Probabilistic Finite Element Analysis and Size Effects of Polymer Matrix Marine Grade Composites

This thesis is divided into two parts. In the first part, the variability of tensile and compressive mechanical properties of a polymer matrix composite material with woven fabric reinforcement is studied using both experimental work and numerical simulations. Eight (1.2 m x 1.8 m x 4.9 mm) E-glass / vinyl ester composite plates were fabricated using the Vacuum-Assisted Resin Transfer Molding (VARTM) by a US Navy contractor. The materials and process selected are representative of marine grade composites typically used by the US Navy. Modified ASTM D3039 tensile coupons and standard ASTM D6641 compression coupons were obtained from the plates and the laboratory results were compared with those of a 3D probabilistic finite element analysis (FEA). In the probabilistic FEA models, elastic properties (En,E22,Gn,vn), strength parameters (Flt,Flc,F2t,F2c,F6, £lt,£lc,e2l,e2c,sl2) and geometric properties (thickness, fiber misalignment) of woven fabric E-glass / vinyl ester coupons were considered as random fields, and generated using Monte Carlo simulations. These material properties were obtained from laboratory test data of [0]4Sf lay-ups, and in-turn used to predict the properties of the [0/90]2Sf lay-ups. The study evaluates the effects of spatial correlation, finite element size, probability distribution functions (PDF) and failure criteria on statistical strength properties of the [0/90]2Sf tension and compression coupons. Comparisons of experimental and probabilistic FEA results provide useful observations on how to assign mean, COV and PDF of material properties to individual finite elements within a mesh. The second part investigates the size effects on bending strength of marine grade polymer matrix composites (PMC) using both probabilistic finite element analysis and laboratory testing. Woven [0/45/-45/0]5Sf E-glass / vinyl ester material specimens with different sizes were evaluated to predict the effect of specimen size on the bending strength. In the probabilistic FEA models, elastic properties (-c,ii\u27-c,22\u27VJi2\u27v/i2)? strength parameters (Fu\u3eFic\u3eF2t\u3eF2c\u3eF6^ £i,\u3e£ic\u3e£2r\u3e£2c\u3e£i2) a^ geometric properties (thickness, fiber misalignment) of the woven fabric E-glass / vinyl ester ASTM D6272 coupons were considered as random fields, and generated using Monte Carlo simulations. These material properties were obtained from laboratory test data of [0]4Sf lay-ups. The relationship between the bending strength and specimen size for the [0/45/-45/0]ssf lay-up was also evaluated experimentally. 32 material specimens were prepared and tested according to ASTM D6272 using four different coupon\u27s size. The probabilistic FEA model bending strength predictions are compared with both the Weibull weakest link theory and the experimental results. Finally a new equation was developed to accurately evaluate the size effect on the bending stress