Inductive monitoring system based fault detection for liquid-propellant rocket engines

The ability of in-time, real-time and reliable fault detection is essential for liquid-propellant rocket engine to monitor its working status. Inductive Monitoring System (IMS) is a data-mining based real-time fault detection method. This paper studies the problem that how to reduce the high false alarm rate when IMS is applied to monitor the system during trials or flights. It is found that the order of sample data and outliers of test data contribute a lot to the high false alarm rate. To reduce the interference evoked by the order of sample data and outliers of test data, an improved IMS method is proposed by rescheduling the sample data and processing outliers that may occur during the monitoring process. Simulation is performed to show that the improved IMS in this paper can decrease the false alarm rate while ensuring timeliness, reliability and accuracy of fault detection.