Multi-Resolution Wavelets Analysis Approach for the Recognition of Concurrent Control Chart Patterns

Abstract

Process control charts have been commonly used as a tool to achieve and maintain acceptable process quality control. The use of Artificial Neural Networks (ANN) for control charts pattern recognition has been studied in recent years. Most of the research has focused on the recognition of single trend, shift, or cyclic patterns, and used simple coding techniques to process the observed data for the neural network. In this article, Multi-Resolution Wavelets Analysis (MRWA) and ANN were used to recognize single and multiple concurrent patterns of control charts. MRWA extracted distinct features for the patterns by providing distinct time-frequency coefficients. A reduced set of parameters was derived from these coefficients and used as input to an ANN classifier. Results have shown that the approach produced adequate recognition of single and concurrent patterns.

Keywords:

• concurrent patterns
• pattern recognition
• control charts
• wavelet analysis
• neural networks
• quality control