| Summary: | 碩士 === 樹德科技大學 === 資訊管理系碩士班 === 98 === The purpose of this thesis is to develop a Neural Network Condition Monitoring and Fault Diagnosis system of a turbofan engine with afterburner. The semi-artificial sensing engine data are normalized and then feeding into the neural network. There are two model of our purposed system: 1. limited-model in which contends 4-node input and 5-node output parameters; 2.extented-model in which contends 6-node input and 7-node output parameters. By the using of gradient method, momentum term method and Levenberg Marquardt (LM) method, the results show excellent effectiveness and accuracy. This shows that the construction of purposed system can be used as a reference of the faultier diagnosis.
As a result, in the case of limited-model, it shows smaller root mean square error in the network architetecture of a 21-node hidden layer neurons using LM algorithm and achieves 90% admeasure rate. In the case of extended-model, the network architetecture of the 25-node hidden layer neurons using LM algorithm can achieve 100% admeasure rate. Finally, the system is then applied for diagnosis of the turbofan engine with hot-section. The effectiveness of the proposed system is verified.
|
|---|
