Adaptive Real Time power plant fault detection: Improving reliability of Industrial Equipment


Dr. Ankur Teredesai, UW Tacoma

Industry Advisors

Dr. Anil Varma, GE Global Research

Dr. Feng Xue, GE Global Research

Dr. Naresh Iyer, GE Global Research

Graduate Committee Member(s)

Dr. Orlando Baiocchi


Pattern recognition techniques can be used for damage detection in industrial equipment. In this scenario, if damage can be detected early then remedial action can be taken before a catastrophic event occurs. We propose to develop, compare, and apply several anomaly detection algorithms to detect multiple types of power plant faults. Information fusion will be used to develop ensemble anomaly detectors with improved robustness and accuracy. We will then compare the performance of our algorithms against anomaly detection using standard statistical process control. The ensemble fault detectors will be incorporated into demonstration system that is built upon a complex event data processing software framework, will consume actual andsimulation power plant data and will flag equipment faults.

Institute of Technology
Research Group Association: