Isolating Multiple Sensor Faults Based on Self-Contribution Plots with Adaptive Monitoring

摘要

Considering the time-varying nature of an industrial process, an adaptive monitoring method based on fast moving window principal component analysis (FMWPCA) was developed. The proposed approach adapted the parameters of the monitoring model with the dissimilarities between the new and oldest data, rather than recursively downgrading and upgrading the parameters. It was found to be more efficient than other approaches tackling similar problems. When process faults are detected, isolating the faulty variables provides additional information to investigate the root causes of the faults. Numerous data-driven approaches require the datasets of known faults, which may not exist for some industrial processes, in order to isolate the faulty variables. For this type of approach, incorrect information would be provided when encountering a new fault that was not in the known event list. The contribution plot is a popular tool to isolate faulty variables without a priori knowledge. However, it is well known that this approach suffers from a smearing effect, which may lead to the incorrect identification of the faulty variables in the detected faults. In the presented work, a contribution plot without the smearing effect was derived, and was named the self-contribution plot. An industrial example, correctly isolating faulty variables and diagnosing the root causes of the faults for the compression process, was provided to demonstrate the effectiveness of the proposed approach for industrial processes.