Application of Data-based Process Topology and Feature Extraction for Fault Diagnosis of an Industrial Platinum Group Metals Concentrator Plant

Abstract

The development and application of a fault diagnosis methodology combining process topology and feature extraction to an industrial platinum group metals (PGM) concentrator plant is presented in this paper. Using the available historical process data, process topology was extracted to generate a connectivity map of the process. This connectivity map was used to separate the process into multiple blocks to be analyzed separately. Feature extraction was applied in these blocks to determine if a known fault could be detected, and which block showed the largest effect of the fault. The connectivity map of this block was then used to determine possible fault propagation paths to identify possible root causes of the fault. The developed methodology provided accurate fault detection and identification results, indicating that incorporation of topology information in fault diagnosis is beneficial for fault diagnosis. A description of the envisioned application of these techniques and recommendations regarding this application is given.