Strength modeling and optimizing ultrasonic welded parts of ABS-PMMA using artificial intelligence methods

This study analyzes variations in metal removal rate (MRR) and quality performance of roughness average (R _a) and corner deviation (CD) depending on parameters of wire electrical discharge machining (WEDM) process in relation to the cutting of pure tungsten profiles. A hybrid method including response surface methodology (RSM) and back-propagation neural network (BPNN) integrated simulated annealing algorithm (SAA) were proposed to determine an optimal parameter setting. The results of 18 experimental runs via a Taguchi orthogonal table were utilized to train the BPNN to predict the MRR, R _a, and CD properties. Simultaneously, RSM and SAA approaches were individually applied to search for an optimal setting. In addition, analysis of variance was implemented to identify significant factors for the processing parameters. Furthermore, the field-emission scanning electron microscope images show that a lot of built-edge layers were presented on the finishing surface after the WEDM process. Finally, the optimized result of BPNN with integrated SAA was compared with that obtained by an RSM approach. Comparisons of the results of the algorithms and confirmation experiments show that both RSM and BPNN/SAA methods are effective tools for the optimization of parameters in WEDM process.