Optimization of wire-EDM process parameters for Ni–Ti-Hf shape memory alloy through particle swarm optimization and CNN-based SEM-image classification

Abstract

Shape memory alloys (SMAs) have the unique ability to regain their shape under specified conditions, making them extremely useful for a wide range of applications. However, non-traditional machining techniques could be more effective for high-temperature SMAs as they provide better shape recovery properties than conventional methods, necessitating this study's unconventional wire electric discharge machining procedure. The surface morphology of Ni–Ti-Hf-based alloys was examined using scanning electron microscopy. A convolutional neural network model was used to categorize SEM images based on the material removal rate, utilizing the pixel intensity histogram of the processed images. The study discovered that the material remelted as the discharge energy increased, leaving lumps and globules on the surface. The size of debris lumps, pores, and globules increased with increasing Ra values. Moreover, widening the inter-electrode gap by expanding the servo-voltage facilitated efficiently removing debris from the machined site. The TOPSIS method for particle swarm optimization was employed to find the optimal solutions, yielding TON = (124.239, 116.228), TOFF = (54.532, 40.781), Servo voltage = (36.216, 43.766), and Wire-Feed = (2.157, 8). These results were validated through experimentation, with minimal error percentages of 2.01%, 2.046% (MRR), and 3.86%, 1.611% (Ra) for both sets of input parameters, respectively.