A Comparison Study of Distribution‐Free Multivariate SPC Methods for Multimode Data

摘要: The data-rich environments of industrial applications lead to large amounts correlated quality characteristics that are monitored using Multivariate Statistical Process Control (MSPC) tools. These variables usually represent heterogeneous quantities originate from one or multiple sensors and acquired with different sampling parameters. In this framework, any assumptions relative the underlying statistical distribution may not be appropriate, conventional MSPC methods deliver unacceptable performances. addition, in many practical applications, process switches operating mode a one, leading stream multimode data. Various nonparametric approaches have been proposed for design multivariate control charts, but monitoring processes remains challenge most them. study, we investigate use distribution-free based on learning work, compared kernel distance-based chart (K-chart) one-class-classification variant support vector machines fuzzy neural network method adaptive resonance theory. performances two were evaluated both Monte Carlo simulations real data. simulated scenarios include types out-of-control conditions highlight advantages disadvantages methods. Real data during roll grinding provide framework assessment applicability these applications. Copyright © 2014 John Wiley & Sons, Ltd.