A novel health probability based engineering system condition and health monitoring method

摘要

Engineering system condition and health monitoring is traditionally concerned with fitting sensors inside the system and analyzing the features of signals from the sensor measurements using appropriate signal processing techniques to reveal the system's condition and health status. However, the conventional signal only based analysis often cannot distinguish the normal changes due to the differences in system environmental or operating parameters from the changes that are induced by damage. This is because the changes revealed by sensor signal analysis can not only show what happens with the condition and health status of inspected systems but may also reflect normal changes in the system such as the changes in system environmental or operating parameters. Motivated by the need to correctly identify the changes in the signal features that are produced by abnormality in inspected systems, a novel health probability based engineering system condition and health monitoring method is proposed in this paper. In this method, the relationship between a signal feature and the normal changes in the system environmental and operating parameters, known as baseline model, is first established. Then, a tolerance range of the signal feature's deviation from what is determined by the baseline model is evaluated via a data based training process. Furthermore, the health probability, which is defined as the proportion of the cases where the system's working status as represented by the signal feature is within the tolerance range, is used to decide whether a system is in a normal working condition or not so as to implement the system condition and health monitoring. Experimental data analyses have been conducted to demonstrate the performance of the proposed new technique.