Health Monitoring of a system is a method to detect a fault in such a system before it leads to a failure (jamming, breakage,…) or extensive performance degradation which hampers normal operation. This monitoring can be done to increase the system reliability (critical aircraft parts like EMAs) or move from interval-based maintenance to need-driven maintenance, reducing operational costs.
To do so, a set of sensor signals is analyzed. Since these faults should be measured as early as possible, they are usually at a stage where they can hardly be measured. This leads to very low signal-noise ratios of these sensor signals. By using dynamical system models of the monitored system, it is often possible to increase the detectability of these incipient faults (Kalman filtering, fault observers). Post processing usually involves frequency analysis methods (FFT, Wavelet) to pinpoint a failure. Furthermore stochastic analyses can be made to further improve detection.
At the moment new algorithms and methods are developed and tested on hardware demonstrators. Furthermore simulation models are developed to simulate faults. This can benefit in the design and optimization of these health monitoring methods
Since a typical system can have many faults, it is often impossible to test if all faults can be detected with the developed algorithms. Therefore simulation of these faults becomes important.
Often the fault mechanisms of system parts are poorly understood. This makes the definition of system models which represent the effect of the faults difficult, and proves to be a challenging task and topic of ongoing research.
Static and dynamic gear test rig