A fault diagnosis method of engine rotor based on Random Forests

Abstract

Rotor is the main part of the engine, the vibration fault is very common in the process of running, it must be monitored, checked, excluded in a timely manner for improving the reliability of engine and aircraft safety. This paper mainly studies four kinds of rotor fault, including unbalance, misalignment, surge, bearing failure. The frequency spectrum of the vibration signal of a rotor system is an important basis for rotor fault diagnosis, using the spectrum of rotor to build decision tree analysis is an important method for rotor fault detection. As the single decision tree's anti-interference ability is very poor, this paper presents an engine rotor fault diagnosis method based on Random Forests. Experimental results show that the accuracy of this diagnosis method is high, the failures can be diagnosed timely and effectively to keep the engine in normal operation. To evaluate the validity of Random Forests, a SVM classifier is trained for comparison. Compare with SVM, we obtain better classification in Random Forests algorithm. This result demonstrates that Random Forests algorithm is a valid method for engine rotor.