Condition Monitoring and Performance Prognosis

Abstract

Many data-driven techniques have been developed and applied for industrial process monitoring and fault diagnosis. In contrast to model-based methods, data-driven approaches can provide efficient process monitoring for sophisticated and highly complex modern industry machines by extracting statistical models based only on historical process data. Various technical systems are investigated in this project, in which data-driven methods are studied with real process data for fault detection, fault identification and performance prognosis. The performance of data-driven methods is considered with examples of an industrial process and a wind energy conversion system.