Methods Research On Aero-engine Fault Diagnosis Based On Gas Path Performance Parameter

Aero-engine is a complex system engineering. Any faults will have a bad effect on the flight and even lead to catastrophic accidents. Most of the aero-engines used in our country are manufactured by foreign engine manufacturers. Although the engine manufacturers always provide the use o f manual, the core technology of the safety assurance is still blocked to the airline companies. In order to break through the ‘bottleneck’, it is very important to find the method that can be applied to the fault diagnosis of aero-engines. This paper has carried out the following research work for fault diagnosis of civil aero-engines.Firstly, the fault diagnosis method of engine gas path based on the r elat ive gradient of performance parameters was developed. Through the analysis of the Customer Notification Report provided by the Original Equipment Manufacturer, the paper could judge that the OEM manufacturers used the set trends of the performance parameter for fault diagnosis. In order to obtain the variation of the performance parameters which was used for fault diagnosis, this paper established a mathemat ical model of the relative gradient of the gas path parameters for aero-engines. According to the historical flight record and analysis of a large number of experiments on the relat ive gradient of flight cycle performance parameters, the paper obtained the safety thresho ld of the relative gradient o f performance parameters, which was based on a given normal flight cycle. The safet y threshold of relative gradient was verified by the test of the fault of exhausted gas temperature indication, the fault of total air temperature indication and the fault of variable bleed valve. The test results showed that the method was accurate for the fault point identification.Then, considering the time sequence of engine performance parameters, a method for aero-engine fault diagnosis based on the trend analysis of mult i performance parameters was developed. In this paper, the neural network model was used to study the performance parameters of the engine gas path with a high degree of nonlinear approximation performance. The performance change trend was mapped to the trend evaluation value of [-1,1], and the trend of t he diagnostic performance parameters was compared with the trend of the fault samples to obtain the result s of fault diagnosis. Two kinds of trend assessment methods were carried out in the paper. One method was to directly use the time series samples as neural network inputs; the other was to use the characteristic parameters obtained by fitting the time series samples as the input of the neural network. These two methods were applied to the test of the specific fault diagnosis and the multi-fault classificat ion of the engine. The test results showed that both of these two methods had high accuracy in fault diagnosis and mult i-fault classification.Finally, a new method of engine fault diagnosis based on weighted D-S evidence fusion was developed. Evidence bodies were identified, as well as the basic probabilit y assignment of above two methods: the relative gradient method and the performance trend analysis method were identified. Besides, the improved evidence theory was used to solve the weighted assignment, the confidence intervals and the uncertainty of multi-evidence bodies. The above two kinds o f fault diagnosis methods were fused and obtained the diagnosis result based o n comprehensive judgement. The test results showed that the accuracy of engine fault diagnosis was significantly improved by using this method.