Pattern identification in statistical process control using fuzzy neural networks

Today customers are demanding more diversified products, with higher quality and shorter delivery times. It has led to the adoption of flexible manufacturing systems (FMS). The quality control of the manufacturing process in FMS is a critical factor, requiring flexible and intelligent quality control systems. Quality control windows (QCW) is an adequate framework for development of automated quality control systems. QCW is formed by five steps: observation, evaluation, diagnostic, decision and implementation. The most important step is evaluation, where quality control charts indicate out of control situations and possible underlying causes. This work studied the performance of three fuzzy neural networks (radial basis functions network, RBF, fuzzy artmap, and a new network, RBF fuzzy-artmap) in the identification of six different "patterns" in quality control charts. The new RBF fuzzy-artmap network presented the best performance of classification (78.8%), while allowing on-line incremental learning.