Using Artificial Neural Networks for Recognition of Control Chart Pattern

Abstract

Control charting is an important tool in SPC to improve the quality of products. Unnatural patterns in control charts assume that an assignable cause affecting the process is present and some actions must be applied to overcome the problem. By its automatic and fast recognition ability the neural network provide best performance to immediately recognize process trends. In this paper, a neural network model is used to control chart pattern recognition (CCPR). Several forms of architectures have been tested and the results point out a network configuration which leads to excellent quality of recognition. General Terms SPC = Statistical Process Control; CCPR = Control charts pattern recognition; CCP = Control charts pattern; CC = Control charts; NOR = Normal; IT = Increasing trend; US = Upward shift; ANN(s) = Artificial Neural Network(s); MLP = Multilayer Perceptron; LM = Levenberg -Marquardt back propagation algorithm; GDA = Gradient descent with adaptive learning rate back propagation; SSE = Error Sum of Squares; MSE = Mean Square Error; Keywords Artificial Neural Networks (ANN), Statistical Process Control (SPC), Control Charts (CC), Control Charts Pattern (CCP).