节点文献
基于建模仿真与振动分析的发动机故障诊断方法研究
Study of Automotive Engine Fault Diagnosis Based on Model Simulation and Vibration Analysis
【作者】 杜灿谊;
【导师】 丁康;
【作者基本信息】 华南理工大学 , 车辆工程, 2013, 博士
【摘要】 汽车发动机结构越来越复杂,安装精度要求也越来越高。在不拆检的情况下,如何判断出故障的类型和位置,是目前研究的重点和难点。研究发动机故障时,往往需要测量故障状态下的相关参数变化情况,了解故障信息并提取故障特征,但是并不轻易具备收集各种故障信息的条件,而靠人为方法设置发动机故障,往往会导致发动机严重的损伤和破坏,试验成本很高,限制了发动机故障诊断的深入研究。针对发动机故障诊断技术研究存在的不足,首次通过建模仿真和实验对发动机整机进行典型故障机理、特征提取和诊断方法研究。利用有限元与多体动力学联合建模,计算结果更精确可靠;通过模型进行了多种类型的故障仿真,从激励力变化这一内在因素分析机体表面振动响应变化,深层次了解故障引起的机体异常振动成因和特性,通过模型正常状态和故障状态的机体表面振动信号变化趋势分析来提取故障特征,为发动机故障诊断研究提供一种新方法。利用多体动力学和有限元技术建立高质量标杆发动机仿真模型,通过相关参数设置,分别模拟发动机失火故障、气门间隙大、凸轮轴承松脱、活塞敲缸和主轴承磨损等故障,从机理上进行了分析,提取了相应故障特征,得到了相关实验验证。提出了应用小波包分解与解调方法来提取气门间隙过大和凸轮轴松脱故障的故障特征。利用发动机仿真模型进行了气门间隙大和凸轮轴承松脱的故障仿真分析,发现由于气门间隙增大,导致气门落座力、凸轮轴承力的增大,这些增大的冲击力使缸盖振动响应时域信号在一个工作循环内出现3个明显冲击,在频谱上激发起中高频频段内的显著振动能量。利用小波包分解到此特征频段后再解调,故障状态的解调谱中出现明显的1.5阶次成分;而对于凸轮轴松脱故障,用相同方法可得到故障状态下解调谱的0.5阶次的明显增大。从理论上说明了此种类型故障具有明显的中高频调制现象,为故障特征提取和故障诊断研究提供有力参考。通过仿真分析发现活塞敲缸故障会激发2500Hz以上高频段振动能量的加剧,提出故障状态与正常状态的小波包分解总能量比随转速变化趋势作为依据进行故障诊断的方法;通过对主轴承间隙过大造成主轴承反力的波动变化,解析了此类故障引起200~1000Hz中低频段振动能量增加的原因,为此类故障诊断研究提供理论依据。
【Abstract】 The structure of automotive engine becomes increasingly complicated and therequirement for installation precision increasing too. How to diagnose the fault’s type andlocation under the non-overhaul condition has been the key and difficult points of researchnowadays. We need to measure the change of relative parameters, know the fault informationand exact fault features while studying the engine fault, however it’s difficult to collect theabove information.And,if the engine’s faults are made artificially, it will result in severedamage of engine and high testing cost, which limit the further survey.Aimming at the existed weakness of engine fault diagnosis techniques, the faults’mechanism, feature extraction and diagnosis methods of the whole engine is studied for thefirst time by modeling and simulation. The results will become more accurate and reliable byjoint modeling under the finite element and multi-body dynamics. By analyzing the vibrationresponse on the surface of engine body considering the inner factor of exciting force, thereasons and characteristics of abnormal engine body vibration caused by the fault is obtained.By the trend analysis of vibration signals of the body surface under the normal and faultconditions, the fault feature is extracted, and which provided a new way for the engine faultdiagnosis. By the multi-body dynamics and the finite element technique, high qualitybenchmarking engine simulation model is builded. Faults including the engine misfire,largevalve clearance, cam bearing looseness, piston knocking and main bearing wears can besimulated by the related parameters setting. The fault features are extracted and verifiedaccordingly.A fault feature extraction method is proposed by the application of wavelet packetdecomposition and demodulation method to extract the faults of the excessive valve clearanceand looseness of camshaft. The faults of the excessive valve clearance and looseness ofcamshaft are analyzed using the engine simulation model established. It is found that with theincreasement of valve clearance, the valve seating force and the bearing force increasing too.And because of the increasing impact forces, three significant impacts in a working cycleappear in the time domain vibration signal of cylinder head, which arouses the notablevibration energy in the high frequency. By wavelet packet decomposition, the signal of the high frequency band is decomposed. For the faults of excessive valve clearance and camshaftlooseness, the demodulation spectrum appears obvious1.5order component and the0.5ordercomponent, respectively. It is found that this kind of fault has significant high frequencymodulation phenomenon in theory, which provides strong reference to failure featureextraction and fault diagnosis. The reason why the energy of high frequency over2500Hzincreased when piston knock is analyzed theoretically. A method of fault diagnosis based onratio’s change trend with rotate speed between the total energy of wavelet packetdecomposition under fault condition and normal condition is presented. The reason why thevibration energy of medium-low frequency between200and1000Hz increased is analyzed intheory, which is caused by the fluctuations of main bearing reaction force caused by excessiveclearance of main bearing. The method can provide theoretical basis for this kind of faultdiagnosis research.
【Key words】 Modeling and Simulation; Finite Element; Vibration; Excitation Force; FaultDiagnosis; Feature Extraction;