An Online Monitoring and Fault Diagnosis Scheme for High-Dimensional Stream

This article proposes a two-stage process monitoring and fault diagnosis scheme for the high-dimensional data stream. In the first stage, we choose the exponentially weighted moving average (EWMA) statistic to construct the online monitoring statistic, then determine the change point of the data stream based on extreme value theory and multiple hypothesis testing procedures. In the second stage, when the data stream has an abnormal change alarm, it enters an out-of-control state, and we perform a fault diagnosis procedure on the data stream. Specifically, when an abnormal alarm occurs, we further utilize the numerical simulation procedures to detect these components that exhibit anomalies. In addition, to verify the performance of the proposed method, this work also conducts numerical simulations and case studies and compares them with other competitive methods.