Modelling and optimisation of hardness in citrate stabilised electroless nickel boron (ENi-B) coatings using back propagation neural network – Box Behnken design and simulated annealing – genetic algorithm

Abstract

ABSTRACT In this work, a novel citrate stabilised electroless bath was developed and the process parameters (concentrations of nickel, reducing agent, and stabiliser) were optimised to achieve the maximum hardness in the ENi-B deposit on 7075-T6 aluminium alloy, using the back-propagation neural network (BPNN), Box–Behnken design (BBD), simulated annealing (SA) and genetic algorithm (GA). The effect of independent variables on dependent variable was modelled using the BPNN and BBD. The models were assessed for their significance using the coefficient of determination (R2) and mean squared error (MSE). The MSE and R2 of 34.18 and 0.9852 were obtained for BPNN model against 20.48 and 0.9911 for BBD, which proved that the BBD fits well to the experimental data. The optimum nickel ion, reducing agent and stabiliser concentrations of 29.86, 0.77 and 30.92 g L−1 were obtained from BBD for the maximum hardness of 592 HV. The local optimum values obtained from BBD were compared with global optimisation techniques, SA and GA, and the values were validated through experiments carried out in triplicate. The maximum hardness from local and global optimisation techniques was identical, with negligible change in the values of optimised process parameters. X-ray diffraction and scanning electron microscopy methods were used to examine the elemental composition and surface morphology, respectively, before and after heat treatment.