube (Figure [17], (b.i)).
This rise is not present for tube specimen 2, which has no woven ply covering its interior wall.
Without this woven inner cover, it is suspected that the tube’s interior walls are less guided and
undergo more severe and intense brittle fracture before meeting and densifying at the center of the
tube, as compared to tube 1. In that situation, they conceivably lost most of their rigidity and post-
testing observations showed that debris were more pushed upwards inside the tube’s structure than
radially inward (Figure [17], (b.ii)). In this instance, since the tube’s interior walls don’t meet, no
increase of the stress is observed on the curve, contrary to situation (b.i).
An increase of the stress at the end could be beneficial and valuable for a surge in the SEA value: the
densification of partially crushed material inside the tubular structure stabilizes the crushing process,
resulting in an increase of the mean crushing stress and therefore the SEA.
Figure [17]: Assumed response and path formed by the tube’s wall under different boundary
conditions: (b) Inner Crushing (i) and (ii), and (c) Inner Conic Crushing (ii)
Figure [18]: Stress/density over Displacement curves for tube sample 2 in 2 configurations for Quasi-
Static (QS) and Dynamic (DYN) testing.
For quasi-static configuration with a conic initiation (c), compressive hoop stress is first generated, as
the inner conic-shaped part constrict the extremity of the tubular structure. Therefore the axial
crushing force in the vertical direction takes longer to apply as the exterior diameter of the tubular
structure slides along the inner conic slope, before being axially loaded. This setback is incidental to
the height of the conic shape (Figure [20]). This can be seen in Figures [16], [18] and [19] for
configurations (c) at the very beginning of the curve between 0 and 3 mm displacement.
Due to the important change in velocity magnitude and the dynamic shock resulting from the impact,
this delay is not as clearly observed in dynamic trials (Figure [16] and [18]). However a focus on the
beginning of the curves (Figure [19]) tends to disclose this progressive and gradual stabilization phase
for dynamic cases, mixed up with the oscillations of the signal, especially visible for tube sample 2.
For the dynamic tests, this transition phase is almost identical for Inner Crushing and Inner Conic
Crushing (mainly due to friction). This observation may be considered with caution as the filtering is
bound to have an impact on the shape of the final curve. However, the trend is similar when the
cutoff frequency of the filtering is increased, confirming this assumption.
This section relates to the crushing stress of the different tube specimens in the three tested
configurations. It aims at explaining the overall crushing behavior through the yield of the structure,
given by the crushing peak, and through the mean crushing stress value, corresponding to the
behavior of the specimen in crushing.
Figure [21] presents the crushing peak value in both quasi-static and dynamic modes for all tubes
samples in all configurations, with the theoretical compressive strength yielding limit (as calculated in
Table 6) displayed in the background. Both quasi-static and dynamic crushing peaks in this graph are
expressed in stress level (MPa) for a straightforward comparison with the compressive stress (max).
According to this figure, configuration (b) (Inner Crushing) is the configuration for which tubes
samples are closest to reach their theoretical yielding limit (as can be seen for tubes samples 2, 3, 4
and 5, but except tube sample 1) and in a lesser extent, configuration (a) (free Crushing) for tubes
sample 3, 4 and 5. In any case, tube sample 1 does not appear to render its full potential, as it stands
at about 50% below its theoretical compressive strength. Usually, this initial peak is detrimental, as it
introduces a substantial stress drop that converts in a large energy drop.
When comparing the mean crushing stress for all tubes samples in all configurations (Figure [22]), the
dynamic effect is clearly visible for tube sample 1, whereas it globally seems less impacting on other
samples. It is strongly supposed that as there are more 0°-oriented fibers present in those samples,
the overall rupture is mainly due to kink-bands effects, leading to global structure failure. Dynamic
effects must be more significant and have a bigger impact on that kind of damage in crushing than on
sole matrix rupture in compression. Furthermore, the compressive stress crushing for dynamic loading
appears to be higher for samples containing 90°-oriented fibers, than those with mainly 0°-oriented
fibers, confirming the low dependence on rate for matrix rupture.
Tubes specimens 4 and 5, which are partially and totally covered with an aramid lay-up on the outer
surface, perform relatively fairly, although the aramid fibers’ rigidity is lesser than the carbon fibers’.
Their core carbon structure and stratification is very similar to tube specimen 3, which explains their
global similar behavior. The major difference therefore lies in the thickness of the tube’s wall, about
half thick for tube specimen 3 compared to other tubes. Tube samples 4 and 5 have not been tested
in dynamic crushing for configuration (b) and (c), nor has tube sample 3 in configuration (c) due to a
lack of samples at the time.
Crushing stress values clearly show that tube specimen 1 performs very poorly in dynamic crushing,
compared to other tubes specimens and also to its own performance in quasi-static crushing.
Overall, in quasi-static loading, the crushing stress always averages around 150 MPa, which
approximately corresponds to the matrix rupture stress in compression. In dynamic loading, only
tube sample 2 in configuration (a) (free-Crushing) reached 150 MPa. It can therefore be supposed
that other configurations ((b) and (c)) have created more brutal ruptures, most particularly due to
global and local buckling, since a pure compressive configuration should produce a 150 MPa
compressive stress. It can therefore be inferred that both the Inner Crushing and Inner Conic
Crushing configurations destabilize the crushing in dynamic loading.
The following discussion, namely Figures [23], [24] and [25] displaying the SEA evolution with the
vertical displacement, presents results based on actual SEA values, that is to say calculated using the
formula given by Equation (3) and using non-filtered stress values. As the SEA equation is based on
an integration, it already acts as a filter, hence the smoothing and difference with the stress/density
curves.
The evolution of the SEA for tube specimens 1 to 5 in dynamic configuration (a) is presented in Figure
[23]. From this figure, it is obvious that tube specimen 2 surpasses all other specimens in terms of
SEA capacity, despite the fact that its peak load is initially identical to tube specimen 1. That latter
specimen performs poorly under dynamic loading, collapsing very easily and quickly.
Compared to quasi-statics, in dynamic crushing, the stabilization phase takes longer: while for quasi-
static loading 5 to 10 mm are required, in dynamic cases, it takes 15 to 20 mm approximately. This, as
many aspects observed in dynamic mode, can be partially explained by the numerous waves
generated during dynamic impact.
Figure [24]: SEA-displacement curve in Quasi-Static (QS) and Dynamic (DYN) loading for tubes
samples 1 to 3 in free Crushing configuration (a).
For the free Crushing (a) and Inner Crushing (b) configurations, the initial peak is generally higher in
dynamics than in quasi-statics (Figure [24] and [25]). This is particularly true for both samples with
mainly 0°-oriented fibers (samples 1 and 2), which coincidently have the same peak profile, whether
in configuration (a) or (b).
For the Inner Crushing configuration (configuration (b), Figure [25]), it seems that the evolution of
the SEA over the displacement tends to decrease at the end of the curves for all three dynamic
crushing and especially for samples 2 and 3. This does not appear to have any particular reason, a
larger number of tests would be required to confirm or correct that tendency.
3.5. Overshoot
Figure [26] attempts a comparison of the overshoot values between quasi-static and dynamic
loadings for the three tested configurations ((a), (b) and (c)).
Figure [26]: Overshoot values in Quasi-Static and Dynamic loading for tubes samples 1 to 3 in
crushing configuration (a), (b) and (c).
It is, however, also needed to keep in mind that the overshoot values presented for the dynamic
testing are originating from a filtered signal, which somewhat affect their factual values.
Figure [27] summarizes the SEA values resulting from experimental testing both in quasi-static and
dynamic loading conditions for all three configurations and tube samples 1 to 5 (when available).
Figure [27]: Comparison of SEA values for tubes 1 to 5 in 3 crushing configurations for Quasi-Static
(QS) and Dynamic (DYN) testing.
Generally, SEA values are lower in dynamic crushing than in quasi-static crushing. This finding has
also been largely reported in previous studies [11,22,26,33,38,41,49].
Overall, based on values reported in Figure [27], tube specimen 2 performs better in dynamic loading
than other tube specimens for all configurations, only matched by tube specimen 3 in inner-crushing
configuration (b) in dynamic loading. This result is strongly suspected to be the consequence of the
90°-oriented fibers present in the core structure of the tubes’ walls, that help stabilizing the structure
and therefore favor and enhance the crushing.
More specifically, tube specimen 2 presents similar results in dynamic and quasi-static loadings for
configuration (a) (free Crushing), but shows a relative to moderate decrease in SEA values for
configuration (b) and (c): 28% and 42% decrease respectively.
Problematic issues observed for tube specimen 1 in dynamic loading may come from the too large
thickness created by the consecutive 0° plies at the center (Figure [28]), whereas for tube specimen
2, interleaved 90° plies must stabilize the structure and therefore smooth the dynamic rupture.
Figure [28]: Highlight of the core structure of tube specimen 1 (left) and 2 (right).
Tube specimens 3 (mostly 90°-oriented), 4 and 5 (identical 90°-oriented basis with an additional
aramid cover) present almost identical results for dynamic and quasi-static loading in free Crushing
configuration (configuration (a)). It is only slightly higher for the quasi-static rate. This trend has to be
confirmed however, as several data points are missing.
However, tube specimen 3 performs better in configuration (b) (Inner Crushing) for dynamic loading.
This can be considered two ways: either a better efficiency in term of energy absorption in dynamic
loading in this configuration, or an almost identical effectiveness in quasi-static and dynamic loading
when taking into account the margin of error and high dispersion in quasi-static loading. Further
testing on a larger number of samples would be required to decisively rule on that trial.
With this in consideration, it can presently be stated that in dynamic loading, tube specimen 3 in
Inner Crushing configuration (configuration (b)) achieves similar results as tube specimen 2 in
dynamic loading. Tube 3 in configuration (b) returns higher SEA values than tube 2 in configuration
(b) and (c) but they are slightly lower than tube 2 in configuration (a).
Generally, tube specimens with 90°-oriented fibers seem to perform better in dynamic loading.
However, overall results and observations lead to the conclusion that in order to have a worthy and
strong SEA both in quasi-static and dynamic rates, it is required to have a stratification with 0°-
oriented plies stabilized by a woven fabric or by few 90°-oriented plies. It also appears that in
dynamic loading it is essential to avoid having too many consecutive 0°-oriented plies in order to
prevent catastrophic failures. In dynamic conditions, the presence of 90°-oriented plies that stabilize
the structure appears to be a major requirement.
II.4. Conclusions:
Dynamic and quasi-static crushing tests were achieved for different circular hybrid composite tubes
samples in various configurations using several trigger initiations. SEA values up to 90 kJ.kg-1 were
obtained, achieving better than most instances from the literature, averaging around 50 kJ.kg-1 for
Carbon/Epoxy laminates. The impact velocity for the dynamic testing was set at 5 m.s-1. Specimens
with 90°-oriented fibers perpendicularly oriented with the direction of compression reached the
highest SEA values while those with mainly 0°-oriented fibers coincidental with the direction of
compression performed unwell. It has consequently been established that in dynamic loading, the
presence of 90°-oriented fibers stabilized by woven plies greatly enhances the energy dissipation
capability of the structure. This was not the case in quasi-static loading, where 0°-oriented specimens
achieved best SEA values and met highest expectation in terms of energy dissipation. Moreover,
unidirectional laminates oriented at 0° performed 50% lesser between quasi-static and dynamic
rates. This has been explained by the too large thickness made by the consecutive 0° plies, yielding
by kink-band effect. Incidentally, an inner constrained containment is more effective in most cases,
reducing the initial peak load without drastically reducing the SEA value. Inner Conic Crushing,
however, although reducing the initial peak load, also reduces the overall SEA value.
Perspective:
In order to complete this study and provide complementary understanding regarding the dynamic
effect in composite tubes crushing, it could be interesting to vary the impact speed for the dynamic
crushing tests, and especially increase it to high-speed velocity, since all trials in the present study
were performed under the same impact velocity and the literature, for the most part, reports a
significant effect of the strain rate.
In addition, a study of the size of the debris and a comparison of their form and size between quasi-
static and dynamic testing might provide more insights regarding the fracture and damage mode that
undergo the tubular structure during these two load types.
Investigation regarding the D/t ratio (Diameter over tallness) of the tubular structures as well as the
wall thickness impact on the crushing behavior and crush strength might be helpful. Similarly, a more
elaborate study on the cone angle, and sloping inclination of the Inner Conic configuration might be
needed, to best adjust that parameter. Finally, testing of chamfered or notched tubular structure, in
combination with the proposed boundary conditions, might help further reduce the peak load and
enhance the SEA capacity at the same time.
Financial support for this study was provided thanks to FEDER funds through the joint project
SKYSEAT co-funded by ARTEC Aerospace Company, the Occitanie – Pyrénées-Méditerranée Region
(ex-Midi-Pyrénées Region) and the ISAE-SUPAERO.
References:
[1] [HEIMBS 2010] HEIMBS, S., STROBL, F., MIDDENDORF, P., GUIMARD, J.M. Composite Crash Absorber For Aircraft
Fuselage Applications, WIT Transactions on The Built Environment, Volume 113, 2010, Page 3-14, ISSN 1743-3509,
DOI 10.2495/SU100011.
[2] [VAN BEEK 1998] VAN BEEK, J.A.B. Dynamic Seat Testing and Head Injury Criteria - An Aircraft Manufacturer’s
Experience and a View on the Future, In: Proceedings of the 21st Congress of International Council of the Aeronautical
Sciences, 13-18 September 1998, Melbourne, Australia.
[3] [OLSCHINKA 2006] OLSCHINKA, C., SCHUMACHER, A., RIEDEL, D., Dynamic Simulation of Flight Passenger Seats. In:
Proceedings 5. LS-DYNA Anwenderforum, 12-13 October 2006, Ulm, Germany.
[4] [FARLEY 1983] FARLEY, G.L. Energy Absorption of Composite Materials, Journal of Composite Materials, Volume 17,
No. 3, May 1983, Pages 267-279, ISSN 0021-9983, DOI 10.1177/002199838301700307.
[5] [THORNTON 1990] THORNTON, P.H. The Crush Behavior of Pultruded Tubes at High Strain Rates, Journal of
Composite Materials, Volume 24, Issue 6, 1 June 1990, Pages 594-615, ISSN 0021-9983, DOI
10.1177/002199839002400602.
[6] [HAMADA 1995] HAMADA, H., RAMAKRISHNA, S., SATOH, H. Crushing mechanism of carbon fibre/PEEK composite
tubes, Composites, Volume 26, Issue 11, 1995, Pages 749-755, ISSN 0010-4361, DOI 10.1016/0010-4361(95)98195-Q.
[7] [MAMALIS 1997] MAMALIS, A.G., ROBINSON, M., MANOLAKOS, D.E., DEMOSTHENOUS, G.A., IOANNIDIS, M.B.,
CARRUTHERS, J. Crashworthy capability of composite material structures, Composite Structures, Volume 37, Issue 2,
February 1997, Pages 109-134, ISSN 0263-8223, DOI 10.1016/S0263-8223(97)80005-0.
[8] [HULL 1991] HULL, D. A Unified Approach to Progressive Crushing of Fibre-Reinforced Composite Tubes, Composites
Science and Technology, Volume 40, Issue 4, 1991, Pages 377-421, ISSN 0266-3538, DOI 10.1016/0266-
3538(91)90031-J.
[9] [PINHO 2004] PINHO, S.T., CAMANHO, P.P., DE MOURA, M.F. Numerical simulation of the crushing process of
composite materials, International Journal of Crashworthiness, Volume 9, Issue 3, 2004, Pages 263-276, ISSN 1358-
8265, DOI: 10.1533/ijcr.2004.0287.
[10] [SUPIAN 2018] SUPIAN, A.B.M., SAPUAN, S.M., ZUHRI, M.Y.M., ZAINUDIN, E.S., YA, H.H. Hybrid reinforced thermoset
polymer composite in energy absorption tube application: A review, Defence Technology, Volume 14, Issue 4, 2018,
Pages 291-305, ISSN 2214-9147, DOI 10.1016/j.dt.2018.04.004.
[11] [BRIGHTON 2009] BRIGHTON, A., FORREST, M., STARBUCK, M., ERDMAN, D., FOX, B. Strain Rate Effects on the Energy
Absorption of Rapidly Manufactured Composite Tubes, Journal of Composite Materials, Volume 43, No. 20, September
2009, Pages 2183-2200, ISSN 0021-9983, DOI 10.1177/0021998309344646.
[12] [ISRAR AHMAD 2014] ISRAR AHMAD, H.A.B., RIVALLANT, S., BOUVET, C., BARRAU, J.-J. Finite element simulation of
0°/90° CFRP laminated plates subjected to crushing using a free-face-crushing concept, Composites Part A: Applied
Science and Manufacturing, Volume 62, July 2014, Pages 16-25, ISSN 1359-835X, DOI
10.1016/j.compositesa.2014.03.014.
[13] [EVANS 1978] EVANS, A.G., ADLER, W.F. Kinking as a mode of structural degradation in carbon fiber composites, Acta
Metallurgica, Volume 26, Issue 5, 1978, Pages 725-738, ISSN 0001-6160, DOI 10.1016/0001-6160(78)90023-8.
[14] [GUTKIN 2011] GUTKIN, R., PINHO, S.T., ROBINSON, P., CURTIS, P.T. A finite fracture mechanics formulation to predict
fibre kinking and splitting in CFRP under combined longitudinal compression and in-plane shear, Mechanics of
Materials, Volume 43, Issue 11, November 2011, Pages 730-739, ISSN 0167-6636, DOI
10.1016/j.mechmat.2011.08.002.
[15] [CHIU 2015] CHIU, L.N.S., FALZON, B.G., RUAN, D., XU, S., THOMSON, R.S., CHEN, B., YAN, W. Crush responses of
composite cylinder under quasi-static and dynamic loading, Composite Structures, Volume 131, 2015, Pages 90-98,
ISSN 0263-8223, DOI 10.1016/j.compstruct.2015.04.057.
Keywords: Composite tubes crushing, Braided Structures, Energy Absorption, Numerical Model, FEA
Abstract:
Various compression tests on CFRP tubular structures were performed and instrumented in order to
understand the damage mechanisms involved. Testing has been carried out under quasi-static and
dynamic loading and using several boundary conditions. The different results of compression and
crushing behavior are then used in order to build and improve a finite element calculation model
allowing to simulate the response of composite structures of different natures subjected to crash by
integrating the geometry and the composition of the structure. The model would ultimately be used
to evaluate the energy that can be dissipated by composite tubular structures under crushing.
The objective of this research work is thus to use the experimental results of composite structures of
different compositions and materials to provide a finite element model representing the behavior of
UD and woven composite structures submitted to compression until damage and fracture of the
structure. Eventually, the aim is to optimize the lay-up configuration and plies orientations, as well as
materials type and shape of the structure, to improve the SEA capacity of the composite structure.
This work presents a first step towards that end.
III.1. Introduction:
Due to the numerous operable and customizable parameters (shape, thickness, materials, initiation
triggers) and possible configurations (stacking order, orientation, plain or braided, bi-axially or tri-
axially), numerical crushing simulation of composite structures has become an essential element in
the predictive capability of composite absorbing structures and structural crash optimization [1-9]
Composite tubular structures are deemed excellent energy absorbers due to the mechanical
properties of the material as well as the structural shape and geometry of the structure and have
therefore been widely investigated. [SUPIAN 2018][10] presents a complete review of hybrid reinforced
polymer composite tubes for energy absorption applications, concluding that the effect of
hybridization in composite tubes exhibited a desirable effect according to the type of failure modes
involved, hence making them very suitable candidates as energy absorbing structural parts in various
applications. [MEIDELL 2009][11] presents a study based on computer aided material selection for
circular tubes designed to resist axial crushing. Almost all studies converge regarding the description
of the damage mechanisms and global failure mode observed for tubular structures under axial
crushing (Figures [1] and [2]).
Stable progressive crushing (4) and high levels of Specific Energy Absorption (SEA) are nowadays
extensively attainable using braided or hybrid composite tubes submitted to crushing [12-19]. The
ensuing crush morphology of such composite structures under quasi-static or dynamic crushing can
generally be categorized into one of three modes [11,19,20]: (I) fragmentation mode – rupture by
brittle fracture and sequential damage of material at the micro-structural level, (II) splaying mode –
rupture by delamination and formation of continuous fronds in combination with transverse tearing
and (III) progressive folding mode – yielding by successive and progressive buckling (Figures [2]). A
final and additional phenomenon which is majorly observed resides in debris creation and
accumulation under the crushing fronts (IV). These uneven fronts generate different sizes of
fragments or debris that accumulate underneath the structure and between the crushing plane and
create wedges that are strongly suspected to increase the delamination (or splaying) occurrence.
Figure [2]: Major damage during the crushing of a composite tube (adapted from [PINHO 2004][2], after
[MAMALIS 1998][21])
Figure [3]: Schematic representation of the ‘shear’ kink formation chronology (a),
schematic of the transition zone between fully-kinked and unkinked zone (b) [EVANS 1978][22],
imaging of a typical kink-band geometry (c) [GUTKIN 2011][23].
Figure [3] displays a schematization of the kink-band phenomenon (a) and (b) and imaging
illustration (c). The kink-band formation is described as a band of matter with its edges lying at angles
+ 0 and . It is induced by the shear stress A due to combination of (11, 12) in a coordinate
system aligned with the rotated fibers. The angle + 0 corresponds to the angle of rotation of the
fibers within the kink-band (with 0 is the initial misalignment and the additional rotation
introduced by the compressive loading) and to the kink-band propagation [23].
The modelization of such phenomena still remains largely problematic. Crushing phenomenon in
composite materials hence involves failure modes which take place at different length scales [24].
Limitations in numerical models can therefore be the result of the choice of modeling scales selected
to predict those crushing mechanisms.
Many of the previously developed models [3,6-8,19,25-27] are based on global test characterizations
(that is to say at a macro-scale level) making the model strongly dependent on the laminate behavior
and crushing response. Even though this methodology is found to make modelization simpler and
largely decreases computational time, it is unlikely to fully represent damage mechanisms that take
place at a sub-ply scale [20] or fully represent the kink-band phenomenon of the fibers under
compression.
Macro-scale models display advantageous computational time, but have the disadvantage of hardly
taking into account material disparity, lay-up variation or woven influence. All of these parameters
expand and enhance the robustness and precision capability of the model.
Behaviors described above are highly and ultimately dependent on many structural parameters such
as geometry, laminate sequences and ply orientations, materials and mechanical properties, contact
conditions and friction [6,17,20,24,29]. When taking that wide range of parameters and phenomena
into account, the capability of existing numerical models to completely and truly describe the
progression of complete crushing events from initiation, stabilization and progressive collapse up to
the point of final failure is still limited.
The model presented and developed in this study takes place at the meso-scale level. The originality
and beneficial attributes of this model lie in the separation of the matrix, fibers, braid or mesh types
and inter-plies interface. It also takes into account the global behavior of the matrix, the fibers, the
woven structure and the interface through behaviors which are independently implemented and
controlled, using own properly adapted behavior laws for each kind of constituents.
Implementation of this numerical work and model development was based on the experimental
results of the crushing of a CFRP tubular structure under quasi-static and dynamic conditions.
This numerical work represents a first step in composite crushing modelization, that was
unfortunately only partially conclusive, since it did not utterly produce the expected results and the
model did not fully behave according to the intended expectations.
UD-Woven CFRP circular tube specimens have been tested in crushing mode in quasi-static and
dynamic loading (Figure [4]). The tube specimen performing best in quasi-static and with an even and
equilibrate structure has been chosen as a guiding thread for the building and developing of the
numerical model (Figure [5]). This specimen which offers a combination of inner UD plies and outer
woven fabric on both sides was found to be an optimal option to fully benefit from inherent
advantages of both types of structure (Figure [5]). Additionally it has previously been established that
in quasi-static loading, a 0° unidirectional core structure stabilized by external woven plies to avoid
buckling and large kinking produced the best results in term of crashworthiness.
In Figure [5] are displayed, from left to right, a picture of an actual tube sample (tube specimen 1)
used during experimental testing, a microscopic image of that tube’s structure from above and its
corresponding schematization.
Figure [5]: Selected CFRP tube sample for the modelization, from left to right: actual tube sample and
dimensions (left), microscopic imaging of the sample (center) and schematization (right)
Geometrical and mechanical properties for this tubular structure were measured, experimentally
assessed in compressive testing and theoretically evaluated using the Classical Lamination Theory.
The corresponding data values are reported in Table 1 and Table 2.
More details on the experimental testing conditions are given in the first part (cf. Chapter I, Energy
Absorption Capacity of Composite Thin-Wall Circular Tubes under Axial Crushing with Different
Trigger Initiations or the corresponding article [30]).
Figure [6] displays the stress-displacement curve (and dispersion) obtained for this specimen in
progressive crushing, under quasi-static loading. It is observed that the dispersion is very low
(averaging -12% and +32% compared to the medium value) for the plateau level, however it is
extremely important and disparate for the initial peak load. The rupture stress value for which the
structure is yielding is the most random measurement coming from the testing data.
Figure [6]: Stress-Displacement curve and dispersion for tube 1 in quasi-static crushing.
A total of three main phases is identified during the crushing process of a CFRP structure: (1) initial
peak load – (2) stabilization – (3) stable crushing (Figure [6]). The first stage corresponds to a loading
phase within the elastic domain, ending by the main rupture of the structure and leading to a
yielding peak. Following the structure global yield, a transition or stabilization phase occurs before a
stable crushing phase settles. This continuous and progressive crushing occurs at a constant level
which defines the Mean Crushing Stress (MCS). The localized fragmentation under the tip of the plies
is pointed as the mechanism leading to the characterization of the ply Mean Crushing Stress value.
Post-testing X-Ray 3D-micro-measurement tomographic imaging was performed (Figure [7]). These
observations were conducted to observe and determine the damage mechanisms involved during
crushing on the inside of the tubes’ walls thickness (Figure [8], [9]). These tomographic images were
achieved using a Micro-Tomography EasyTom 130 machine, manufactured by RX Solutions, France.
Figure [8]: Schematic presentation of damage and Figure [9]: Picture of the crushed sample
Figure [8] represents major occurring damage based on tomographic observations (Figure [7]) and
post-crushing tube samples (Figure [9]). Fragmentation by brittle fracture, splaying (delamination)
leading to the creation of continuous fronts and axial splitting, as well as debris creation are clearly
visible.
These observations and testing data are thereafter used to build and improve a numerical model.
The challenge in developing a robust and reliable predictive numerical model is to take all that
information into account and realistically simulate all those observed damage, namely, brittle
cracking, kink-bands, splaying, fragmented and torn fronds, debris… Needless to say, classical finite
elements are not principally or accurately developed for that kind of modelization issues, as they are
particularly built and most suited for continuous problems. This issue remains largely problematic in
numerical modelizations, hence the advantage of the present developed model, which is both
discrete and continuous as it integrates a semi-continuous approach.
Developing and adding to existing works accomplished by [NAVARRO 2012,2015] [TAWK 2012] [AUBRY 2013]
[PASCAL 2015,2018][31-37], a numerical model was extended for crushing application and simulation.
This model was built to be implemented in Altair HyperWorks Radioss FE software, (Radioss, Altair
HyperWorks solver, version 12.0 (10.0), Altair Engineering Inc., 2013), using an explicit calculation
method. It was initially developed for the modelization of impacts on woven composite sandwich
structure [31,38], and more particularly the oblique impacts on composite helicopter blades [31,32].
This specific composite laminates numerical model presents the particularity of being built on a
matrix – fibers de-cohesion and differentiation. It has first been developed for the specific case of
woven fabric structures modelization, and then extended to both woven and unidirectional
modelization for the need of this study. The woven and unidirectional composite plies are connected
using specific interface elements (or cohesive elements) that are able to take into account the
compressive and bending behavior of the shell fabric elements.
A specific user element has been developed, improved and adapted over the years and many
research works [31,33,35,37] to bring an innovative solution for Finite Element Modeling of crash and
impact of composite laminate structures.
This user element consists of an 8-node-element in which are implemented: 1 plane shell element, 4
truss (or rod) elements, 1 interface element (Figure [10]). Rigid bodies are used to ensure kinematic
coupling between the shell element and the rods (or truss) elements.
This model is fully modular and adaptable. The size of the element as well as the geometry is fully
configurable for any wanted application and structural stratification. Moreover, the size of the
elements is based on the braided structure and it is directly linked to the size of the woven pattern.
This is an original feature and a major advantage of this model, compared to classical FE model which
are mainly nondimensional.
Coupling between nodal efforts of the shell and rods is carried out by rigid bodies: rod efforts at the
virtual nodes at both ends are transmitted and summed to actual real nodes of the shell elements. A
total of 8 integration points are used within the shell elements, for membrane normal stress and
membrane planar shear calculations. The flexion stress is partially resulting from the shell element
but it is also largely conveyed by the offset position of the rods. All the forces and moments resulting
from the calculations are directed and dispatched to the 4 “real nodes” of the shell element.
The shell element is comprised by 4 initial nodes (1-2-3-4) that establish the base of the whole
element (Figure [10], left). These 4 first nodes constituting the perimeter of the shell element are
designed as “real nodes”. The next 4 nodes (5-6-7-8) finishing the parallelepipedic structure are
utility construction nodes that become active only if a top upper layer is added. In that case, they
become boundary nodes for a new shell element therefore constituted (Figure [10], right).
Rod elements are 2 by 2 offseted by a minimal distance (customizable offset), , set at ¼ of the
element thickness h0. To genuinely represent the nature of a braided structure, 2 of the 4 rods are
positioned above the mean plane defined by the shell element while the other 2 are positioned
bellow. To achieve that, “virtual nodes” at both ends of the rods are created. These virtual nodes are
orthogonally positioned on both sides of the shell relative to the real nodes. The force in the rods is
transmitted to the “real nodes” via the “virtual nodes”. Rigid body elements are used to link these
“virtual nodes” to the “real nodes”.
An interface user element is also included within the 8 nodes elements, and is activated if necessary
(Figure [10], center and right).
To fully put this modelization into effect, the nature of actions that have been accomplished was on
two aspects: on the geometry and the behavior of the element. Those two aspects are presented
thereafter.
Figure [11]: Sixteen unitary elements to create each and every kind of 2D braided patterns.
To complete this library, 2 more elements were added to represent the unidirectional (UD) unitary
elements – one for the 0° orientation, the other for the 90° orientation (Figure [12]).
Figure [12]: Two additional unitary elements to create unidirectional patterns (0° and 90°).
The advantages of this model are hereby clearly illustrated: once the behavior of the resin and the
fibers are known and identified, any UD or Woven structure is then theoretically fully modelizable.
As an illustration, Figure [13] displays the basic pattern formed by unitary elements and their
connections to reproduce a taffeta woven fabric corresponding to the upper and bottom layers of
the composite specimen stratification under investigation.
The section of the fibers (rods) needed to calculate the stiffness is given by (Eq. 1):
1
Sf = × Vf h 0 L0 Eq. (1)
4
where Vf is the density ratio of fibers, h0 the initial thickness of the element, and L0 the initial length
of the element (Figure [10]).
The ¼ factor is due to the geometrical aspect of the element and the philosophy behind its
constitution. A first ½ coefficient is introduced since half the fibers tows are in warp direction and
half are in weft direction. A second correction with a ½ coefficient is introduced since fiber-rod
elements are communally shared between two consecutive elements, therefore only half a rod
section is considered for each direction for the woven-type elements, and two halves in the
longitudinal direction for the unidirectional-type elements (Figure [14]).
: Stress
: Strain
rc : Compression stress rupture level
rc : Compression strain rupture level
p : Plastic stress level, corresponding to the MSC plateau
p : Plastic strain limit, corresponding to the beginning of plasticity plateau
Ec : Compression modulus
Ets : Tensile modulus
rt : Tensile stress rupture level
rt : Tensile strain rupture level
eq : Equivalent Von Mises stress
0 : Shear stress rupture level
Rplat : Hardening coefficient
: Hydrostatic pressure coefficient
Er : Resin stiffness modulus
Ef : Fiber stiffness modulus
Vf : Volume fraction of fibers
f : Fiber strain
El : Longitudinal modulus of the element
Et : Transverse modulus of the element
Glt : Shear modulus of the element
lt : Poisson’s ratio of the element
The objective of the numerical work is to represent the brittle behavior in traction (red) and the
plastic crushing phase in compression (blue) (Figure [15](b)). Therefore the current model aims at
modelizing a linear elastic behavior, followed by a brutal rupture in traction and a perfect plasticity
level in compression, so as to represent the crushing plateau, based on the Mean Crushing Stress
(MCS) definition [39,40]. The tensile and compressive stiffness modulus are chosen to be the same,
that is to say, Ets = Ec. The rupture in compression (logically following the elastic behavior and before
the crushing plateau, Figure [15](a)) was not modelized, as further explained thereafter.
Figure [15](a) represents the schematization of the real compressive behavior, based on the
experimental testing results (Figure [6]). However, for the compressive-related behavior of the
model, it was decided to limit the modelization to the compressive stress decrease and the plateau
level and phase out the initial peak load (Figure [15](b)).
That choice was based on experimental observations (Figure [6]), from which resulted that the
rupture level in compressive strength occurred most randomly, returning dispersive and
unpredictable peak stress values. For a purpose of simplicity it was therefore decided not to take the
initial peak into account in a first approach.
That decision was also based on a numerical reason, namely the impossibility to establish
communication between elements when using the current Radioss developed model. That
communication would have been essential in order to “notify” a neighboring element in a vertical
column that the previous below it has already yielded and therefore avoid re-creating the initial
loading peak load after each instance of an element’s deletion.
This communication notion has already been established in other numerical studies. The operating
principle used in the modelization work by [ISRAR AHMAD 2014][20] is illustrated in Figure [16]. Initially,
only the bottom element at the extremity of the ply (1) and in contact with the crushing plane is
considered as a free-face-crushing element. When subjected to compression, this element is crushed
under an appropriate localized behavior law. Once totally damaged, the element is then deleted, and
the next element above it (2) becomes a new free-face-crushing element, with its behavior law
automatically changing from inside ply law to localized crushing law.
In order to take into account the localized fragmentation, [MATZENMILLER 1991][41] introduced the
concept of advancing crashfront. This concept was based on the use of a parameter which reduces
the strength of the elements located in the crushing zone, that is to say only the elements directly in
contact with the crushing surface. Results of simulations based on that concept disclosed that the
deletion of elements led to a series of unrealistic peaks in the load–displacement curve of the
crushing simulation.
To overcome such limitations, works by [RIVALLANT 2013][40] and [ISRAR AHMAD 2014][20] introduced a
similar free-face-crushing concept, combined with a specific behavior law. In order to represent the
localized crushing, (i.e. with a constant mean crushing stress), elements located at the extremity of
plies (that is to say elements in direct contact with the crushing surface by having a free-face in the
direction of the crushing) have a different behavior than other elements, higher in the structure. For
these upper elements, corresponding to inside ply elements, a classical failure criterion is applied.
The behavior law governing the elements is therefore adapted and changed based on the state of the
neighboring elements.
The use of bottom “discrete” elements with subsequent complete removal to represent a continuous
mechanism still leads to limitations in the model, such as generating excessive distortion before
deletion and creating successive peak load after each element deletion.
Such an operating modelization, i.e. based on elements communication, was unfortunately not
supported in Radioss, hence the choice in modelling strategy is not to represent the initial peak load.
Although such a communication between elements is not categorically unachievable in theory, it is
not a standardized option in the commercial version of Radioss Finite Elements software.
Furthermore, this initial peak load, along the inherent stiffness and compressive strength failure
values, can separately be numerically obtained using a distinct classical homogenized model that
limits itself to the first crushing stage. Achieving that would consist in providing a separate specific
numerical model aiming at modeling only the first phase of the crushing process.
Figure [17]: Expected behavior expressed in the Principal axes system, with both tensile (brittle) and
compressive (plastic) behaviors.
Stresses are calculated in the Principal basis with specific behavior laws for woven and unidirectional
elements (refer to the description below) while the damage parameter is updated at each time step.
Shell/resin behavior
The behavior is based on an elastic damageable modelization, using the general following law: (Eq. 2)
In the Principal coordinate system (XI,XII), the behavior law is given by: (Eq. 3)
EI (1 d I ) EI (1 d m )
(
plast
) Eq. (3)
EI (1 d m ) EII (1 d II ) ( X I , X II )
1
with and d m max(d I , d II ) Eq. (4)
1 I II II I
and where EI is the stiffness modulus in direction I, EII the stiffness modulus in direction II, and I II
represents Poisson’s ratio in (XI,XII).
The damage parameter di is updated if necessary (see section below). This update is done
permanently, without possible reversion.
For the Woven elements, no distinction is made for the stiffness modulus Ei in either direction, as the
woven structure has similar properties in warp and weft directions. Furthermore, experimental
testing on woven samples showed a mechanical uniformity in both warp and weft directions.
Damage parameter
Initially, the damage parameter d is set at 0. The following criterion (Eq. 5) is used to update the
damage parameter, from 0 to 1 (more accurately 0 to 0.99 to avoid numerical instability).
In order to express the resin rupture using a maximal Principal stress criterion, and taking into
account the hydrostatic pressure, a Drucker-Prager criterion is used: (Eq. 5)
where 0 is the shear stress rupture level and λ is a material coefficient used to take into account
the effect of the hydrostatic pressure [42]. It can be noted that in pure shear mode, I + II = 0, hence
confirming that 0 is indeed the shear rupture stress.
When the criterion given by Eq. 5 is reached, the damage parameter d is updated to 1 (or 0.99). Then
it is attributed to the corresponding direction I or II, according to the following conditions process:
Plastic behavior
In addition, a perfect plasticity law is used to represent the homogenous crushing of composite
elements, using equivalent Von Mises stress, eq, and a plastic loading function, fplat (Eq. 6).
where P is the plasticity limit, set at P= 250 MPa (according to the MCS) and Rplat an isotropic
hardening term.
Rods/fibers behavior
The behavior law for the fibers is the same in both geometric structural designs. It is given as a brittle
rupture behavior, with a rupture controlled in elastic strain.
where Frod is the force inside the rod elements, Ef the stiffness of the fibers, rod the strain of the
fibers, and Sf the section of the fibers.
The tension rupture of fibers bundle is of brittle nature. The fiber rupture is given by a maximum
tensile strain criterion (Eq. 8):
where c is the maximum strain value of the fibers in tension (positive term), and c the maximum
r r
In order to smooth the brutal breakage of the rods elements leading to a surge in energy release and
avoid numerical instabilities, instead of immediate brutal rupture, the normal force Frod is
progressively decreased using a given characteristic time when the maximum strain criterion is
reached (Eq. 9).
* t* t
Frod Frod
1 exp Eq. (9)
where t* is the time at which the rupture criterion is reached and Frod
*
the actual force at that time.
The characteristic time is a constant time value, fixed at 1 µs, conforming to a compromise
between stability and sensitivity of the numerical results.
Ultimately, the compression failure mechanism generating the kink-band was taken into account
within the MCS expression in the shell element, instead of using the rod elements, mainly due to
stability issues. These stabilization difficulties will be presented and discussed thereafter.
Unfortunately, including parts of the fibers behavior within the shell element – initially representing
the resin alone – has prevented from making the best of this unique model, specifically based on
fiber-matrix decohesion.
For the delamination modelization, each ply is connected to the next using a shell-to-shell interface
user element with a cohesive law, developed and detailed by [NAVARRO 2015][34]. The interface is
considered as a thin ply of matrix, with a small thickness compared to the element’s dimensions. The
constitutive parameters of the developed interface were obtained from inverse identification and
usual tests of interlaminar crack propagation were performed, namely pure mode I Double Cantilever
Beam (DCB) test and pure mode II End-Notched Flexure (ENF) test [43,34].
Element deletion
The removal of the whole element (shell, rods and interface, as defined above in Figure [10]) is
achieved via a maximum negative strain deformation criterion surpassing 85% of the initial length
(L0) of the element in compression. In traction, the element is deleted if all the 4 rods elements are
damaged, after reaching the maximum tensile strain criterion.
This also means that debris are not maintained and taken into consideration in this model, as
crushed and broken elements are deleted and totally removed. However, the loading effect of the
debris is already accounted for through the MCS plateau level in the shell element.
- In Traction by the fibers (and to a lesser extent by the resin, with a damage parameter) to modelize
breakage by brittle fracture,
- In Compression both by the fibers and the resin, with the fibers related to compressive breakage
and the resin related to crushing damage, notably to include kinking and plasticity.
In order to verify the behavior law that was implemented, elementary numerical testing
corresponding to basic and common mechanical stress was performed on a single element. These
numerical trial cases are presented in Figure [19], where the boundary conditions and applied load
are indicated. For a greater clarity, schematization is portrayed from the top of the element, and only
the shell mean plane section is shown.
By varying the direction of applied loads and the coefficient ratio of the imposed velocity in case of
traction-compression, a full circular picture of the stress was drawn along Principal Stress I and II axes
(Figure [20]).
Figure [20] presents the response of the element to the basic individual tests with and without the
damage parameter implemented. The scenario with activated damage parameter provides two clear
fronts of behavior: a brittle rupture surface in traction-dominated loading and a plasticity surface in
compression-dominated loading.
The final shape of the stress curve was found to correspond to the desired behavior, therefore
validating the implementation of the behavior law within the coded element.
Figure [20]: Numerical testing results for the elementary trial cases, presented with and without the
damage parameter implementation, for the woven-type elements.
The UD-type element (Figure [12]) was also tested in traction and compression. Although tensile
testing correctly rendered brittle rupture after reaching the rod maximum tensile strain, compression
testing of this element was not satisfying, as it was anomalously behaving (Figure [21]).
The modeling strategy used in this model, which consists in incorporating the compression of the
fibers within the rod elements, led to highlight a strong instability of those rods elements when used
in compression, as displayed in Figure [21], presenting a chronological evolution of a compression
test on a unitary UD-element.
Figure [21]: Compressive behavior of the unidirectional element, highlighting its instability.
Rod elements (fibers) have been graphically added on the successive deformed step-images in Figure
[21], becoming shorter with the increase of the imposed displacement (dz in mm) (red rods) and
remarkably rotating and shifting on the last illustration (blue rods), when they became unstable.
These UD elements became highly unstable under this type of stress, leading to the emergence of a
buckling mode of the element. In order to recover stability, the rods elements tended to rotate and
reach a more stable position, which is less energy consuming. This was revealed when testing the
unidirectional-type element. This negative behavior was not observed in the woven-type elements,
which possess perpendicular rods of identical stiffness in each direction, preventing the whole
element from shifting on the side.
Further illustration of the rods unexpected abnormal behavior is given by the Stress-Displacement
curves in Figure [22], which shows the compressive stress (z) of the rods elements in a unitary
unidirectional-type element in compression. From the curve representing the stress in the 2 “blue”
rods in Figure [22], yielding occurs for a displacement value of – 0.125 mm, corresponding to the rods
axial shifting and element distortion observed in Figure [21] (transition from – 0.125 to – 0.126 mm).
A small but clearly visible burst, due to a shock-wave effect, is simultaneously detected for the same
displacement value on the curve representing the stress in the 2 “red” rods, which behave correctly
and in accordance with the behavior law governing them, from a linear elasticity evolution up until
the compression strain rupture level is reached, at which point the rods brutally break.
The encountered problem highlights the limitations of a “rods-based model”. To remedy this
problem, it would be required to upgrade the current version of the element to a “beams-based
model”, in order to take into account and reliably reproduce the flexion rigidity of the fibers. The
current approach stands at the limit of possibilities of this unconventional fibers/resin uncoupling
model. In traction, it is working relatively well, however it causes major issues in compression,
especially to modelize the kink-bands.
It was consequently decided to remove the rod-elements’ contribution to the whole element, when
it was loaded in compression and homogenize the behavior of the fibers/resins inside the shell
element.
A lead of amelioration would be to instore a coupling between warp and weft rod elements, (that is
to say between 2 fibers), for the Woven-type elements. This additional condition could either be a
fiber-fiber coupling or a fiber-matrix coupling, which could be more beneficial. This is all the more
true for the unidirectional-type elements, which were the most unstable. This instability pointed a
limitation to this numerical model: when used in compression, the model extends too far in the fiber-
resin dissociation. That specific aspect of the model is particularly problematic for representing the
kinking effect, as this mode of fibers rupture is directly linked to the shear behavior of the resin [23].
Tube specimen 1 was chosen as the thread for model building. Its structure is a symmetrical and
even 10 plies stacking, consisting of one 0.34 mm 0/90° woven taffeta, eight 0.175 mm 0°-oriented
plies and one 0.34 mm 0/90° woven taffeta. This stratification and its simplified schematization are
presented in Figure [23]. To optimize the modelling computation time and gain in structure
simplicity, only 2 UD central plies of the structure are built in the numerical model, using the
symmetrical and even geometry of the initial structure, although actual dimensions for those plies
are respected, as displayed in Figure [23]. Moreover, the geometric constitution of the UD-elements
used in the model helps simulating the equivalent of 2 UD plies per elements, therefore raising the
number of 2 modelled plies to a 4-equivalent-ply pattern. In the end, as a result, every other central
UD ply of the actual structure, that is to say a total of 4 plies, is modelled using 2 rows of elements.
Figure [23]: Schematization of the modelled structure with nodes positioning and element size (right)
based on the laminate structure schematization (left)
This stacking is therefore composed of 2 woven plies, using elements (6) and (11) from the unitary
elements library and 2 central unidirectional plies, using element (17). A graphical illustration of the
modelled structure based on the actual tube is given in Figure [24].
Table 3 gives the numerical input values for different key parameters on which the model relies.
Due to the instabilities of the rods elements encountered in compression, it was required to revert to
a more classical homogenized model for the compressive aspect, therefore needing the
homogenized modulus given by a classical rule-of-mixture for composite materials. Those modulus
values are reported in Table 4.
Table 4: Material stiffness properties for the Woven and Unidirectional (UD) plies, resulting from the
Classical Lamination Theory and rule-of-mixture for composite materials.
It can be pointed out that both the traction and compression homogenized stiffness, Ets and Ec, were
taken at the same value, mostly by choice of simplicity.
The tubular structure described above (Figures [23], [24]) is contrived and submitted to crushing
between 2 infinite rigid planes, one of which is fixed and acts as the crushing surface (bottom plane)
while the other is vertically moving downwards and applies the crushing load on the whole structure
(upper plane) (Figure [25]).
Contact and friction conditions have been tested and varied to optimize the crushing simulation,
leading to a configuration where the tube’s top nodes in contact with the upper-moving plane are
tied and the tube’s bottom nodes in contact with the bottom-fixed wall are sliding freely or with
friction (Figure [25]).
The load applied to the structure to induce crushing is an imposed velocity (Vimp), which general
expression corresponds to a linear progressive increase, as shown in the schematization (Figure [25]).
That imposed velocity was chosen to be progressive to solve some instability issues and avoid a
brutal loading which created numeric wave-rebounds within the structure.
Figure [25]: Schematization of the boundary conditions for the numerical tubular structure testing.
Numerical results
- with 1 ply
As a first step, the numerical response of tubular structures consisting of 1 individual ply of
equivalent thickness (2 mm) constituted with Woven-elements only then with UD-elements only was
tested. Optimistic results of these two configurations are presented in Figure [26].
- with 4 plies
As a second step, the full realistic tubular structure presented in Figure [23] was tested.
Unfortunately, that tubular structure with a Woven-UD-UD-Woven pattern did not produce expected
results (Figure [26] and [27]).
Figure [26] presents the mean vertical stress (z) over the axial vertical displacement (dz) for 3
circular tube configurations: one tube constituted of 1 ply of pure UD elements ((1), Figure [14]), one
tube constituted of 1 ply of single Woven elements ((17), Figure [14]), and one tube constituted of 4
plies with the real structure Woven-UD-UD-Woven (Figures [23], [24]). In each case, the total wall
thickness is identical for all tubes, and equal to 2 mm.
The behavioral response of the 1-ply tubes, whether UD or Woven, is consistent with the expected
results and the implemented behavior law: a vertical stress (z) decreasing down to -250 MPa, and a
relatively stable plasticity level, starting from that point and lasting throughout the crushing phase. In
addition, the respective compression stiffness for each structure are coherent, with the
unidirectional’s stiffness (blue curve) being twice the woven’s stiffness (red curve) value. The half
stiffness value for the Woven tube compared to the UD tube is logical considering there are twice as
many fibers oriented in the loading direction for the latter.
The lack of “opening” at the bottom of the 4-ply tube, on the free-crushing-face and high
concentration of stress on the inside plies lead to a total collapse of the structure, through the rapid
yielding of the unidirectional plies, causing rapid delamination spreading.
Figure [27] depicts the state of the 4-ply-W-UD-UD-W tube in crushing simulation and highlights the
instability of the model for this structure under compressive crushing. The inner unidirectional plies
are the first to undergo plastic deformation, which destabilize the whole structure as it leads to a
rapidly propagating delamination of the entire UD-plies.
Figure [27]: Mean vertical stress z (above) and delamination indicator (below)
for the Woven-UD-UD-Woven tube structure at the beginning of the crushing simulation.
Approximations with a 4 consecutive woven plies (W-W-W-W) structure tended to the expected
results. That configuration was tested as the W-UD-UD-W structure didn’t work properly. Results
with the approximate W-W-W-W stratification structure are presented below. Some approximations
using an inner central Woven structure with UD properties were also attempted. Although outcomes
from those simulations also tended toward acceptable results, they are not presented in this study.
Associated results were very similar to those of the 4 consecutive woven plies (W-W-W-W) structure.
Figure [28] illustrates the difference between removing the elements (i) and stacking them (ii). To
highlight the stacking effect, illustrations for Figure [28](ii) are given for 2 different imposed
displacement values (dz in mm), a few time-step apart. Deformed elements that have fully plasticized
are kept in the model and they keep piling up (Figure [28](ii) right), while they are simply deleted in
the first case (Figure [28](i)) after reaching a maximum strain deformation.
Figure [28]: Numerical simulation with elements deletion (left) and elements stacking (right)
This simulation was achieved with a 4-woven-ply (W-W-W-W) structure. The stacking option in the
modelization is not utterly proper, but it helps avoiding numerical instability and is rather easy and
straightforward to implement. It should however be avoided if the crushing of a significant number
of elements is intended, as it keeps a large number of damaged elements in the model, and because
it has no physical meaning after some extensive strain is reached.
Load signals resulting from the numerical calculations of these simulations were filtered using a time-
dependent cutoff filter given by (Eq. 10):
Significance and influence of this filter are illustrated by Figure [29], displaying the outcome of the
same crushing simulation, with filtered(*) and unfiltered results.
Figure [30] presents the filtered numerical calculation results for the two previous simulations:
Deletion or Stacking of the elements (as presented in Figure [28]). The initial element size is 1.4 mm.
Figure [30]: Stress-Displacement curve for tube crushing with deletion or stacking of the elements.
A small burst is periodically and regularly observed on the curve corresponding to the stacking
simulation (red curve). Every burst corresponds to a new element entering its plasticization stage. It
represents the passing to the next vertical element, when the previous just below it has fully
plasticized and this is reproduced for each element.
It can be noticed that in case of the deletion simulation, the burst periodicity corresponds to 3
elements which are removed simultaneously (a vertical height of 3 elements, actually). Since those
elements are in a similar plastic state and have reached a very close strain, the shock wave generated
by the deletion of the first element to reach the deletion criterion also deletes its immediate
neighbors, explaining why more than 1 vertical element is removed at a time.
Due to the uniform vertical axial loading and the chosen deletion condition, a full radial row of
elements (i.e. a full torus of elements with the same vertical altitude) is deleted at once (cf. Figure
[31]). Simulations were attempted using a tilted plane as the crushing surface in order to avoid
complete and simultaneous deletion of an entire torus of elements, but was not effective.
Several attempts were made to try and open the wall’s structure of the tube to create fronts:
- Shifting the ground nodes
- Acting on the sliding and friction properties
- Changing the delamination properties and interface parameters
- Delay the removal of elements
Most of those actions were taken on the mechanical, geometrical, or boundary conditions properties
of the ground nodes, that is to say the first raw of elements in direct contact with the crushing plane,
as illustrated in Figure [31].
Figure [31]: Schematization of the crushing process and localization of the nodes/elements where
most actions were taken.
As the initial behavior was not able to numerically reproduce the splaying in a progressive way,
several attempts were made to open the composite between consecutive plies and control this
rupture. Two of these attempts are presented in Figure [32].
Figure [32]: Illustration of some attempts to force the delamination opening in the structure:
(a): acting on the initial geometry, (b): acting on the friction parameter.
Diagrams in Figure [32] are drawn using a single ply to ease reading. The first configuration (a)
corresponds to a geometrical action consisting in radially shifting the ground nodes by half an
element’s initial length. The second configuration (b) is focused on contact properties and consists in
varying the boundary condition and friction sliding coefficient. Results of those two attempts are
presented thereafter.
Radially shifting the first row of bottom elements (Figure [32](a)), in contact with the crushing plane,
by a distance of half the element’s size, to force an open-splaying mode was not successful, as
delamination spread upward too fast. This is illustrated by Figure [33], where graphic animations are
displayed one time-step apart. The imposed displacement dz (given in mm) progressively increases.
The delamination indicator for the successive animations shows the delamination spreading upward
too quickly compared to the loading, leading to instabilities that destabilize the entire structure and
ends the simulation.
The friction coefficient between the row of bottom elements and the crushing plane was varied from
0.5 to 0.01 gradually to try and replicate a realistic resistance to the sliding of the bottom of the tube
on the infinite plane (Figure [32](b)).
Also this coefficient was adapted to try and enhance the opening and progressive separation of the
plies to recreate delamination. It has been observed that a complete “tie” of the bottom nodes to
the rigid wall would generate higher stress on upper nodes and not produce any kind of opening or
progressive crushing from the bottom. Conversely a complete sliding without any friction would let
the plies slip too easily and quickly, creating a delamination opening that would go up the ply and
create the structure to implode as a result of lack of rigidity (combination of buckling and loss of
rigidity of “independent” thinner plies). A quick study was done by varying the sliding friction
coefficient. Intermediate sliding using this friction coefficient would bend either way, without being
able to converge on an appropriate friction coefficient parameter.
It was also attempted to reduce the properties of the cohesive elements of the interface between
plies, without further success.
To summarize, numerous numerical instabilities were still persistent and most likely due to the
explicit nature of the calculations. Instabilities within the explicit model seem to destabilize the tube
structure. Usually, such instabilities may be controlled using numerical viscosities parameters
inherent to the elements. However, in the present case, even changing and tampering those
viscosities didn’t help solving the problem, as such actions were still unable to stabilize the model.
This work is a first approach in building and developing a FEA model for composite structures
behavior response prediction under crushing loading, based on a user-developed element at the
woven pattern scale that dissociate resin/fibers behavior and tension/compression behavior.
Although displaying some promising results, notably regarding the woven-based structure, the model
led to highlight some limitations and instabilities.
One of these instabilities came from the irregular behavior of rods elements in compression. The
others came from the modelization strategy.
This specific model separates matrix and fibers by its constitution, which is useful in tension. It allows
for the resin to damage while the fibers withstand and continue to work in the form of a network,
either interweaved for the braided or continuous for the UD structure.
In compressive crushing, kinking occurrence holds a prominent role. Yet, the kink-band phenomenon
has the fibers and matrix working simultaneously and closely tied. During kinking, the strain strength
of the fibers is narrowly tied to the shear strength of the matrix. Hence, the current model, which
separates the two, may not have been fully adapted for a composite crushing modelization.
Perspective:
It could be attempted to replace the truss or rods elements by beams elements, to modelize the
fibers. This elements substitution could add a fibers/resin coupling that might help stabilize the
problem encountered in compression. On the other hand however, that addition would cancel the
fibers/resin decohesion, which is the key point of this model.
Expansion of the “3D” elements in contact with the crushing surface and change them in a
trapezoidal shape to help open the plies and therefore force delamination (and also represent the
debris).
Implement or update the current model with a version of the Radioss software withstanding the
communication between elements. It is however uncertain that it will help in solving the instabilities
issues.
Once the different options and possibly beneficial additions to the model mentioned above have
been tested and are considered effective, it would be interesting to try and test other crushing
modes, especially by changing the boundary conditions to reproduce different trigger initiations.
Correspondingly, other braided patterns, fiber orientations and material types could be numerically
tested to verify the validity and robustness of the model.
Thanks are duly addressed to the CALMIP High Performance Computing platform for the numerical
support and the computational power it provided. Financial support for this work was obtained
thanks to FEDER funds through the joint project SKYSEAT co-funded by ARTEC Aerospace Company,
the Occitanie – Pyrénées-Méditerranée Region (ex-Midi-Pyrénées Region) and the ISAE-SUPAERO.
References:
[1] [AL GALIB 2004] AL GALIB, D., LIMAM, A. Experimental and numerical investigation of static and dynamic axial
crushing of circular aluminum tubes, Thin-Walled Structures, Volume 42, Issue 8, August 2004, Pages 1103-1137, ISSN
0263-8231, DOI 10.1016/j.tws.2004.03.001.
[2] [PINHO 2004] PINHO, S.T., CAMANHO, P.P., DE MOURA, M.F. Numerical simulation of the crushing process of
composite materials, International Journal of Crashworthiness, Volume 9, Issue 3, 2004, Pages 263-276, ISSN 1358-
8265, DOI: 10.1533/ijcr.2004.0287.
[3] [BISAGNI 2005] BISAGNI, C., DI PIETRO, G., FRASCHINI, L., TERLETTI D. Progressive crushing of fiber-reinforced
composite structural components of a Formula One racing car, Composite Structures, Volume 68, Issue 4, May 2005,
Pages 491-503, ISSN 0263-8223, DOI 10.1016/j.compstruct.2004.04.015.
[4] [ZENG 2005] ZENG, T., FANG, D.-N., LU, T.-J. Dynamic crashing and impact energy absorption of 3D braided composite
tubes, Materials Letters, Volume 59, Issue 12, May 2005, Pages 1491-1496, ISSN 0167-577X, DOI
10.1016/j.matlet.2005.01.007.
[5] [MAMALIS 2006] MAMALIS, A.G., MANOLAKOS, D.E., IOANNIDIS, M.B., PAPAPOSTOLOU, D.P., The static and dynamic
axial collapse of CFRP square tubes: Finite element modelling, Composite Structures, Volume 74, Issue 2, 2006, Pages
213-225, ISSN 0263-8223, DOI 10.1016/j.compstruct.2005.04.006.
[6] [FLEMING 2008] FLEMING, D.C., MORROW, C., CLARKE, C.W., BIRD, C.E. Finite Element Simulation of Delamination
with Application to Crashworthy Design, Journal of the American Helicopter Society, Volume 53, Number 3, 1 July
2008, Pages 267-281, DOI 10.4050/jahs.53.267.
[7] [XIAO 2009] XIAO, X., BOTKIN, M.E., JOHNSON, N.L. Axial crush simulation of braided carbon tubes using MAT58 in LS-
DYNA, Thin-Walled Structures, Volume 47, Issues 6–7, June–July 2009, Pages 740-749, ISSN 0263-8231, DOI
10.1016/j.tws.2008.12.004.
[8] [XIAO 2009] XIAO, X., McGREGOR, C.J., VAZIRI, R., POURSARTIP, A. Progress in braided composite tube crush
simulation, International Journal of Impact Engineering, Volume 36, Issue 5, May 2009, Pages 711-719, ISSN 0734-
743X, DOI 10.1016/j.ijimpeng.2008.09.006.
[9] [TONG 2018] TONG, Y., XU, Y. Improvement of crash energy absorption of 2D braided composite tubes through an
innovative chamfer external triggers, International Journal of Impact Engineering, Volume 111, January 2018, Pages
11-20, ISSN 0734-743X, DOI 10.1016/j.ijimpeng.2017.08.002.
[10] [SUPIAN 2018] SUPIAN, A.B.M., SAPUAN, S.M., ZUHRI, M.Y.M., ZAINUDIN, E.S., YA, H.H. Hybrid reinforced thermoset
polymer composite in energy absorption tube application: A review, Defence Technology, Volume 14, Issue 4, 2018,
Pages 291-305, ISSN 2214-9147, DOI 10.1016/j.dt.2018.04.004.
[11] [MEIDELL 2009] MEIDELL, A. Computer aided material selection for circular tubes designed to resist axial crushing,
Thin-Walled Structures, Volume 47, Issues 8–9, 2009, Pages 962-969, ISSN 0263-8231, DOI
10.1016/j.tws.2009.02.003.
[12] [THORNTON 1990] THORNTON, P.H. The Crush Behavior of Pultruded Tubes at High Strain Rates, Journal of
Composite Materials, Volume 24, Issue 6, 1 June 1990, Pages 594-615, ISSN 0021-9983, DOI
10.1177/002199839002400602.
[13] [HULL 1991] Hull, D. A Unified Approach to Progressive Crushing of Fibre-Reinforced Composite Tubes, Composites
Science and Technology, Volume 40, Issue 4, 1991, Pages 377-421, ISSN 0266-3538, DOI 10.1016/0266-
3538(91)90031-J.
[14] [CHIU 1998] CHIU, C.H., TSAI, K.H., HUANG W.J., Effects of Braiding Parameters on Energy Absorption Capability of
Triaxially Braided Composite Tubes, Journal of Composite Materials, Volume 32, Issue 21, 1 November 1998, Pages
1964-1983, ISSN 0021-9983, DOI 10.1177/002199839803202105.
Conclusions
Le but de cette étude est d’évaluer la capacité de diverses structures composites à dissiper de
l’énergie lors de l’écrasement, ainsi que de comprendre les phénomènes mécaniques et
d’endommagements entrant en jeu lors du crushing, dans le but d’une optimisation de ces structures
menant à proposer une structure optimale de tenue au crash tout en permettant une grande
dissipation d’énergie.
Différents drapages et orientations de fibres, ainsi que plusieurs matériaux, ont ainsi été testés.
Plusieurs systèmes originaux d’initiation du crash ont également été expérimentés, toujours dans le
but d’améliorer le comportement du crushing et ainsi augmenter la dissipation d’énergie.
Une première étape dans la création d’un modèle numérique simulant le crushing de structures
composites a été réalisée. Ce modèle numérique a été construit sur la base de précédentes études
menées au laboratoire et enrichi sur la base des essais expérimentaux réalisés au cours de l’étude.
Le point de départ et la ligne conductrice de cette étude a été le développement d’un siège
aéronautique en composite respectant la réglementation en vigueur et permettant ainsi une
dissipation rapide de l’énergie perçue en cas de choc ou de crash.
Dans un premier temps, une étude préliminaire sur le comportement et la tenue au crash des sièges
d’avions avec leurs passagers a été menée. Cette étude a permis de comprendre les dynamiques
mises en jeu lors de tels phénomènes, d’enrichir les connaissances relatives aux systèmes dissipatifs
d’énergie ainsi que sur la réglementation en vigueur dans ce domaine, et enfin de dimensionner les
structures composites nécessaires à la réalisation d’un tel système pour répondre au cahier des
charges fixé. La seconde partie du travail préliminaire a porté sur la confection et l’acquisition de
structures tubulaires composites à des fins de tests expérimentaux.
Lors de l’étude quasi-statique, des valeurs de SEA atteignant jusqu’à 140 kJ.kg-1 ont été obtenues
pour des tubes mixtes UD/Tissés Carbone-Epoxy, conduisant à des résultats supérieurs à la plupart
des exemples de la littérature, de l’ordre de 50 à 60 kJ.kg-1 en moyenne, et atteignant jusqu’à
100 kJ.kg-1 pour des structures tissées Carbone-Epoxy récentes.
D’une manière générale, il a été observé qu’en quasi-statique les structures contenant des plis
orientés à 0° atteignaient une meilleure valeur d’absorption d'énergie que les structures n’en
procédant aucun.
En outre, il a été établi qu’en chargement de type statique, un composite unidirectionnel orienté à 0°
stabilisé par des plis tissés en peau extérieure et intérieure correspondait à la structure offrant la
meilleure tenue au crushing et renvoyait ainsi les meilleures valeurs de SEA.
Un renfort tissé positionné sur les faces internes et externes des tubes s’est avéré être un moyen
efficace pour stabiliser la structure et guider son crushing. De surcroît, ces plis tissés aident
fortement à tenir l’ensemble des plis orienté à 0°, évitant ainsi un « splaying » – ou délaminage – ou
un évasement trop important et trop rapide.
En outre, comme ces plis sont nécessaires structurellement, en particulier pour la tolérance aux
dommages d’impact, il s’agit de les mettre à profit et en tirer avantage, notamment avec le cas du
confinement conique vers l’intérieur, qui contraint fortement les fibres tissées transverses de ces plis
lors du crushing.
En parallèle, un confinement contraint vers l’intérieur semble être le moyen le plus efficace dans la
majorité des cas pour réduire le pic initial de chargement tout en conservant une bonne valeur de
SEA.
Enfin, concernant les renforts aramides tissés utilisés sur certains tubes, bien que ces derniers
n’apportent aucun avantage d’un point de vue de la dissipation d’énergie, ce type de renfort
recouvrant partiellement ou totalement la structure peut agir comme un filet et ainsi retenir les plis
centraux en carbone, en réduisant fortement l’évasement et l’expansion vers l’extérieur de la
matière fragmentée et contenant les débris à l’intérieur de la structure, agissant comme une gaine se
repliant autour d’elle-même. Ceci peut être extrêmement avantageux dans le contexte aéronautique
afin de réduire la multiplication et l’expulsion de débris hors de la structure.
En dynamique, des valeurs de SEA atteignant 90 kJ.kg-1 ont été obtenues pour des tubes
multidirectionnels Carbone-Epoxy, supérieures à la plupart des valeurs de la littérature se situant en
moyenne autour de 50 - 60 kJ.kg-1.
Il a été montré qu’en chargement dynamique, la présence de plis orientés à 90° stabilisés par des plis
de tissus permettait grandement d’augmenter les capacités de dissipation d’énergie de la structure.
Les structures contenant des plis orientés à 90° par rapport à la direction de chargement en
compression ont atteint les valeurs de SEA les plus élevées, tandis que celles composées de plis
orientés à 0° coïncidant ainsi avec la direction de chargement et étant les plus performantes en
quasi-statique sont parmi les plus faibles en dynamique.
Ces stratifiés, constitués majoritairement de fibres orientées à 0° ont enregistré une baisse de leur
performance de plus de 50% entre le quasi-statique et le dynamique. Ceci pourrait s’expliquer par la
présence d’un trop grand nombre de plis consécutifs orientés à 0° qui tendrait à réduire le taux de
restitution d’énergie critique en rupture de fibres avec un phénomène de type « kink-band ».
Par ailleurs, en dynamique, un crushing avec un confinement forcé vers l’intérieur est encore une fois
plus efficace qu’un simple crushing sur une surface plane, réduisant le pic initial sans faire chuter
drastiquement la valeur de SEA atteinte.
Cependant un confinement conique vers l’intérieur, bien qu’ayant toujours pour effet la réduction du
pic initial, réduit également la valeur globale de SEA. Ceci est toutefois vrai pour les seules
dimensions de cônes utilisées. Une étude plus complète sur la taille et les dimensions des cônes,
notamment le degré d’inclinaison de la pente, aurait le mérite d’être réalisée.
Enfin, pour parachever et compléter cette étude, un modèle numérique simulant le crushing de
structures composites a été développé. Ce travail de modélisation représente une première étape
dans la représentation des phénomènes et dans la réponse aux sollicitations de type « crash »
permettant la prédiction du comportement en écrasement de structures composites.
Ce travail a permis d’apporter plusieurs éléments originaux, à la fois sur le plan expérimental et
numérique. Notamment, ont été introduits et testés différents mécanismes originaux d’initiation du
crushing, permettant dans certains cas d’augmenter les contraintes de compression et ainsi
optimiser la dissipation d’énergie. Des images réalisées par tomographie ont également apporté des
réponses sur les mécanismes d’endommagements intervenant lors du crash de ces structures.
Cette thèse constitue une première étape dans la compréhension des phénomènes, la mise en
application et l’optimisation d’un absorbeur d’énergie en composite, dans le cadre d’une application
dans le domaine aéronautique.
Une meilleur sélection et connaissance des matériaux utilisés pour la confection des tubes semble
évidente. Dans le même ordre d’idée, il pourrait être utile de s’intéresser au moyen de mise en
œuvre et de fabrication, ce qui est peu présent dans la littérature.
A ce sujet, une piste envisagée pour l’amélioration de la fabrication des tubes confectionnés avec le
CRITT Mécanique & Composites serait de remplacer la méthode de fabrication utilisée par un
procédé de type RTM et ainsi mieux contrôler le taux d’imprégnation de la résine afin de réduire les
porosités.
Une interrogation peut être portée sur les aspects géométriques de la structure « tube »,
notamment sur le ratio diamètre/épaisseur et diamètre/longueur, toujours à des fins d’optimisation
de la structure.
L’étude menée sur les systèmes d’initiation coniques – intérieurs et extérieurs – a permis de mettre
en évidence l’influence marquante et l’intérêt de tels moyens, notamment à des fins d’optimisation
de la valeur atteinte de SEA. Cependant, un nombre limité de cas a pu être testé. Il serait utile et
nécessaire de mener une étude complémentaire sur les dimensions de ces pions (ou gorges)
coniques, et particulièrement sur l’inclinaison de la pente de ces pièces coniques, engendrant les
contraintes circonférentielles. L’étude du frottement pourrait également permettre d’optimiser ces
systèmes.
Par ailleurs, en se rapportant à la littérature, une amélioration des performances en crushing a été
observée grâce à l’utilisation de structures chanfreinées. Il pourrait être intéressant de combiner cet
autre mode d’initiation aux systèmes de cette étude.
Enfin, toujours d’après la littérature, de nombreuses études relèvent les meilleures performances
atteintes par d’autres types de matériaux composites, et en particulier ceux à base de résines
thermoplastiques. S’ils sont certifiés FST (Flame, Smoke and Toxicity), ces matériaux peuvent
constituer une piste sérieuse pour l’élaboration de sièges aéronautiques.
Une étude de la taille des débris générés, ainsi qu’une comparaison de la taille et de la forme entre
essais quasi-statiques et dynamiques, pourraient être utiles et apporter des éléments
supplémentaires concernant les mécanismes d’endommagements mis en jeu lors de ces deux modes
d’essais. Ceci pourrait permettre d’apporter des explications sur les nombreuses différences
observées en termes de contraintes de crushing, et donc de SEA atteintes, entre ces deux modes de
chargement pour des structures et configurations identiques.
Afin de s’assurer de la validité de tous ces résultats, un nombre d’essais plus conséquent en
dynamique devra être envisagé, et notamment profiter de cette étude pour faire varier les vitesses
initiales d’impact. L’état de l’art pointe une influence non négligeable de cet effet de vitesse sur les
résultats, mais ces résultats ne sont pas toujours cohérents entre les différentes études.
Pour mieux rendre compte de la flexion, prendre en compte l’effet de « kink-band » et palier aux
nombreuses instabilités liées aux éléments barres en compression, il faudrait remplacer ces éléments
« barres » par des éléments « poutres » dans le présent modèle.
L’interface implémentée et utilisée dans la version actuelle du modèle a été paramétrée et validée
précédemment dans le cas de plis tissés superposés. Le développement d’autres interfaces
spécifiques à la jonction de plis UD-Tissé est peut-être nécessaire.
Enfin, pour palier à de nombreux problèmes d’instabilité mais aussi pour rendre compte de l’effet
d’endommagement, il serait utile de pouvoir instaurer une communication entre éléments. Ceci
aurait pour effet de tenir compte de l’état des éléments voisins.
Finalement, une fois toutes les pistes d’amélioration proposées ci-dessus testées et prises en compte
dans le modèle numérique, ce dernier pourra être utilisé de manière à tester différentes
configurations de stratifiés et conditions limites pour correspondre aux essais réalisés, notamment
avec la représentation des pions et gorges coniques, ainsi que faire varier les vitesses de chargement
imposées.