Support Vector Machine Control Chart for Multivariate Data

The emergence of quality 4.0, driven by industry 4.0 technologies, introduced complex, high-dimensional, and non-normal data that challenges the effectiveness of traditional control charts. The objective of this study is to proposes a support vector machine-based control chart that can handle multi non-normal variables. The proposed Support Vector Machine (SVM) chart can also handle one class data which is very common in todays automated processes. The approach uses median rank nonparametric probability estimation to generate out-of-control vectors followed by kernel-based function hyper plane parameter estimation. The distance between vectors and hyperplane distance is used to develop the control chart and flag any out-of-control events. The proposed method is demonstrated using real life case study.