Validation of a method using Taguchi, response surface, neural network, and genetic algorithm

Abstract

Piston is one of the important parts for aircraft engine, and the quality of piston affects the efficiency and safety of the engine. This study applies Taguchi method, response surface methodology (RSM), and back-propagation neural networks (BPNN) combining with genetic algorithm (GA) on the quality improvement of piston manufacturing processes to enhance the process yield. The Taguchi parameter design concerns three nominal-the-best specifications, including ring groove diameter specification, inner groove diameter specification, and inner diameter of pistons. Together with five control factors consisting of (1) type of carbon steel, (2) type of cutting fluid, (3) cutting depth, (4) spindle speed, and (5) chuck pressure, the L27(313) orthogonal array was selected for this experiment. Three models: (1) Taguchi model, (2) Taguchi_RSM model, and (3) Taguchi_BPNN_GA model were constructed to find the parameter combinations of five control factors for each model. Confirmation experiments were done for each model and the performances of three models were also compared to indict the enhancement of manufacturing quality of piston.