Title: Denoising and feature extraction for control chart pattern recognition in autocorrelated processes
Authors: Hui-Ping Cheng, Chuen-Sheng Cheng
Addresses: Department of Business Administration, Ming-Dao University, 369 Wen-Hua Rd., Peetow, Changhua 52345, Taiwan, ROC. ' Department of Industrial Engineering and Management, Yuan-Ze University, 135 Yuan-Tung Rd., Chung-Li, Taoyuan 320, Taiwan, ROC
Abstract: The main purpose of this paper is to develop a neural network-based recogniser for control chart pattern recognition in autocorrelated processes. First, we apply a multi-resolution analysis approach based on Haar Discrete Wavelet Transform (DWT) to denoise, decorrelate and extract distinguished features from autocorrelated data. Second, we introduce a supervised neural network for control chart pattern recognition. The performance of the neural network using features extracted from wavelet analysis as the components of the input vectors is explored and compared. In this study, we investigated three types of unnatural patterns, namely increasing and decreasing trends, cyclic patterns, upward and downward shifts. Extensive comparisons based on simulation study indicate that the proposed neural network performs better than that using raw data as inputs.
Keywords: SPC; statistical process control; pattern recognition; multiresolution analysis; DWT; discrete wavelet transform; neural networks; autocorrelated processes; denoising; feature extraction; control charts; simulation.
International Journal of Signal and Imaging Systems Engineering, 2008 Vol.1 No.2, pp.115 - 126
Published online: 24 Oct 2008 *Full-text access for editors Access for subscribers Purchase this article Comment on this article