摘要
根据发动机的结构特点,将其表面划分成不同的测试区域进行声强信号采集;依据声强特征,确定不同区域对应零部件的工作状况;利用模块化神经网络,建立基于声强特征的故障诊断模型,该模型中包含发动机低速与中速诊断模块、决策模块和故障知识库;在建模过程中,利用特征函数强化故障特征作为网络输入。结果表明,该方法具有诊断精度高、速度快、实时自学习等特点,为建立更为完善的发动机智能化故障诊断系统提供新途径。
According to the characteristics of engine structure, the engine surface is divided into different test areas for acquiring sound intensity signals. Based on the features of sound intensity, the working status of the engine parts and components can be determined. By utilizing modular neural network, a fault diagnosis model is built based on sound intensity features. The model includes four sub-modules for low speed diagnosis, medium speed diagnosis, decision-making and fault knowledge database respectively. As network input, the fault feature is enhanced by eigen-function in modeling. The results show that the diagnosis method has advantages of high diagnosis accuracy, short diagnosis time, and the ability of real time self-learning, which provides a new way to build a more perfect intelligent fault diagnosis system for engines.
出处
《汽车工程》
EI
CSCD
北大核心
2006年第4期401-404,共4页
Automotive Engineering
基金
南京理工大学青年学者基金(njust200202)资助。
关键词
声强
神经网络
发动机
故障诊断
Sound intensity, Neural network, Engine, Fault diagnosis
