基于神经网络的柴油机故障诊断

基金项目:

山东省自然科学基金项目 Y2002F17

Fault diagnosis of diesel engine based on neural network

1.
School of Traffic and Vehicle, Shandong University of Technology, Zibo 255049, China
2.
Department of Precision Instruments and Mechanics, Tsinghua University, Beijing 100084, China
3.
The 53RD Research Institute of China North Industries Group Corporation, Jinan 250031, China

摘要: 基于虚拟仪器技术, 开发了柴油发动机信号处理系统, 该系统数据采集由传感器与PCI-6023E数据采集卡完成, 系统软件采用LABVIEW6.0, 根据柴油发动机的典型故障, 对系统参数赋值, 利用BP神经网络对采集的数据进行处理, 进而判断发动机的故障, 试验表明该系统是可行的。
- 汽车工程 /
- 神经网络 /
- 虚拟仪器 /
- 故障诊断 /
- 柴油机
Abstract: Based on virtual instrument technology, the fault diagnosis system for diesel engine was developed. Its digital treatment system was consisted of sensors and PCI-6023E collectors, its design software was LABVIEW6. 0. This system character parameters were defined by engine's typical faults. In this system, the collected data were treated with BP neural network to diagnose engine's faults. The tested results show that this system is feasiable.
- automotive engineering /
- neural network /
- virtual instrument /
- fault diagnosis /
- diesel engine

图 1 智能检测仪组成

Figure 1. Composition of intelligent instrument

下载: 全尺寸图片幻灯片

图 2 智能检测仪软件功能

Figure 2. Software functions of intelligent instrument

下载: 全尺寸图片幻灯片

图 3 正常压力曲线

Figure 3. Normal oil pressure curve

下载: 全尺寸图片幻灯片

图 4 故障油压曲线

Figure 4. Abnormal oil pressure curves

下载: 全尺寸图片幻灯片

图 5 神经网络结构

Figure 5. Structure of neural network

下载: 全尺寸图片幻灯片

图 6 网络训练和测试误差曲线

Figure 6. Error curves of network training and testing samples

下载: 全尺寸图片幻灯片

表 1 网络输出向量与故障种类的对应关系

Table 1. Relations of network output vectors and fault kinds

下载: 导出CSV

表 2 特征参数测试样本

Table 2. Testing samples of characteristic parameters

下载: 导出CSV

[1]	Kajior Watanabe. Incipient fault diagnosis of chemical proces-ses via artificial neural networks[J]. Journal of AICHE, 1989, 3(11): 1803-1812
[2]	王文成. 神经网络及其在汽车工程中的应用[M]. 北京: 北京理工大学出版社, 1998.
[3]	赵振宇. 模糊理论和神经网络的基础和应用[M]. 北京: 清华大学出版社, 1993.
[4]	Bosch W. The fuel indicator-a new measuring instrument fordisplay of the characteristics of individual injection[J]. SAE660749.
[5]	Fan J Y. An approach to fault diagnosis of chemical processesvia neural network[J]. Journal of AICHE, 1993, 39 (1): 82-87. doi: 10.1002/aic.690390109
[6]	徐丽娜. 神经网络控制[M]. 哈尔滨: 哈尔滨工业大学出版社, 1993.
[7]	戴葵. 神经网络应用技术[M]. 北京: 国防科技大学出版社, 1998.
[8]	申焱. 基于Kohonen网络的机械故障模式识别[J]. 交通运输工程学报, 2002, 2(2): 55-59. http://transport.chd.edu.cn/article/id/200202013 SHEN Yan. Pattern recognition of mechanical fault based on Kohonen neural network[J]. Journal of Traffic and Transportation Engineering, 2002, 2(2): 55-59. (in Chinese) http://transport.chd.edu.cn/article/id/200202013
[9]	John J Hopfield. Artificial neural networks[J]. IEEE Circuits and Devices Magzine, 1988, 18(12): 3-10.