This paper reports on performance enhancement of electrochemical honing (ECH) through its multi-performance optimisation for finishing the bevel gears. Implicit nature of ECH, complex interactions among its process parameters and conflicting objectives makes the multi-performance optimisation very challenging. In such cases, soft-computing tools such as artificial neural network (ANN) are found to be very effective. In this work, multi-performance optimisation of five important ECH parameters (namely concentration, temperature and flow rate of the electrolyte, rotary speed of the workpiece gear and voltage) was performed by developing back-propagation neural network (BPNN) and multiple regression models of percentage improvement in average and maximum surface roughness and MRR. The experiments were conducted using Taguchi's L27(313) orthogonal array to generate the experimental data required for the models. The optimised results were validated through confirmation experiments. The BPNN-based models were found very effective and superior to regression models for the multi-performance optimisation. Use of optimal values of the ECH parameters yielded better surface finish and productivity of the gears.
|