The SNL/MSU/DOE Fatigue Program: Composite Testing Trends

John Mandell, Dan Samborsky, David MillerMontana State University

2014 Wind Turbine Blade WorkshopAugust 26th-28th, 2014

Outline• Overview of MSU Fatigue Program on Wind Blade

Materials: Testing and Research• Recent Findings: Resin and fabric structure and biax fabric

issues• Environmental effects for wind and MHK applications• Ongoing work:

– characterization of effects of defects in fatigue

– Subscale testing

– Acoustic emission

Research Group

• PI’s: John Mandell, David Miller• Current Group:

– CoPI Research Engineer: Daniel Samborsky– Grad Students: Tiok Agastra, Michael Schuster, Austin Lolatte, Paul

Murdy– Undergraduate Assistant: Nathan Fritz

• Sandia PI’s: Brian Naughton, Bernadette Hernandez-Sanchez (Environmental Effects)

• Interface with Doug Cairns MSU/Sandia BRC and Manufacturing Studies

Thanks• To Sandia/DOE for long term support• To our many partners in the industry

And….. Looking forward to continued, and expanding relationship.

More details

Contractor reports, database and publications onwww.coe.montana.edu/composites/

additional information including resins, adhesives, fabrics, etc.Contact: dmiller@me.montana.edu.

DOE/MSU Fatigue Database for Wind Blade Materials (Public, Sandia Website)

– Over 250 Materials – 12,000+ test results– Updates each March– Now Excel based– Trends analyzed in contractor reports (www.coe.montana.edu/composites/)

Outline• Overview of MSU Fatigue Program on Wind Blade

Materials: Testing and Research• Recent Findings: Resin and fabric structure and biax fabric

issues• Environmental effects for wind and MHK applications• Ongoing work:

– characterization of effects of defects in fatigue

– Subscale testing

– Acoustic emission

Aligned Strand Composites• PPG-Devold L1200/G50-E07 (MSU Fabric H, 1261 gsm)• Dry wound/infused UD laminate (PPG)• RodPack pre-cured rods

Front Back

Aligned Strand

Aligned Strand (AS) vs UD Fabric H (02) Fatigue data, Three Resins

• AS laminates fabricated by PPG/Reichhold by dry strand winding/infusion;

• same strands and resins as in the fabrics.

• Aligned strand laminates higher Vf, stronger, significantly more fatigue resistant compared to UD fabrics)

Neptco RodPack Fatigue data compared to UD fabric H/epoxy (baseline laminates)

  RodPack Baseline LaminateDirection L T L TE, GPa 48.4 18.7 40.5 12.8UTS, MPa 1174 32.2 974 56.6Ultimate tensile strain, % 2.5 0.17 2.5 0.36 / 1.6*UCS, MPa -986 -141 -706 -161Ultimate compressive strain, % -1.9 -1.0 -1.7 -1.3*Transverse strain to first cracking / strain at failure; due to fabric 0’s presence.

Fatigue of Biax Laminates

• Creep effects in tensile fatigue at tested stress levels.

• Significant strain accumulation and stiffness decrease for R = 0.1

• Biax fabric P: PPG-Devold DB810-E05, with epoxy 135/1366, Vf = 55%

• Failure comparison with two other biax fabrics

Biax Fabrics

Transmitted light images of dry fabrics L, M and P

Failed fatigue coupons of fabrics L(SWA), M(DH) and P (DH3) (bottom).

Typical tensile and compressive stress-strain curves for biax fabrics; L , M, and P in the warp direction.

• Nonlinear, rate sensitive stress-strain response. • In reversed loading, residual stress-strain curves show lower initial modulus but higher

stresses at high strain (shear dominated, but compression curves higher due to transverse compressive stress)

• Left, coupon tested at R = 0 (0-tension), max stress 60 MPa, stress-strain loops with residual tensile stress-strain test immediately after cycle 10,000; shows significant stiffness loss and permanent creep.

• Right, compression loops, R = 10, max abs stress 96.5 MPa, with residual compressive stress-strain curve after cycle 8,246. Less stiffness change than tension.

Stress-Strain Loops, creep and Residual Stress-Strain

0-Tension

Comp., R = 10

Cumulative strain increase

• Maximum running strain over lifetime at different max loads: R = 0.1 (top left), 10 (top right) and -1 (bottom: left, tension part of cycle; right, compression part).

Cumulative strain increase criterionMaximum running strain vs cyclesat various stress levels, R = 0.1

Damage accumulates and strain increases rapidly after the max strain increases by 50% over the first cycle strain (dashed line)

Similar for all R-values

• The 50% max strain metric is selected since it approximates the point of rapid upturn on the max strain plots and can be determined for all R-values

Damage Metric Comparisons

• Total fatigue failure results are compared to cycles for • 25% decrease in stiffness and • 50% increase in cumulative max abs strain.

• “Total failure” is poorly defined except in tension.

• Hysteretic heating becomes significant in reversed loading beyond 50% strain increase even at 1 Hz frequency

Damage Metric Comparisons So, what identifies a good failure criterion?

50% strain increase (occurs first in most cases)25% stiffness decrease

Constant Life Diagrams

Based on 50% cumulative strain increase (creep data for 0-amplitude, R = 1.0, considered separately)Cycle Based…

Tensile and compressive creep data for use as R approaches 1.0, zero amplitude.

CLD’s give the lifetime under constant amplitude loadingat any mean stress and stress amplitude. At the extremes, as the amplitude approaches zero, creep data should be substituted (using an assumed frequency).

For Spectrum loading of blades, the CLD must be combined with a known loads spectrum for the turbine and blade, and a validated cumulative damage law for lifetime prediction under variable amplitude loading. While this methodology has been demonstrated for fiber dominated failure modes, additional work is required for this cumulative strain increase, time under load based approach to resin dominated failure.

CLD’s and Design

In tensile and compressive fatigue (R = 0.1 and 10) the cumulative (sine-wave) test time to 50% strain increase correlates with the corresponding value from creep tests at the same max stress [1].

In reversed loading (R = -1), square waveforms are easier to interpretfor cumulative time under tension and compression.

Square Wave Results: Separating Creep and Cycling Effects in Reversed Loading (R = -1, ±37.9 MPa, 0.01) Hz)

ActuatorPosition

ExtensometerStrain

Creep on both tensile and compressive parts of cycle; somewhat greater creep on tensile side.

Plotted as cycles, thestrain curves separate according to frequency

Square wave, R = -1,±37.9 MPa

Strain vs cumulative time in tension and compression,square wave and creep, various frequencies, ±37.9 MPa

Within scatter of abouta decade, similar cumulative times to reach point of high strain increase (~0.75% strain)except for pure creep,which remains stable.

Strains follow creep curves, then deviate higher at some time, (unlike R = 0.1 and 10). The timescale is consistent for all frequencies.

Conclusions: S-N Fatigue and Residual Properties

1. The cumulative strain criterion for a 50% strain increase is most generally applicable to all seven R-values as well as creep, for the resin-dominated laminates.

2. A constant life diagram has been assembled based on the cumulativestrain criterion. Its application to wind blade design requires additional testingto establish a consistent cumulative damage approach for spectrum loading.

3. Residual stress-strain data show significantly reduced stiffness only for R-values with a significant tensile component.

4. Cyclic stress-strain loops show large accumulating creep for tensile and compressive fatigue, but remain centered close to the origin in fully reversed loading.

1. Tensile and compressive fatigue data trends are consistent with creep data to beyond the 50% cumulative strain increase condition.

2. Under reversed loading, relative to simple creep, stress reversals even atvery low frequency accelerate the onset of damage to a cumulative time under load range which is frequency and cycle independent, for a particularload level.

3. Square wave strain data follow the creep trend with time until resin damage develops; the strains then increase rapidly relative to creep strains at the same max stress.

4. The controlling parameter in determining the fatigue lifetime for off-axis laminates is cumulative time under load, not stress cycles, for the frequency range 0.001 to 0.5 Hz. Data also agree with sine wave data at 0.5 to 2 Hz, above which hysteretic heating is prevalent for the thick coupons.

Conclusions: Creep/Fatigue Interaction

Outline• Overview of MSU Fatigue Program on Wind Blade

Materials: Testing and Research• Recent Findings: Resin and fabric structure and biax fabric

issues• Environmental effects for wind and MHK applications• Ongoing work:

– characterization of effects of defects in fatigue

– Subscale testing

– Acoustic emission

• To cultivate a successful industry it becomes pertinent to develop a comprehensive understanding of immersed MHK structures

• Well documented that composite materials absorb moisture– Significant mechanical and physical degradation– Primarily unstressed systems investigated

• Structure will be subjected to stresses– Becomes vital to understand what effects these

stresses have on the moisture absorption process in composite material systems

Effects of Tensile Stress on the Moisture Diffusion Characteristics of Epoxy Glass Composites

Problem Definition

• Seek to fully characterize the effects of tensile stresses on the moisture diffusion characteristics of Epoxy Glass composites

– To gain a clear understanding of the mechanisms at work the effects of varying both fiber angle and magnitude of applied stress will be investigated

0 200 400 600 800 1000 12000.0%

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%

1.4%0 MPa

(0)2(20)2(45)2(90)2

Time Soaked (hours)

% W

eigh

t Gai

n

0 200 400 600 800 1000 12000.0%

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%18 MPa

(0)2(20)2(45)2(90)2Control

Time Soaked (hours)

% W

eigh

t Gai

n

0 200 400 600 800 1000 12000.0%

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%

1.4%30 MPa

(0)2(20)2(45)2(90)2Control

Time Soaked (hours)

% W

eigh

t Gai

n

0 200 400 600 800 1000 12000.0%

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%

1.4%

Time Soaked (hours)

% W

eigh

t Gai

n

Summary• Began with moisture absorption of composite

materials, Springer (1976).– D1,2,3, Dx,y,z, and D for unstressed composite plate

• Free volume theories to describe diffusion in polymers– Free volume changes Changes in diffusion parameters

– Neumann (1986): – Hurt (1980):

Continued…• Laminate Plate Theory to calculate volume change of

only the polymer matrix

• All input parameters are known quantities:– Stress (σx), fiber angle (θ), volume fraction (φ), densities of

fluid and matrix (ρ), and elastic properties of the constituents (E and ν for composite and fibers).

Finite Element Analysis• ANSYS 13.0 – strong time dependent analysis tools• Thermal-Moisture Diffusion Analogy as presented by

Wong and Koh (2002)– Fourier Heat diffusion Fickian Mass Diffusion

Property Thermal MoistureField Variable Temperature, T Saturation Ratio, w

Density ρ (kg/m3) 1Conductivity k (W/m °C) D × M∞ (mm2/hr)

Specific Capacity c (J/kg °C) M∞

𝜕C𝜕 t =D ( 𝜕2𝐶

𝜕𝑥2 + 𝜕2𝐶𝜕𝑦2 +𝜕

2𝐶𝜕𝑧2 )

𝜕T𝜕 t =

𝑘𝜌𝑐 ( 𝜕2𝑇

𝜕 𝑥2 + 𝜕2𝑇𝜕𝑦2 + 𝜕

2𝑇𝜕 𝑧2 )

Maximum Moisture Content    σx M¥ (%) Percent Error (%)

θ(deg) ϕf (MPa) Experimental ANSYS Theory ANSYS Theory

0 0.520 0.9692 1.0652 1.0652 9.91 9.91

18 0.9453 1.0703 1.0676 13.22 12.9430 0.9758 1.072 1.0718 9.86 9.84

20 0.52

0 0.9466 1.0651 1.0652 12.52 12.5318 1.0235 1.0773 1.0776 5.26 5.2930 1.151 1.085 1.0852 -5.73 -5.72

45 0.52

0 0.9559 1.0652 1.0652 11.43 11.4318 1.0644 1.1031 1.1027 3.64 3.60

30 1.2523** 1.1354 1.1349 -9.33 -9.37

90 0.52

0 1.0102 1.0652 1.0652 5.44 5.4418 1.1246 1.1363 1.1358 1.04 1.00

30 1.4057** 1.1836 1.1829 -15.80 -15.85

** Sample fracture prior to achieving full saturation

ANSYS and Model:

    σx D (mm2/hour) * 10-2 Percent Error (%)θ(deg) ϕf (MPa) Experimental ANSYS Theory ANSYS Theory

0 0.520 0.1073 0.1046 0.1076 -2.52 0.2818 0.1156 0.1118 0.1075 -3.29 -7.0130 0.112 0.1132 0.1074 1.07 -4.11

20 0.520 0.125 0.1197 0.1134 -4.24 -9.2818 0.1374 0.1296 0.1366 -5.68 -0.5830 0.1813 0.1619 0.1559 -10.70 -14.01

45 0.52

0 0.1237 0.1187 0.1211 -4.04 -2.1018 0.1444 0.1429 0.1482 -1.04 2.6330 0.1911 0.1691 0.1743 -11.51 -8.79

90 0.52

0 0.1195 0.1151 0.1177 -3.68 -1.5118 0.1705 0.1631 0.1699 -4.34 -0.35

30 0.2132 0.1977 0.1987 -7.27 -6.80

Diffusivity

Conclusions• The model successfully predicts maximum moisture

content and diffusivity values for stressed unidirectional composite samples.

• The model uses commonly known composite input parameters (σx, θ, φ, ρ, E, ν) in addition to neat resin properties D and M¥

• ANSYS FEA code has shown very good agreement with experimental data, validates thermal-moisture diffusion analogy

Outline• Overview of MSU Fatigue Program on Wind Blade

Materials: Testing and Research• Recent Findings: Resin and fabric structure and biax fabric

issues• Environmental effects for wind and MHK applications• Ongoing work:

– characterization of defects in fatigue

– Subscale testing

– Acoustic emission

Characterization of effects of defects in fatigue

• Initial Results

Coupon T-B-3-2145, R=0.1, 103/10.3 MPa, 12750 cycles, 50 mm widthAverage waves: amplitude = 6 mm, wavelength = 25 mm, wave severity = 0.24

Characterization of defects in fatigueMaximum applied tensile stress vs cycles to failure for unidirectional (UD)

control laminates, UD laminates with waviness flaws, and ±45 control laminates, tensile fatigue at R = 0.1.

Substructure Test FrameDesign Specifications

Test frame description led to the following list of design specifications:• Load Capabilities

• 525 kN-m in cantilevered loading• 175 kN-m in 4 pt bending• 50 kN-m in torsion

• Deflection Capabilities• Max specimen displacement in bending: 10”• Max angle of twist: 30 degrees

• Fatigue testing• Test frequency: 5 Hz

• Measuring area 300x250mm• Camera distance was 192cm

• much further than has been normally used by us at MSU

• FEA to Exp load-displacement compares very well

• Images shown at 32000lbf and 0.32in displacement – final failure occurred several seconds later.

Strain in the y (vertical) direction-Good correlation FEA to DIC.

C-channel Beam Manufacturing• Develop in-house beams for baseline static and fatigue

characterization• Impart flaws for sub-structure effects.

Completed Box Beam

Projected Beam Testing• This testing will involve testing the built up box beams with in-

plane (IP) waves and comparing them to unflawed box beams.

• This will also be modeled in Abaqus, building off of Dr. Nelson’s coupon models.

• A preliminary test matrix for this testing in development.

IP wave on surface of wind turbine blade

Acoustic Emission Technology

• Based on piezoelectric technology• Elastic waves traveling through a

material excite the piezo• Produces a voltage waveform• Recording equipment consists of:

– Amplifiers– High/Low Bandpass Filters– High bitrate A/D converter– To software– Software filters– Display

• Sensors can have different ranges and tuned to specific frequencies

Frequency Domain Features

AE Waveform: Time Domain AE Waveform: Frequency Domain

•Fast Fourier Transform performed by AEWin software.•Highest magnitude in the frequency domain is Peak Frequency (P-FRQ)

Application to Composites

•Can divide frequency spectrum into four “bins”•Identified as particular failure mechanisms

•Division per waveform• Partial Powers Analysis

•Division per full test• Frequency Distribution

Peak Frequency Bin RangesBin Freq Range Identified MechanismF1 0-120kHz Matrix CrackingF2 120-200kHz Fiber slip/pulloutF3 200-300kHz Fiber/Matrix DebondF4 300kHz + Fiber Break

0 0.5 1 1.5 2 2.50

50

100

150

200

250

300

350

P-FRQ Bin Ranges

Percent Strain

P-FR

Q (k

Hz)

Test CouponManufacture

• Plates manufactured w/ VARTM process

– Uni plate for all materials– Biax plate for materials A,B and D

• Dimensions: 300mm X 30mm• Coupon edges and sensor locs sanded• High-load coupons tabbed with G10• Positioning marks for AE attachment• Designed to fail within limits of

electrically driven, quieter load frame

Static & LUR

Materials

L1200/G50 ELT5500 EBX0900 CLA2012

Layup

[0]4 [0]2 [45's]4 [0]2

[90]4 [90]2 [90]2

[90/0]s [90/0]s [90/0]s

Total 60

Test Progression and Characterization• Glass-A [0]n Static Test• Things to note:

• First hit strain• P-FRQ changes

• First hits mid freq• Low freq hits at 1.5%• Clusters of high freq late

• Hit Energy • Hits before 1.5% minor

• Accumulated Energy• Sharp uptake 1.6%• Nearly 0 increase after• Then steady

0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.40

50100150200250300350400450500

Hit Peak Frequency

Percent Strain

P-FR

Q (k

Hz)

0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.41.00E+01

1.00E+02

1.00E+03

1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

Hit Absolute Energy

Percent Strain

Abs.E

(aJ)

0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.40.00E+002.00E+004.00E+006.00E+008.00E+001.00E+011.20E+01

Accumulated Absolute Energy

Percent Strain

AbsE

(aJ)

LUR Test Progression• P-FRQ progression is same as in

static, just discrete steps• No AE activity during 1st cycle• Sensors removed after 90%• Noise present upon unload• Appears to be fewer hits but just

grouped together• Coupons generally failed in the

110% to 120% cycle

0 100 200 300 400 500 600 700 800 900 10000

2

4

6

8

10

12

0102030405060708090

Hit Peak Frequency for [90/0]s Carbon-D

Time (s)

P-FR

Q (k

Hz)

Load

(kN)

0 100 200 300 400 500 600 700 800 900 10001.0E+01

1.0E+02

0102030405060708090

Absolute Energy for [90/0]s Carbon-D

Abs Ener Acc. Abs.E Load

Time (s)

Abso

lute

Ene

rgy

(aJ)

Load

(kN

)

Questions?