A methodology for fault propagation analysis based on nearest neighbors method combined with the information on process connectivity
Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Professional
In such cases, it is essential to detect and eliminate the root cause of the faulty condition as early as possible in order to minimize its adverse effect on the entire process. Investigating the causal dependencies among the process variables can assist in retracing the fault propagation path and ultimately the root cause.
In recent times, several data-based methods have been developed and were successfully applied in order to capture causality. However, data-based methods suffer from several limitations and deficiencies. This paper proposes a new methodology for implementing the nearest neighbors method in order to retrace the propagation path of a fault in a complex system. The methodology is implemented in two phases. In phase I, the information on the process connectivity is incorporated into the data-based analysis using a unique search algorithm in order to consider only the interactions which are direct based on the plant topology. In phase II, a new multivariate directionality measure is introduced in order to exclude indirect paths and is applied in conjunction with the search algorithm. The methodology is successfully demonstrated on an industrial board machine exhibiting oscillations in its drying section.
|Title of host publication||Automaatiopäivät22|
|Publication status||Published - 23 Mar 2017|
|MoE publication type||D3 Professional conference proceedings|
|Event||Automaatiopäivät - Vaasa, Finland|
Duration: 23 Mar 2017 → 24 Mar 2017
Conference number: 22
|Period||23/03/2017 → 24/03/2017|
- nearest neighbors, causality, process connectivity, fault analysis, industrial application