Optimization of Kerf Width in WEDM of Sandwich Woven CFRP-An Ensemble Machine Learning Based Approach

Abstract

Machining CFRP with WEDM is extremely challenging and produces kerf of poor quality. Therefore, the present research venture is intended to improve the kerf quality produced in WEDM of woven CFRP through a machine learning-based metaheuristic algorithm. Two ensemble-based machine learning algorithms i.e., the Random Forest (RF), and Adaptive Boosting algorithm (AdaBoost) have been used to model the kerf width. The performance of RF is found to be superior to AdaBoost in terms of generalization prowess as the box plot corresponding to the predicted KW by RF closely resembles the box plot of experimental KW whereas the box plot corresponding to the predicted KW by AdaBoost has a varying distribution with the box-plot of experimental KW. Furthermore, the kerf width optimization has been conducted using a broad range of optimization techniques from nature-inspired to mathematically driven approaches such as the Moth flame optimizer (MFO), Grey Wolf optimizer, Chimp optimization algorithm, and sine cosine algorithm in an attempt to compare the computational performance of the algorithms. It has been revealed that MFO discovered the minimum KW (global optimum solution) and exhibited rapid convergence as compared to its counterparts. The optimal results are Ton = 26 microsecs, Toff = 50 microsecs, I = 7A, and V = 70 V. Additionally, the proposed optimization's durability has been examined using the traditional desirability approach. The percentage improvement in KW through the proposed optimization as compared to the desirability approach is 5.6%. Lastly, FESEM images are provided for varying process parametric conditions.