SPC without Control Limits and Normality Assumption: A New Method
Control Charts (CC) are important Statistic Process Control (SPC) tools developed in the 20’s to control and improve the quality of industrial production. The use of CC requires visual inspection and human judgement to diagnoses the process quality properly. CC assume normal distribution in the observed variables to establish the control limits. However, this is a requirement difficult to meet in practice since skewness distributions are commonly observed. In this research, a novel method that neither requires control limits nor data normality is presented. The core of the method is based on the FuzzyARTMAP (FAM) Artificial Neural Network (ANN) that learns special and non-special patterns of variation and whose internal parameters are determined through experimental design to increase its efficiency. The proposed method was implemented successfully in a manufacturing process determining the statistical control state that validate our method.
KeywordsControl Charts Neural Networks Pattern Recognition
- 5.Guh, R.S.: Real-time pattern recognition in statistical process control: a hybrid neural network/decision tree-based approach. IMechE. J. Engineering Manufacture 219, Part B (2005)Google Scholar
- 6.Al-Hindi, H.A.: Control Chart Interpretation Using Fuzzy ARTMAP. Journal of King Saud University. Engineering Sciences 16(1) (2004)Google Scholar
- 11.Vazquez-Lopez, J.A., Lopez-Juarez, I., Peña-Cabrera, M.: Experimental Design applied to the FuzzyARTMAP neural network to the statistical special pattern recognition. In: XVIII Congress of the Automatic Control Chilean Association-IFAC-IEEE (2008) (in spanish)Google Scholar