基于BP神经网络的发动机故障诊断研究

【作者】 袁科新；

【导师】 刘玫；

【作者基本信息】 山东大学 ， 控制理论与控制工程， 2006， 硕士

【摘要】 柴油机是许多设备和军事装备的心脏，保证其良好的工作状态对维护其整体工作效率是至关重要的。目前柴油机故障诊断多利用台架试验法，此方法耗时长，台架结构庞大，造价高，难以在应用部门推广。而且由于此类复杂机械的非稳态多元激励特性、时变传递路径特性以及复杂运行机理等因素，使得这一研究方向具有相当的难度。 本文在分析了国内外柴油机故障诊断研究现状的基础上，给出了一种通过燃油压力信号来进行柴油机故障诊断的设计方案。 燃油喷射系统是柴油机的心脏部分，其工作质量直接影响柴油机工作过程及整机性能，决定柴油机的经济性和可靠性。同时，它又是柴油机的主要故障源。而燃油喷射系统的状态信息主要体现在高压油管的压力波形中，当某处发生故障时，必然使原有供油状态发生变化，燃油流动的压力和流速等参数会有相应的改变，反映在压力波形上将导致波形形态和波形参数值的变化。 本文所提出的柴油机故障诊断方法是：利用传感器测出燃油压力信号，根据燃油压力信号绘制出燃油压力波形，并且提取出特征参数，再将这些特征参数当作BP神经网络的输入样本，并利用LM改进学习算法训练所建立的BP神经网络，然后利用训练好的神经网络进行柴油机故障诊断，得出诊断结果。这种方法不仅大大简化了故障诊断的方法，而且还使故障诊断具有人工智能化。 仿真试验表明，基于神经网络的故障诊断结果与实测值具有良好的一致性，只要选择足够典型的原始故障样本训练BP神经网络，网络的容错性和稳定性就较好。针对设备运行的复杂性，仅选用单一的诊断参数往往会做出错误的判断，而基于神经网络的故障模式识别方法能充分利用信息特征，实现输人与输出之间的映射关系，得出准确的诊断结果。人工神经网络为状态监测和故障诊断提供了新的理论方法和技术手段，这种故障诊断方法完全可用于柴油机、大型旋转机组等的故障诊断，因而具有广泛的应用前景。更多还原

【Abstract】 The diesel engine is the heart of many equipment and the military equipment, so it is important to ensure its well working condition and maintains its whole working efficiency. At present, diesel engine fault diagnosis most use the method of gantry testing, but this method need to keep long time and huge structure and high cost, so it is difficulty in promoting the application. Moreover, as such complex mechanical multi-state excitation and operation mechanism and time-varying and other factors enables this research direction to have the suitable difficulty.Based on the analysis of the domestic and foreign research about diesel engine fault diagnosis at the current status, it gives a design proposal in this paper to carry on the diesel engine fault diagnosis through the fuel pressure signal.The fuel injection system is part of the heart of diesel engine. Its work quality has the directly influence about the diesel engine work process and entire machine performance, and it decides the efficiency and the reliability about diesel engine. Meanwhile, it is the main diesel engine failure source. The state information of fuel injection system is mainly manifested in the high-pressure oil pipe profile, When somewhere breaks down, it must causes the original feed condition to change inevitably, and the pressure and speed of the fuel oil flow must be changed corresponding and so on. All of reflections will lead to change the parameters and shapes of the pressure waveform.In this paper, it proposed the method of diesel engine fault diagnosis is: first, determines the fuel pressure signal using the sensor; second, draws up the fuel pressure profile according to the fuel pressure signal, and withdraws the characteristic parameter; third, treats these更多还原