SN CURVE APPROACH FOR
Transcription
SN CURVE APPROACH FOR
VALIDATION OF THE MASTER SN CURVE APPROACH FOR SHORT FIBER REINFORCED COMPOSITES Atul Jain, Yasmine Abdin, Stefan Straesser, Wim Van Paepegem, Ignace Verpoest, Stepan Lomov, [email protected] COMPTEST 2015, Madrid Contents Motivation Master SN-curve approach (MSNC) Experimental validation MSNC vs. prevalent approaches Conclusions 2 Short fiber composites Usually injection molded with glass fiber reinforced thermoplastic Different orientations, length distribution at every point Different static and fatigue properties at every point!! σ S ԑ N 3 5/12/2015 Short fiber composite and industry Industrial application of such composites is increasing, esp. in automotive- fatigue is a major issue for this application!! Engine mounts Spare wheel recess Engine cover cam Efficient fatigue simulation can help fully exploit weight reduction potential Goal is to reduce extra testing by proposing hybrid multiscale fatigue methodologies 5/12/2015 Image sources: www.reinforcedplastics.com 4 Multi-scale analysis of short fiber composites Micro-macro analysis: Every point in the Finite Element model treated as an RVE Each RVE is characterized by an orientation distribution of inclusions and a length distribution of inclusion Static and fatigue properties must be calculated efficiently at every point 5 5/12/2015 Multi-scale analysis of short fiber composites Detail of orientation at every point in the component Manufacturing simulation Mapping of meshing Static properties: MoriTanaka formulation FE solver Fatigue properties: ???? Fatigue solver First choice scheme for static properties is full Mori-Tanaka formulation No analytic counter part for fatigue properties 6 5/12/2015 Master SN-curve approach More than one damage event takes place at stress to failure For a certain number of cycles, damage needs to be “big enough” or “spread enough” to propagate and cause failure “How much damage is enough damage to cause cyclic failure?” Damage modeling and quantification of damage is needed 7 5/12/2015 First cycle- micromechanics Fiber matrix debonding, matrix non-linearity and fiber breakage is modeled Fiber matrix debonding Fiber breakage 8 5/12/2015 Master SN curve approach Each data point in SN-curve gives two values o number of cycles to failure (abscissa) o the stress to failure for the number of cycles (ordinate) Value of damage parameter is calculated for the ordinate Same value of damage parameter is assumed for different SN-curves having the same abscissa 9 5/12/2015 Algorithm Read reference SN data – stress level 0and corresponding number of cycles SN-ref Calculate Initial Young’s modulus E0 – MT formulation σ Calculate the modulus and damage parameter at stress level, S1 read from SN curve Dfirstcycle = 1-E1/E0 For second RVE: loading increments is modelled and the modulus and damage parameter is calculated S1 E0 N ԑ S2 E02 Stress at which damage parameter Dfirstcycle is reached, S2 is the stress to failure for initially chosen cycles Dfirstcycle = 1- E1RVE / E0RVE E12 ԑ 5/12/2015 10 Experiments 50% GF reinforced PBT is injection molded into plates Coupons are milled in 3 directions w.r.t. to the matrix flow Fatigue and static tests are performed Frequency = 10Hz R-ratio = 0.1 11 Results of experiments Clear dependence of fatigue properties on orientations is observed 12 13 ANOVA analysis confirms with 90% confidence that the loss of stiffness curves are independent of the FOD 14 MSNC Simulation Good predictions for the SN curve are seen 15 MSNC simulation Good match is observed for the SN curves independent of the input SN curve!! 16 Collected data-1 Data set #1 De Monte et. al. 35% weight fraction GF reinforced polyamide (PA) 17 Collected data-2 Data set #2 Klimkeit et. al. 30% weight fraction GF reinforced PBT Apart from the SN curve the FOD is also provided in both the papers 18 Results ~ De Monte et al. Similar good results are obtained also for the published data 19 Results ~ Klimkeit et al. For the three sets of data considered: The MSNC approach is seen to be independent of the input SN curve Errors increase when lower number of cycles are taken as input 20 MSNC approach and FOD of reference SN curve Error of the proposed scheme is defined as: Group 0-degree No. of samples 30 Mean ± deviation -0.02 ± 0.054 Standard Mean of abs (error) 0.043 45-degree 21 0.06 ± 0.061 0.069 90-degree 21 -0.03 ±0.056 0.049 The magnitude of error in the proposed scheme is independent of the FOD of the reference SN curve This is confirmed further by students T-test with a confidence of 95% 21 MSNC approach vs. UTS based scaling A prevalent practice is to scale the SN curves based on one input SN curve and UTS of the reference and the target RVE This approach would require one SN curve as input and knowledge of UTS at every point MSNC approach is seen to be more accurate than the UTS based scaling! 22 MSNC approach vs. test based interpolation This approach requires many SN curves as input MSNC approach is seen to be more accurate than the test based interpolation! 23 Scaling vs. test based interpolation 1 SN-curve as input is Test based interpolation necessary No specific requirement on orientation of coupon Possible to predict SN-curves with different volume fractions, length distribution etc. Volume-fraction 2, 3 or even more curves are needed as input Coupons need to be orientated in specific directions- testing becomes expensive!! Orientation only parameter on which interpolation could be done Orientation Length distribution State of art methods allow only scaling across “orientation axis” 24 5/12/2015 Conclusions A hybrid method of scaling which involves both experimental and micromechanical analysis was presented The MSNC approach is independent of the FOD of the reference SN curve Only 1 SN curve is needed for such an analysis This could lead to a breakthrough in industrial deployment of composite materials as collection of fatigue data is a bottleneck for simulation The presented methodology has been implemented in industrial software LMS Virtual.Lab Durability part of Siemens Industry Software with partnership with PART Engineering 25 5/12/2015 Future Work Process integration and component level simulation is performed First results are promising!! Critical areas are correctly identified Life time to failure is off by a factor of about 5-8 Multi-axial fatigue and mean stress corrections are studied 26 Thank you! The authors wish to thank the IWT Vlaanderen for funding this research as a part of the project “Fatigue life prediction of random fiber composites using hybrid multi-scale modeling methods” - COMPFAT Baekeland mandate number 100689 27 5/12/2015