【作者】 刘玉梅；

【导师】 苏建；

【作者基本信息】 吉林大学 ， 载运工具运用工程， 2009， 博士

【摘要】 汽车悬架装置是保证汽车的动力性得到充分发挥,保证汽车具有良好的乘坐舒适性和行车安全性的重要总成。为了保证汽车悬架装置可靠工作,应定期进行悬架性能检测与故障诊断,以对其进行合理的维护和修理,确保其使用性能始终处于良好状态。论文对悬架系统的故障诊断方法进行了深入研究,提出了以车身加速度、车轮相对动载荷、悬架动挠度、车轮振动滞后相位差和悬架间隙等多源信息作为悬架故障诊断参数,采用多源信息融合技术,解决悬架故障多因素耦合问题。构建了基于传递函数的故障诊断参数模型,并通过Laplace变换和大量的实车试验、仿真实验,建立了基于幅频特性特征参数与特性曲线的诊断标准数据库;在构建基于Euclidean距离函数的悬架系统故障诊断模型基础上,实现基于悬架幅频特性的悬架系统故障诊断,并定义了悬架优度以评价非故障悬架性能的优劣程度,使悬架性能的评价定量化、科学化。对于已偏离正常性能的悬架系统,采用模糊诊断方法,建立了基于最大隶属原则的悬架故障诊断模糊综合评判模型,对悬架系统进行综合评判,评判结果表明,利用一种或数种故障征兆可以准确、快速地诊断悬架故障产生的原因和部位。论文还建立了基于Fisher有序聚类分析的汽车悬架间隙识别模型,通过全局优化获得曲线分段的最优解,并采用最小误差函数通过逐级迭代准确识别测试曲线上的最佳分类间隔点,从而实现了悬架间隙的准确识别。总之,论文在悬架故障诊断的数值建模、算法应用、技术实现等方面取得了部分研究成果,对提高汽车检测与故障诊断技术水平和实现车辆故障诊断智能化有指导意义和推动作用,所研究的故障诊断方法在工程应用领域具有理论意义和实际应用价值。更多还原

【Abstract】 Automotive suspension is an important assembly which not only ensures riding comfortableness, but also fully displays the vehicle power performance, and keeps handling stability, braking performance and active safety in a reliable state. When driven at high speed, the smoothness and safety are related to the quality of the suspension device. With the improvement of road and traffic conditions, especially the rapid development of highway, the speed of modern automobile is rising; at the same time, along with society advancement and people’s living standard improving, the requirements of operation performance and riding comfortableness are getting higher and higher, and the requirements of technical state of vehicle especially suspension system is even higher. Therefore in order to ensure vehicle suspension device works reliably, it’s necessary to detect suspension performance at fixed period, and carry out reliable fault diagnosis aiming at the ones with performance degradation to find out the location and the reasons of the fault to maintain and repair the suspension reasonably, so that suspension performance always keeps in good condition..However, as the complexity of the suspension system structure and the uncertainty and non-linear property of dynamic characteristics, the fault diagnosis of suspension system becomes very complex. One fault symptom is always coupling interaction of multi-factor, and there is not a simple one to one correspondence between the fault symptom and fault reason, what’s more, several faults occur simultaneously and are induced mutually, and there is no clear-cut boundaries between different faults exist similarity and ambiguity. At present, however, suspension performance testing equipments adopt single or one-class sensor, with only one detection parameter, so they can not evaluate the systematization and integrity of the suspension overall, besides, they are lack of the cooperative application and comprehensive treatment of multi-source and multi-dimensional information of complicated suspension, so they not only can’t evaluate the suspension system objectively and accurately, but also can’t provide reliable foundation for suspension system fault detecting and fault diagnosis. Therefore it’s essential to carried out the further study on suspension system fault diagnosis method in order to make qualitative and quantitative analysis scientifically by means of the results of suspension performance testing, determine the suspension technical condition, ascertain the fault’s location and the reasons, establish comprehensive suspension fault diagnosis system, and realize suspension detecting and diagnosis rapidly and efficiently to provide accurate and rich information about the suspension technical condition for the follow-up maintenance work, besides, provide guidance for the design and alteration of the suspension in the future.Under the guidance of the research thought put forward, suspension system fault diagnosis method was mainly studied in this paper. Main contents and conclusions were summarized as follows:1. Put forward multi-source such as body acceleration, wheel relative dynamic load, suspension dynamic deflection, wheel lagging phase difference and suspension gap as fault diagnosis parameters, made use of multi-sensor differences in performance and complementary, integrate variety of data information from each senor and adopt multi-source information fusion technique to evaluate suspension performance scientifically and diagnoses suspension fault reliably.2. Through the analysis of the vehicle vibration system and suspension fault detection system,’vehicle-board’two degree-of-freedom vibration model was established. And according to the Newton’s second law, 1/4‘vehicle-board’system dynamics model was established in this paper which prove that the’vehicle-board’two degree-of-freedom vibration model was made of the first order mode and second order mode through the mathematical analysis of the model. When the external excitation frequency approaches to the first order mode, low-frequency resonance occurs, with body vibration mainly, while that approached to second order mode, high-frequency resonance occurs, with wheel vibration mainly. The proof of the law above provides theoretical and experimental basis for suspension fault diagnosis.3. Through the Laplace transform of the’vehicle-board’vibration system mathematical model, the mathematical model with body acceleration, dynamic deflection, wheel relative dynamic load, wheel lagging phase difference as fault diagnosis parameter based on transfer function as well as corresponding frequency response function and the amplitude-frequency characteristic was established which provided theoretical basis for the suspension intelligent diagnosis.4. The vehicle suspension test simulation system which selecting linear sinusoidal variable frequency disturbing function as incentiveness was developed under Simulink environment. The road roughness linear variable frequency incentiveness mathematical model and simulation model were also established. Besides, 1/4 suspension two degree-of-freedom vibration model and 1/4 suspension simulation model which based on some fault diagnosis parameters such as body acceleration, dynamic deflection, wheel relative dynamic load, wheel lagging phase difference were constructed. By comparing results between simulation and real vehicle test, the accuracy of the model established was proved to be correct, and verifies the feasibility and affectivity of linear sinusoidal variable frequency incentiveness model was also verified.5. Through a lot of simulation tests on suspensions of different models and different types, the mapping relationship between suspension inherent parameters variation and the amplitude-frequency characteristic curve of fault diagnosis parameters, The suspension inherent parameters which including the body acceleration, dynamic deflection, wheel relative dynamic load, wheel lagging phase difference affect suspension structure performance and technical condition directly. The response of amplitude-frequency characteristic curve can be used to evaluate suspension performance and diagnose suspension fault.6. Suspension fault reason and fault symptom are parameterized, with the wheel stiffness k1 , suspension stiffness k 2and damping coefficient c 2respectively represents the technical performance and technical conditions of the wheel, elastic components and shock absorber. The corresponding relationship between the fault symptom which based on body acceleration, dynamic deflection, wheel relative dynamic load, wheel lagging phase difference and fault reason was concluded through a lot of simulation tests on suspension of different models and different types. That provided the test basis for fault diagnosis database based on amplitude-frequency characteristic parameters and curve and reduced real vehicle test greatly, and provided theoretical basis and technical support for development of vehicle suspension fault diagnosis system.7. By means of Euclidean, put forward a new method of suspension system fault diagnosis distance based on amplitude-frequency characteristic. The suspension system fault diagnosis model based on Euclidean distance was established with which to determine whether the suspension exist fault. The concept of suspension goodness was used to evaluate the virtues or defect degree of non-fault suspension performance to make the evaluation quantitative and scientific. Real vehicle test proved that the model established and the methods used were correct.8. The flexible four-link mechanical model of wheel assembly and suspension gap detecting mechanical model were established which provided theoretical basis for the intelligent inspection and identification of the suspension gaps. Besides, the suspension gap identification mathematical model based on Fisher ordered cluster analysis was also established, therefore, a new suspension gap detecting method based on Fisher cluster analysis was presented. The best subsection coefficient was introduced as an optimization evaluation index for the first time so as to obtain the optimal subsection number of the curve to be identified. By global iterative approximation, the index can make the subsection curve get the optimal solution in the global. The minimum error function was used as the objective function to identify the best subsection points of the curve. Through iterating step by step, the optimal subsection interval points could be identified and the suspension gap can be identified and extracted accurately. Some vehicles with typical suspension structure such as JettaGix and BJ2020SA were selected for suspension gap identification testing. The testing results proved that the Fisher ordered cluster analysis had higher identification precision in detecting suspension gap which was better than that of traditional methods. In summary, the suspension gap identification mathematical model established was proved to be correct and the Fisher ordered cluster analysis method which was used to identify the suspension gap was feasible.9. Multi-source information fusion theory and fuzzy mathematical diagnosis method applied to suspension fault diagnosis system. The multi-source information fusion technology based on fuzzy logical algorithm was introduced to solve the fuzzy causality between complicated suspension faults’reason and the fault symptom which realized the accurate diagnosis of the suspension fault. The three-class progressive function model including main function of information fusion, databases and the interaction course of the various components of the information fusion was designed which started from the multi-information fusion course of the suspension fault diagnosis system. This model realized multi-level data fusion from data collection, the choice, merge and assembly of the information, to the state identification of the components, and finally to the assessment of the situation of the system. Combining with the structure feature, the multi-level fusion topology diagnosis model was designed which was a mixture of focus and distribution type. This model not only ensured the integrity of the original data, but also reduced the transmission of the redundant information so that the multi-source information diagnosis system have some feature such as higher reliability, strong robustness and good real-time property. Bedworth waterfall model was introduced on fusion model. Making full use of the powerful processing function of this model’s low-level system, then the detecting system can make full use of the original information to realize the consistency of the structure model’s orientation. 10. By means of fuzzy logic algorithm, the fault reason set suspension structure and fault reason fuzzy vector were established, and fault symptom set based on Euclidean vector of suspension fault diagnosis index and fault symptom fuzzy vector were also established. Using the fuzzy operator M (? ,⊕), put forward the fuzzy comprehensive evaluation model of suspension fault diagnosis; State parameters method was introduced to establish membership function mathematic model. At the same time, through a large number of simulation tests and real vehicle test, combining with expert experience, the typical suspension system fuzzy relations matrix was verified. Carried out comprehensive evaluation on suspension system according to‘Maximum Subordination Principle. The evaluation result showed that the specific reasons of suspension fault can be found out accurately and quickly by means of one or more symptoms. The real vehicle test also proved that all models established in the paper were correct and fuzzy logic diagnosis method based on multi-information fusion as well as the judgment criterion adopted in fault diagnosis of suspension system were reliable.The main innovations in this study are as follows:1. The wheel vibration lagging phase difference was defined as the diagnosis parameters for the first time. Regarded the multi-sensor information such as the body acceleration, the wheel relative dynamic load, the suspension dynamic deflection, the wheel vibration lagging phase difference and the suspension gap as the suspension fault diagnosis parameters. Solved the multi-factor coupling problem of suspension fault with multi-source information fusion technology which in the end achieved the goal of scientific evaluation of the suspension performance and reliable diagnosis of the suspension fault.2. The reason and the phenomenon of the suspension fault were parameterized in the study so as to quantitative evaluation of the suspension performance and quantitative analysis of the suspension fault which provided a theoretical basis and technical support for the intelligent diagnosis of the suspension system.3. Put forward a new method of suspension system fault diagnosis based on amplitude-frequency characteristic by means of Euclidean distance for the first time, and suspension system fault diagnosis model based on Euclidean distance was established for suspension system fault diagnosis. The concept of suspension goodness was presented for the first time to evaluate the virtues or defect degree of non-fault suspension performance, making the evaluation quantitative and scientific.4. The suspension gap identification mathematical model based on Fisher ordered cluster analysis was also established, therefore, a new suspension gap detecting method based on Fisher cluster analysis was presented. The best subsection coefficient was introduced as an optimization evaluation index for the first time so as to obtain the optimal subsection number of the curve to be identified. By global iterative approximation, the index can make the subsection curve get the optimal solution in the global so that the suspension gap can be identified accurately.5. Multi-information fusion theory and fuzzy mathematics diagnosis method applied to suspension fault diagnosis system. The fuzzy causal relation between complex fault reasons and fault symptoms was solved by virtue of multi- information fusion technology based on fuzzy logic algorithm, realizing suspension fault diagnosis exactly.6. Some models such as the three-class progressive multi-source information fusion function model, the multi-level fusion topology structure model, the Bedworth waterfall fusion model were introduced so as to design the key technology of the multi-source information fusion. So that the multi-source information diagnosis system have some feature such as higher reliability, strong robustness and good real-time property.In a word, the numerical simulation technology, the theory and method of cluster analysis, the multi-source information fusion technology, some advanced computer technology on fuzzy logical algorithm, the information theory and the mathematical methods applied to the researching of the suspension fault diagnosis. Some research findings have been obtained, which concern numerical modeling, algorithm applications and technology realization which had a certain guiding significant and promoting role in improving the level of vehicle detection and fault diagnosis technology as well as the intelligent diagnosis of vehicle fault. And the fault diagnosis method studied in this paper has theoretical significant and practical application value in the field of engineering applications.更多还原