Implementation of Neural Network in Classifying the Product Quality for the ABS Metallization

Abstract

The surface treatment conditions of a plastic surface are related to the quality of finished products. Usually, more than 20 causes dominate the success of electroplating for acrylonitrile butadiene styrene (ABS). Thus, the quality control is very complicated and challenging. Even nowadays, most of the production quality still relies on the operator's experience and intuition. This research takes a company of water hardware in Taiwan as the research object. We propose a revolutionary concept of quality management, combining artificial intelligence and surface treatment process altogether. We set up a parameter monitoring system during production to predict the quality of ABS metallization using neural network models such as artificial intelligence forms the basis of the intelligent manufacturing system. It can be used as a quality control tool to improve quality yield and industrial competitiveness. Totally 13 operational parameters (causes) and one quality parameter (consequence) of the electroplating tanks were collected from time to time to build the NN models. Interestingly, we finally find the fuzzy NN model performs better than the precise NN model. We conclude this is resulting from the limitation and vagueness of data.