The diagnosis of Aero-engine's state based on rough set and improved BP neural network

Abstract

Operating parameters and life parameters are the factors that determine the state of the engine. However, the operating parameters and life parameters also include many factors, such as cycle life damage rate, speed life damage rate, low and high pressure rotor rate and so on. Based on these factors, the evaluation and diagnosis system of aero-engine is built. Due to many factors that need to be considered, the training time of failure diagnosis based on traditional neural network is long. After taking account of these reasons, this paper is based on the rough set theory and a kind of improved BP neural network to realize the diagnosis more rapidly and accurately. According to a variety of performance indices, the key parameters are extracted as evaluating indicators. And then, this paper uses the rough set theory based on genetic algorithm to reduce these factors. In the condition of keeping the ability of classification, a kind of improved BP neural network is used to train the simplified parameters. Subsequently, we input the sample data into the neural network. By comparing with the final output, we can know whether the simplified evaluation indicators are feasible. The testing results show this paper offers a feasible way to solve comprehensive estimation of aero-engine performance. And this method shortens the training time, improves the accuracy of diagnosis, and provides the reference for the maintenance of the engine.