A comparative study of spark assisted bending process using teaching–learning based optimization, desirability approach and genetic algorithm

Abstract

The present work deals with the application and comparison of advanced meta-heuristic-based optimization techniques on the micron-thin sheet bending process. Nature-inspired Teaching-Learning Based Optimization Algorithm (TLBO), Genetic Algorithm (GA), and desirability function-based opti-mization techniques have been used to predict the optimal parametric levels for obtaining desired bend angles. Spark discharges were applied to bend sheets using electro-discharge machining. Process parameters, namely, peak current (Pc), duty factor (Df), and gap voltage (Gv), were varied to obtain the response, i.e., bend angle (theta b). Box-Behnken design in Response Surface Methodology (RSM) was used to obtain a regression model. Statistical analysis of the developed model was done using analysis of variance (ANOVA), which showed that theta b was statistically affected by variation in Pc, Df, and Gv at a 95% confidence level. Minimum (theta bmin) and maximum (theta bmax) bend angles obtained from the experiments were reported to be theta bmin = 8.57 degrees and theta bmax = 26.48 degrees at Pc = 6A, Df = 30% and Gv = 40V and Pc = 10A, Df = 50% and Gv = 50V, respectively. Further, developed model adequacy was inspected using standard error design plots and analysis of residuals. The developed quadratic regression model was used to optimize the desired response (theta b). The results revealed that the genetic algorithm provided the desired output corresponding to the requirement of bend angle. The values obtained after the optimization of bend angles by performing a confirmatory test were theta bmin = 8.454 degrees and theta bmax = 28.015 degrees. Hence the values obtained were better concerning the initial practical experimental data set.