Soft computing models and intelligent optimization system in electro-discharge machining of SiC/Al composites

In this paper, a multi-variable regression model, a back propagation neural network (BPNN) and a radial basis neural network (RBNN) have been utilized to correlate the cutting parameters and the performance while electro-discharge machining (EDM) of SiC/Al composites. The four cutting parameters are peak current (I_p), pulse-on time (T_on), pulse-off time (T_off), and servo voltage (S_v); the performance measures are material remove rate (MRR) and surface roughness (Ra). By testing a large number of BPNN architectures, 4-5-1 and 4-7-1 have been found to be the optimal one for MRR and Ra, respectively; and it can predict them with 10.61 % overall mean prediction error. As for RBNN architectures, it can predict them with 12.77 % overall mean prediction error. The multivariable regression model yields an overall mean prediction error of 13.93 %. All of these three models have been used to study the effect of input parameters on the material remove rate and surface roughness, and finally to optimize them with genetic algorithm (GA) and desirability function. Then, an intelligent optimization system with graphical user interface (GUI) has been built based on these multi-optimization techniques, in which users can obtain the optimized cutting parameters under the desired surface roughness (Ra).