Establishment and application of high-precision A286 superalloy constitutive model considering initial hardness

Compression specimens of A286 superalloy with different hardness were obtained by different annealing processes. A286 superalloy specimens of different hardness were compressed by the universal testing machine. Based on the true stress-true strain curve under different initial hardness conditions, a BP neural network (BP) with Vickers hardness and strain as input parameters and stress as output parameter and a BP neural network model optimized by genetic algorithm (GABP) were established, respectively. A finite element model of axial compression of a thin-walled tube with inhomogeneous hardness distribution of A286 superalloy was established. The flow behavior of the thin-walled tube axially compressed into a bulge was analyzed. The results indicated that the absolute average relative error (AARE) and correlation coefficient (R) of the GABP model were 0.52% and 0.997, and 2.5% and 0.996 for the BP model, by the experimental data with initial hardness of 286 HV and 235 HV, respectively. The prediction accuracy of the GABP constitutive model is higher than that of the BP model at different initial hardness. The C1 region of the bulge has the densest metal flow lines, the most significant stress concentration, and the greatest degree of plastic deformation.