This paper describes an online diagnosis approach for leakage and oil contamination faults in hydraulic disc brake system using feature extracted from model and an SVM classifier. A feature signal, which can reflect both faults directly, is derived from the mechanism model of hydraulic cylinder. Feature parameters extracted from the feature signal are related to the magnitude and the category of the faults. An SVM classifier is used to isolate the two kinds of faults. After identifying the fault category, the fault magnitude can be estimated based on the relationship between the fault magnitude and the feature parameter. Simulation results show that this method fulfills the function of online fault diagnosis effectively.