Canonical variate analysis for performance degradation under faulty conditions

Abstract

Condition monitoring of industrial processes can minimize maintenance and operating costs while increasing the process safety and enhancing the quality of the product. In order to achieve these goals it is necessary not only to detect and diagnose process faults, but also to react to them by scheduling the maintenance and production according to the condition of the process.

The objective of this investigation is to test the capabilities of canonical variate analysis (CVA) to estimate performance degradation and predict the behavior of a system affected by faults. Process data was acquired from a large-scale experimental multiphase flow facility operated under changing operational conditions where process faults were seeded. The results suggest that CVA can be used effectively to evaluate how faults affect the process variables in comparison to normal operation. The method also predicted future process behavior after the appearance of faults, modeling the system using data collected during the early stages of degradation.