An adaptive robust framework for model-based fault detection

Phanindra V Garimella, Purdue University


For the design of reliable control systems, one of the integral parts is the ability of the system to react to sudden unexpected changes in the operating conditions like sensor or actuator failure. Model-based fault detection and diagnosis systems have found extensive use because of the fast response to abrupt failure and the ease of implementation of these schemes in real-time algorithms. The major challenge in the design of model based FDD systems is that they rely on an idealized assumption that a perfect mathematical model of the physical process is available and this model is utilized in the design of the state observers or parameter estimation schemes to detect faults. In practice, however, this assumption can never be completely satisfied since an accurate mathematical model of the physical system is not usually available either because of the uncertainty in the parameters of the system or because of the unmodeled dynamics. This leads to a mismatch between the actual physical system and the mathematical model used to design the FDD algorithm which could lead to unwanted false alarms. Hence, the robustness of the fault detection scheme to modeling errors without losing sensitivity to faults is the key problem in the application of model-based FDD algorithms. This dissertation proposes the concept of an adaptive robust framework for the detection and isolation of state and sensor faults. The causes of the uncertainty in the model of the physical system used in the design of the FDI algorithm are investigated and different tools are utilized to attenuate the different causes of the uncertainty. The proposed framework utilizes extended filter structures to attenuate the effect of unmodeled dynamics and intelligently controlled on-line parameter adaptation to reduce the effect of the unknown parameters. In order to validate the proposed scheme, analytical results on robustness and sensitivity of the proposed scheme are presented. In addition, the FDD scheme is applied to the swing arm of a three degree of freedom hydraulic system to detect typical faults like the lack of sufficient supply pressure, contamination of the hydraulic fluid and typical sensor failures like position and pressure sensor faults.




Yao, Purdue University.

Subject Area

Mechanical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server