Application Of Metaheuristics To The Training Of Artificial Neural Network For Control Chart Pattern Recognition

Abstract

Quality management particularly and total quality management in general have always been a major concern on various perspectives of industrial engineering. In the 4.0 era, quality control process has become one of the most crucial issues in intelligent manufacturing. The purpose of the dissertation is to present the application of metaheuristics to the training of artificial neural networks for control chart pattern recognition. This dissertation not only researches the accuracy of training pattern recognizers but also figures out the lack of requirements to fulfill the expected demand that the application of metaheuristics entail. As the most pratical and effective tools for continously monitoring, control chart patterns (CCPs) can be intransitively recognized to determine the defect of quality control process. Therefore, this dissertation implies the implementation of artificial neural network (ANN) network for recognizing patterns in control chart. The ANN network were trained by deploying an advanced optimization algorithms. The algorithms which can be Genetic Algorithm (GA) or Bees Algorithm (BA). This dissertation first presents the algorithms and explains how the algorithms are deployed to the ANN network. It then compares the accuracy of control chart pattern by ANN networks optimized practicing those algorithms and concludes the best technique.