Robust Fault Diagnosis for DK-2 Brake Based on Both Data-driven and Model

In this paper, according to the dynamic features of strong coupling, non-linearity and uncertainty for the DK-2 brake, a fault diagnosis strategy based on data-driven and model is proposed for online monitoring and fault isolation. Firstly, based on the analysis of structure and operating principle of the DK-2 brake, the model is established by using bond graph. Secondly, based on the large amount of historical data in normal operating condition, the empirical regression model which can be used to estimate the output by the observations is set up using the auto-associative kernel regression (AAKR). Thirdly, based on prediction intervals (PIs), the uncertainty of the system is described quantitatively in the regression model, so the robust on-line monitoring of the system is realized. Combining the two methods, the online monitoring and fault isolation of the DK-2 brake can be achieved. Finally, the strategy proposed in this paper is verified by simulation and the results show that it can realize online monitoring and fault isolation accurately and quickly.