Composite Failure

LARGE COMPOSITE SPACE STRUCTURES: FAILURE ANALYSIS AND EXPERIMENT

Vibration Suppression – Precision Motion Control

Emmett Nelson, Firehole Technologies

Adam Biskner, CSA Engineering

Presented to:

AIAA Rocky Mountain SectionAIAA Rocky Mountain Section

January 29th, 2009

1

STRU

CStructural Failure Test Program

AFRL Static Test Facility• AFRL is directing an investigation to d th bilit f it

CTURA

LFA

advance the capability of composite failure analysis from the coupon level to full‐scale structures

• Testing conducted by CSA Engineering in th AFRL St ti T t F ilit

AILU

RETEST

the AFRL Static Test Facility

• Firehole Technologies provided analysis using Helius:MCT

• Tested three previously flight qualified EDU t t t f il

TPRO

GRA

M

EDU structures to failure • CASPAR MPA (Minotaur IV), Atlas V ISA,

and Delta IV PAF

• Compared to failure predictions from conventional FE models and advanced FE

Conic ISA

MPA

M

conventional FE models and advanced FE models to test data

• Evaluating the validity of the design by comparing the structural capacity to the flight conditions

CASPAR

g

• Scaled worst case qualification load profiles and increased applied load until structural failure was achieved• CASPAR and ISA experienced a “flight

Delta IV 1780 PAF


• CASPAR and ISA experienced a flight‐like” composite failure

2

Pump

LOADC

Hydraulic Service Manifold (HSM)

Distribution Manifold (S V l )

CONTRO

LA

(Servo Valves)

Actuator

ANDDATA

A

Load Cell

Load

ACQ

UISITIOLoad

Controller

UPS

LVDT

ON

Data Acquisition System


System

3

INTEG

RA• Agilent® data acquisition system

• 256‐channel front‐end ATED

INST

• Fully integrated with the load controller with MTS software

• All channels recorded at 1% load intervals or as required RU

MEN

TAT

intervals, or as required

• Sensors • Typical test includes only strain and

displacement TION

displacement• Able to condition anything with a voltage output

• Full bridge strain gage based deflection t d i f 0 25” t 5”transducers, ranging from 0.25” to 5”

• Digital video recorded during loading operations


4

• Load control parameters are t d

TEST

OActuator Control Profile

custom programmed per experiment

• Parameters are redundantly reviewed by QA engineer

PERATIO

N

reviewed by QA engineer

• All channels are controlled simultaneously in accordance with a load profilewith a load profile

• Concurrently subjected to 9 limit and error detectors

• Live data displayed during testPl d i l i l di i

250H ld P i

Live Data Comparison

• Plotted against analytical predictions

• Test can be paused or aborted at any time per engineering request 150

200PredictionGauge 1Gauge 2

Hold Points

0

50

100

0 20 40 60 80 100 120 140


0 20 40 60 80 100 120 140Percent of Flight Load

5

WHOIS

• Composite Adapter for Shared PAyload Rides

• Multi‐payload adapter (MPA) for Minotaur IV CASPA

R?

Vehicle • Utilizes excess Peacekeeper missile motors to provide low‐

cost LEO launches (~$20 mil.)

• Nominal payload capability of 4000 lbm

• Designed to integrate two primary spacecrafts (1000‐2000 lbm) per Minotaur launch• Different design approach than previous MPA’s

• Composite material minimizes payload mass penalty• IM7/8552 unidirectional tape• 2 Identical monocoque shells • 60 inches tall, 74 inches in diameter• Integrated composite flangesIntegrated composite flanges • 62.01” diameter bolt pattern

• One primary stowed, other placed atop adapter• Requires Latching Lightband (LLB) low shock separation

d l d b l S C i


systems, developed by Planetary Systems Corporation• Bonded only joint between LLB/CASPAR

6

CASP• Test designed to drive failure in the transitional radius between the conic

section and the aft flange

PAR T

ESTD

section and the aft flange

• Shear, moment, and axial load combination balanced to maximize aft compression while preventing failure in other critical regions

M i t FWD H d t < 3X li it

DESIG

N

• Max compression at FWD Hg adapter < 3X limit

• Max tension at lightband < 4X limit

• Maximize aft CASPAR flange compression

opposite access doors (270o)

Critical Aft Flange (Failure Region) Line LoadCritical Separation System Line LoadCritical Forward Adapter Line LoadLine Load (lbs/in) Line Load (lbs/in) Line Load (lbs/in)

Max Compression at Limit -252 Max Tension at Limit 114 Max Compression at Limit -252Max Applied Compression -755 Max Applied Tension 440 Max Applied Compression -1421

Predicted Failure -1000

Percent above Limit 564%


M.S. at Max Applied Load 0.00 M.S. at Max Applied Load 0.03 Percent above Predicted Failure 142%

7

CASPA

R

A t t L d (lb )*

• Two axial actuators apply pure compression (no bending)

• Lateral actuator applies moment

Test Stack

Applied Loads R TEST

Lateral AXI090 AXI27035760 -20250 -20250

Actuator Loads (lbs)*

*A positive load indicates a tensile actuator load

and shear

• Axial actuators biased to offload the weight of the load head and Hg adapter

• 100% represents flight line load used to

100 kip Axial Actuators

100% represents flight line load used to design the structure

• Failure test includes:• 50% and 100% checkout run

% f fl d

900

44 kip

Lateral Actuator

Load Head

• 250% Aft flange strain demonstration

• 884% failure run

Failure Test Profile

400

500

600

700

800

900

Lim

it Lo

ad

Primary LoadsCounter Balance

Actuator

Peak t i

Forward Hg Adapter

0

100

200

300

400

0 200 400 600 800 1000 1200 1400

% o

f L stress in aft flange 180° from the door

Failure Test Setup


Load Step Failure Test Setup

8

ATLA

SV

What is This Beast?• Atlas V CCB Conical ISA

Test Design• Test designed to drive failure in the top

f th d

V CONICIS

• On all Atlas V 400 series launches

• 12.5 foot diameter to 10 foot diameter, 65 inches tall cone

• Demonstration unit for lightweight

corner of the door • Failure mode suggested by ULA analysts

• First load case applied to as‐built ISA

• Failure load represents 200% of the A ANDTES

g gtooling development program

• Graphite/epoxy and honeycomb mandrel

• Typical composite tooling made from Invar• Long lead times and expensive

pgreatest FWD Flange Peq in ULA test plan

• Limited by the actuator capacities

• Second door installed since original door section did not fail ST

DESIG

N

Long lead times and expensive

• Heavy and difficult to work with

• Currently manufactured in Spain• Creates schedule and cost issues for ULA

C t d i d it b l t

section did not fail• 180o opposite original door, eliminating

pad up around door

• Door section cutout using original ATK tooling , process, and technician• Component redesign render unit obsolete

at the end of a 5 year, $6 million effort

g , p ,

• Second load case same as the first, applied max compression over new door


9

ISA T

• Test stack modified from Qualification Configuration

• Forward stiffness simulator removedR d

Qualification Test SetupTEST• Aft LOx tank simulator removed

• Center actuator biased to offload the weight of the load head

• Failure test includes:

Removed

• Failure test includes:

• 40% and 100% checkout run

• 200% failure run

• As‐built structure successfully withstood 200%

Removedy

load condition

• Second door installed

• Second case conducted by reversing the direction of the applied loads

Failure Test Setup

180%

200%

direction of the applied loads

Load Profile

20%

40%

60%

80%

100%

120%

140%

160%

% of Failure Load Primary Loads

Counter Balance


0%

0 100 200 300 400 500 600 700 800

Load Step

10

FAILU

• Produce a composite failure in the test article

• Record critical strain and displacement data at each load step RETEST

O

• Record critical strain and displacement data at each load step

• Acquire adequate load, strain, and displacement data such that quantitative assessment of the load bearing capacity of O

BJECTIVES,

the structures can be made and to allow comparison to pre‐test analytical models

• Identify resulting initial and final failure mechanism , AKA

SUCC

• Identify resulting initial and final failure mechanism

• Assemble test results, conclusions, and disseminate to community CESS

CRITEERIA


11

ANALY

Real World Failure Exercise• Provide blind failure predictions of large composite structures YSIS

OBJEC

• Provide blind failure predictions of large composite structures

• Predict initial failure, progression, and final failure

• Transfer new technology to the commercial analysis community CTIVES

• Model entire structure with a single detailed model

• Reasonable time frame (weeks)

• Why Firehole: Under the direction of AFRL, Firehole Technologies has been developing an advanced composites analysis technology for several years. The Structural Failure Test program was an great y p g gopportunity to validate the software, or learn where improvement was needed.


12

COMP

Firehole Technologies

• Firehole was founded in 2000 PANYOVER

• Our mission is to deliver tools and services that enable wide‐spread application of composite materials leading to lighter, stronger, safer and more fuel efficient structures RVIEW• Two distinct business areas:

Software DevelopmentStructural Analysis

• Firehole is a profitable, employee owned company focused on delivering more accurate results and a higher degree of confidence in

Software DevelopmentStructural Analysis


delivering more accurate results and a higher degree of confidence in composite simulations

13

FIREH

Current Products Upcoming ProductsHOLESOFTW

• Online, searchable database of composite material datasheets for material selection and comparison

• Cyclic loading simulation• Currently in Alpha• Development partnerships with a large W

ARE

SOL

material selection and comparison Development partnerships with a large Helicopter OEM and a major Naval contractor

LUTIO

NS

• Multiscale composites progressive failure technology • Layerwise Finite Element technology

• Simple transition between 2‐D (Shell) and ( l d) l h3‐D (Solid) Elements using the same

element/model

• Online, micromechanics based composite material simulator

• Composite simulation package for sustainable industries (Wind Turbine Blades Hydrogen Fuel Cells Lighter


14

Blades, Hydrogen Fuel Cells, Lighter Automobiles)

HELIUU

S:MCT

Helius:MCT™ is an enhancement to commercial finite element packages specifically for efficiently improving the accuracy of composite structures analysesanalyses.

• Uses fiber and matrix stresses to predict failure

• Extremely efficient

d d l• Standard material inputs

• Easy to adopt

• Always converges

• Proven results


15

CASPA

• First attempt at the CASPAR analysisAR: BLIN

DP

Model Details• Continuum shell elements PRED

ICTION

• 60 + plies through thickness (1 element)

• 15,000 elements• Handbook material N

SHandbook material characterization

• Fixed constraints at boundary• Continuous run time: overnight

l l d• Entire analysis completed in < 2 weeks

• Initial matrix failure: 1269% FLL%• Initial fiber failure: 1944% FLL


16

CASPACASPAR was successfully tested to ultimate

failure on April 14, 2008 AR: F

AILU

R

% FLL Failure Event

234 I iti l t i ki

RETEST

234 Initial matrix cracking sounds

319‐469 Occasional matrix cracking noisenoise

470 + Continuous matrix cracking noise

500 L b d i500 Lap band gapping

644 Door debond

792 Lower radius failure

847 Ultimate failure


17

CASP

Model Improvements

• 3D layer solid elements PAR : LESS

• 4 elements through thickness

• General mesh refinement

• Contact and nodal ties at Load Head(steel) O

NLEA

RN

Contact and nodal ties at bolt locations

Code Improvements

Test Adapter(7075 T7451 Alum.)1560 elements

(steel)54 elements

Forward Adapter(IM7/8552)

58600 elements EDCode Improvements

• Convergence algorithm improved

1560 elements

Access Doublers(IM7/8552)

558 elements each

Lightband(7075 T7451 Alum.)

464 elements

58600 elements

Aft Adapter

Model Details

• 122,146 elements

Aft Adapter(IM7/8552)

58600 elements

Base Plate(steel)1696 elements

• Continuous run time ~1½ days

• 8 node desktop p.c.

18

UPD

A

% FLL F il E t

Fiber FailureMatrix FailureNo Failure

ATED

ANALY

% FLL Failure Event

260 Initial matrix failure 520 % FLLFailure State

Fiber FailureMatrix FailureNo Failure

YSISRESU

L

261‐480 Matrix failure progression

500 + Rapid matrix failure

LTS500 + Rapid matrix failure progression

740 First fiber failure 800 % FLLFailure State

Fiber FailureMatrix FailureNo Failure800 Fiber failure in lower

radius

980 Ultimate Failure13001450

(discontinuity in load vsdisplacement)

840 % FLL


19

UPD

A3ATED

ANALY

2

2.5

in)

Potential ultimate failure

YTICALLO

A1.5

2

Displacem

ent (

ADVS

DISP1

Compressive DPLA

CEMEN

T

0.5

T

0

0 200 400 600 800 1000 1200 1400 1600

Flight Load (%)


20

Flight Load (%)

CASPPA

R: FAILULower Radius Failure

Ultimate Failure

Helius:MCT Ultimate Failure

URE

EVEN

TDoor Debonding

Helius:MCT Initial Fiber Failure

COMPA

RIS

Lapband Gapping

Helius:MCT Initial Matrix Failure

SONInitial Matrix Cracking

Occasional Matrix C ki N i

Continuous Matrix C ki N iCracking Noise Cracking Noise


21

STRA

INCOMPA

RRISON

IV13153

IV33151IM22701OV33151

IM12701

IV73151OV73151

IM12701


22

3 CASP

980

2

2.5

in)

PAR F

AILU

740

T U

ltim

ate:

9

mat

e: 8

47

1.5

2

Displacem

ent (

REPRED

ICT260

s:M

CT

Fibe

r:

Hel

ius:

MC

ASPA

R U

ltim

1

Compressive D TIO

NCOM

MC

T M

atrix

: 2

Hel

ius

: 110

0

r: 13

00

mat

e: 1

950?

CA

0.5

PARISO

NHel

ius:

M

ashi

n M

atrix

:

Has

hin

Fibe

r

Has

hin

Ulti

m

0

0 200 400 600 800 1000 1200 1400 1600 1800 2000

Flight Load (%)

H


Flight Load (%)

23

REA

SOONSFO

RD

First analytical discontinuity in global stiffness occurs 15% higher than actual ultimate failure.

DIFFEREN

CE

• Model does not capture lapband gapping

• Model does not capture door debonding

• Material disorganization occurring at flange radii not captured

ES• Material disorganization occurring at flange radii not captured

• Possibly reduce residual stiffness (ongoing work)

• Difficult to determine where ultimate failure occurs

24

ISA A

• ISA AnalysisANALYSIS

Model details• 3D model of entire structure• 3D layered solid elements M lti l l t th h thi k• Multiple elements through thickness

• Coupon material characterization• Model generation ~ 2 weeks • 192,000 elements

Load Head‐1440 solid, linear, reduced‐integration elements (Abaqus:C3D8R)

Forward Adapter/Splice‐2902 C3D8R elements

,• Continuous run time ~1 ½ days

• 8 node desktop p.c.

Composite Conic‐186,152 solid, linear, reduced‐integration, composite elements (Abaqus:C3D8RC3)‐These elements have one integration point per ply

Aft Adapter/Splice‐1518 C3D8R elements


25

ISA A

ISA As Built Helius:MCT Failure PredictionsASBUILT

HELIU

S:MCFailure of Structure

340% FL

Limit of Load Frame200% FL

T FAILU

RE

340% FL

PRED

ICTIOONS


26

ISA A

ISA As Built Test ResultsASBUILT

TE

The ISA was successfully tested to 200% FL on October 3, 2008.

•The structure responded nearly linearly to the loading.k bl

ESTRESU

LT

•Testing was remarkably quiet.

TS

1500

-1000

-500

00 50 100 150 200

DOOR233EMIN

800

1000

1200

14000 50 100 150 200

DOOR233EMAX

-3500

-3000

-2500

-2000

-1500

µStr

ain

% Flight Load

0

200

400

600

800

µStr

ain

% Flight Load


Experimental Data Firehole Technologies' Predicitons Experimental Data Firehole Technologies' Predicitons

27

MODI

ModificationsIFICA

TIONS

A second access door was cut into the ISA

l d

S

•180° opposite original door•No pad up around new door•Original ATK tooling was used•Honeycomb edge potted as original

•Loads were reversed


28

HELIU

Helius:MCT Predictions of Modified ISAUS:M

CT PRRED

ICTIONS

Initial fiber failure180% FL

SOFM

ODUltimate Failure

Initial matrix failure110% FL

IFIEDISA

187% FL


29

HELIU

Helius:MCT Predictions of Modified ISAUS:M

CT PRRED

ICTIONSSOFM

OD

186.72 % Flight Load 187.52 % Flight Load

IFIEDISA


187.68 % Flight Load

30

VIDEO

• VideoO


31

FAILU

Failure Test

The modified ISA was successfully tested to failure on October 24 2008 RETEST

The modified ISA was successfully tested to failure on October 24, 2008.

• 183% Flight Load

• Linear response

• Instantaneous event

• Door corners

Failure initiated at door corners

Rapidly propagated around circumference


32

ISA F

ISA FailureAILU

RE

Close up of upper door corner

Failure occurred on interior face sheets


33

RESU

LResults: Interior Strain Gauge

2000

ess

LTS: INTERI

1500

train

Tsai

-Wu

Max

Str e

IORSTRA

IN

1000

Prin

cipa

l St

nt CT

Hash

in

NGAUGE

500Max

P

Expe

rimen

Heliu

s:M

C

00 50 100 150 200 250 300 350 400

E


% Load

34

CONC

• Structural Failure Test Program Successful

• Two large space structures were tested to failure

CLUSIO

NS

• Two large space structures were tested to failure.

• Analytical results within 15% of ultimate failure on CASPAR

• Analytical blind predictions with 2.5% of ultimate failure on ISA

• Traditional composites analysis technologies over predict failure by a minimum of 1.5.

“I had anticipated that most large aerospace composite structures were considerably over‐designed, and this program proved that on all structures tested With innovative analysis technologies such as Helius:MCT fromtested. With innovative analysis technologies such as Helius:MCT from Firehole Technologies, I am convinced that these composite structures could remove as much as 40% mass, which translates into tremendous savings for many space applications.” g y p pp

Dr. Jeffry Welsh

Program Director

Chief Tier 3 Division ORS Office


Chief, Tier‐3 Division, ORS Office

35

Yellowstone Park’s Firehole River

BACKUP

COMP

Composites Failure TechnologiesPO

SITESFA

Conventional technologies treat composites like “black aluminum”

• Mask interactions

l l

ILURE

TECH

• Failure single event

• Unusable degradation models

• Exotic material parameters HNOLO

GIES

• Computationally unfeasible

Action in composites occurs in the Fibers OR the Matrix SAction in composites occurs in the Fibers OR the Matrix


37

TRA

DI

T i Hill

• Tsai‐Wu • Extension of Tsai‐Hill or

H ff t l t

ITIONALC

O

• Tsai‐Hill• Extension of von Mises to

orthotropic materials.

Hoffman to a general stress state.

• Invariant under coordinate rotation O

MPO

SITE

• Hoffman• Extension of Tsai‐Hill for

differing tensile and

• Use of tensors make mathematical operations easy

• Biaxial data required to determine failure parameters F

AILU

REC

compressive properties

• Simple to use in design

determine failure parameters.

• Hashin• Differentiates between fiber

d t i f il

CRITERIA

and matrix failure

• Distinguishes between tensile and compressive modes

All of these criteria are applied to a homogeneous composite and neglect the interaction between the fiber and matrix

38

MULT

Multicontinuum Theory (MCT)TICO

NTIN

UU

MCT decomposes composite stress into fiber and matrix stress• Based on Hill (1963)• Development @ Univ of Wyoming since 1988 U

MTHEO

R

Development @ Univ. of Wyoming since 1988

Accurately represent material phenomena

RY(M

CT)MCTMCT

Composite Stress


Fiber and Matrix Stress

39

COMP

Composite Under Mechanical LoadingPO

SITEUNDComposite Stress: DER

MECH

Fiber Stress:

σ11f = 108 Ksi

σ22f = ‐205 Ksi

σ11 = 0

σ22 = ‐200 Ksi

σ33 = ‐200 Ksi ANICA

LLO

σ22f 205 Ksi

σ33f = ‐205 Ksi

OADING

Matrix Stress:

σ = 143 Ksiσ11m = ‐143 Ksi

σ22m = ‐192 Ksi

σ33fm= ‐ 192 Ksi3


1 2

40

en1

Slide 40

en1 Update lamina pictureemmett_nelson, 12/15/2008

COMP

Composite Under Thermal LoadingPO

SITEUNDComposite Stress: Fib St

DER

THERM

Composite Stress:

σ11 = 0

σ22 = 0

σ33 = 0

Fiber Stress:

σ11f = ‐44.5 (MPa)

σ22f = ‐17.25 (MPa)

σ33f = ‐17.25 (MPa) MALLO

ADI

σ33f 17.25 (MPa)

NG

Matrix Stress:

σ11m = 66.75 (MPa)

σ = 25 87(Pa)

ΔT = ‐216 °C

σ22m = 25.87(Pa)

σ33fm= 25.87 (Pa)

3


1 2

41

PRO

GProgressive Failure Analysis

RESSIVEFA

matrixmatrix fiberfiber AILU

REANA

matrixmatrixfailurefailure

fiberfiberfailurefailure

ALYSIS

Material State 1Material State 1undamaged matrix,undamaged matrix,

Material State 2Material State 2failed matrix,failed matrix,

Material State 3Material State 3failed matrixfailed matrixundamaged matrix,undamaged matrix,

undamaged fibersundamaged fibersfailed matrix,failed matrix,undamaged fibersundamaged fibers

failed matrix,failed matrix,failed fibersfailed fibers

σc matrix failure eventmatrix failure event

11fiber failure eventfiber failure event


22 3342

en2

Slide 42

en2 Color Codeemmett_nelson, 12/15/2008

FINITE

Finite Element ImplementationEELEM

ENTTIM

PLEME= N

TATIO

N

=1 element 11148 elements

Using MCT, a one element model gives the same averaged fiber and matrix stress as a micromechanics model


43

Documents

Composite Failure