摘要
针对目前汽车发动机的传感器易损坏而导致发动机状态分析的结果产生重大偏差的特点,对人工神经BP网络模型做了改进,使其具有很强的自适应能力而能使网络的收敛方向和速度得到优化,并编制了相应的程序。作为实例,文章对某一实际发动机进行了仿真试验,结果表明该改进的BP网络具有很强的自适应能力,所有的误差控制在3%以内,可以满足工程实际的需要。由于人工神经网络在实际应用中不涉及具体的物理模型,因此该模型对发动机的状态参数在线仿真、减少传感器的维护量,特别是对发动机故障诊断技术水平的提高有很大的意义。
The abnormity of sensor of automotive engine will lead to great difference of result of engine status analysis. This paper applies artificial neural BP network to keep track of engine＇ sensor. A simulation experiment is performed to show that the improved BP network is self- adaptive and the error within 3% could be maintained. It could meet practical need of the project. The physical model is not involved in the application of artificial neural BP network, hence the significance of the model is far - reaching. It might exert profound influences on the online simulation of parameter, on reducing sensor maintenance, and on the development of the fault diagnosis technology.
出处
《计算机仿真》
CSCD
2007年第11期246-248,323,共4页
Computer Simulation
基金
广东省科技计划项目(2005B10201014)
关键词
发动机
故障诊断
神经网络
Engine
Fault diagnosis
Neural network
