Neural network prediction of mechanical properties of porous NiTi shape memory alloy

A multilayer back propagation learning algorithm was used as an artificial neural network tool to predict the mechanical properties of porous NiTi shape memory alloys fabricated by press/sintering of the mixed powders. Effects of green porosity, sintering time and the ratio of the average Ti to Ni particle sizes on properties of the product were investigated. Hardness and tensile strength of the compacts were determined by hardness Rockwell B method and shear punch test. Three-fourths of 36 pairs of experimental data were used for training the network within the toolbox of the MATLAB software. Porosity, sintering time and particle size ratios were defined as the input variables of the model. Ultimate strength and hardness were the outputs of the model. Results indicated that seven neurons in the hidden layer yielded the minimum normal error. The modelling outcomes confirmed the feasibility of the model and its good correlation with the experimental information.