Toughness prediction in functionally graded Al6061/SiCp composites produced by roll-bonding

Abstract Functionally graded aluminum matrix composites (FGAMC) are new advanced materials with promising applications due to their unique characteristics in which composite nature is combined with graded structure. Different architectures of Al6061/SiC p composite laminates were fabricated by successive hot roll-bonding. For FGAMCs, two composite layers as outer strips and a layer of Al1050 as interlayer were applied. To investigate laminate toughness, the quasi-static three-point bending test was conducted in the crack divider orientation. Genetic programming as a soft computing technique was implemented to find mathematical correlations between architectural parameters and experimentally obtained results. Doing this, reliable data were achieved and randomly divided into 63 training sets and 12 testing sets. Input parameters of “specimen thickness”, “SiC content in composite layers”, “rolling strain”, “maximum achieved stress”, and “strain in utmost stress”, as well as, “total strain of specimen” were utilized for modeling. Models of different types each used 8 independent parameters, wholly and/or partially, were proposed. In order to introduce the best model, four criteria of “higher regression”, “minimum errors”, “minimum subtraction of errors summation” and “minimum sum of absolute subtractions of predicted and experimented toughness” were considered. Based on different evaluations, a model which used 5 inputs, as well as, 30 chromosomes, 8 head sizes, 2 genes, addition as linking function, RRSE as fitness function while implemented 5 different mathematical operators was selected. The study showed that predicted results and experimented data had good conformity while all models had reliable applicability for predicting fracture tolerance of aluminum matrix composites with graded structures under quasi-static bending condition.