Development of statistical and soft computing regression models for WEDM machining of aluminum composites

Producing low-cost, high-quality products are the primary aim of machining. To accomplish this appropriate choice of machining parameters is a challenging issue. This paper compares three feature learning techniques to investigate their performances for predicting manufacturing quality using regression model development for predicting machining parameters. Wire-cut electrical discharge machining is carried on 9% activated carbon reinforced Al7075 aluminum composites using a response surface design matrix. Data from experimentation is used for the training of the proposed models. The proposed general regression neural network, support vector regression model and relevance regression model are developed, trained and tested with the data from experimentation. From the developed models, it has been found that the GRNN model has shown an MAPE 2.28% and an R-value of 0.99 for all data and 5.622% for SVM model and 8.217% form RVM model, indicating excellent prediction capability of the developed model. From the ANOVA analysis, the parameter that has the most significant effect on MRR is the discharge current and servo speed is the significant effect on surface roughness. From interaction study it is found that R² value for SR and MRR are 0.99 and 0.98 respectively the developed models. The developed models empirical relationships effectively predicted MRR and SR values, with average prediction errors of 3.84% and 4.21%, respectively. This methodology will significantly increase the efficiency of decision making and offers a solution for cutting intricate shapes that may pose challenges for traditional cutting tools with high accuracy in manufacturing industries.