摘要
随着电控技术发展,汽车发动机结构变得更加复杂、紧凑和精密,传感器也越来越多,如何利用传感器信号来诊断发动机故障类型(尤其是机械与性能方面故障)成为研究热点。选取反映汽车发动机运行工况的3个重要传感器信号作为输入向量,分别建立BP神经网络与概率神经网络模型对发动机失火故障、进排气管堵塞、火花塞间隙过大等非传感器直接监测的典型故障进行识别。结果表明,基于传感器信号的神经网络模型能较好地识别出故障类型,并且通过测试和验证对比可知,概率神经网络(PNN网络)在速度和准确性上均优于BP神经网络,且效率更高。因此该方法在汽车发动机故障识别和诊断中应用更为有效。
With the development of electronic control technology, the structure of automobile engine has become more complex, compact and precise, and there are more and more sensors. How to use sensor signals to diagnose engine fault types(especially mechanical and performance faults) has become a research hotspot. Three important sensor signals reflecting the running condition of automobile engine are selected as the input vectors, and BP neural network and probabilistic neural network models are established to identify the typical faults which are not directly monitored by sensors, such as engine misfire, intake and exhaust pipe blockage, excessive spark plug gap, etc. The results show that the neural network model based on sensor signals can better identify the type of failure, and through testing and verification, it can be seen that the probabilistic neural network(PNN network) is superior to BP neural network in speed and accuracy, and more efficient. Therefore, this method is more effective in the application of automobile engine fault recognition and diagnosis.
作者
李雯
喻菲菲
杜灿谊
李锋
龚永康
Li Wen;Yu Feifei;Du Canyi;Li Feng;Gong Yongkang(School of Electromechanical Engineering,Guangdong Polytechnic Normal University,Guangzhou,Guangdong,510635,China;School of Automobile and Traffic Engineering,Guangdong Polytechnic Normal University)
出处
《小型内燃机与车辆技术》
2020年第5期78-83,共6页
Small Internal Combustion Engine and Vehicle Technique
基金
广东省自然科学基金(2018A030313947,1914050006111)。
关键词
发动机
传感器
故障识别
BP神经网络
概率神经网络
Engine
Sensor
Fault identification
BP neural network
Probabilistic neural network
