航空发动机故障诊断与振动预测技术研究

发布时间：2018-06-29 12:37

本文选题：航空发动机 + 故障诊断　；参考：《西北工业大学》2015年硕士论文

【摘要】：随着航空发动机结构日益复杂,性能不断提高,其振动故障也越发多样,危害越来越严重,直接导致发动机可靠性降低。航空发动机故障预测与健康管理系统(PHM)是先进发动机的标志和关键技术之一。美国自上世纪50年代开始研究航空发动机预测健康管理技术,并在F119发动机上配装PHM系统,发挥了重要的作用。因此开展航空发动机故障诊断和振动预测研究对提高发动机可靠性,推动我国发动机综合设计水平的提高具有重要意义。本文将发动机历史数据的数理统计、动力学分析和振动预测方法有机结合,形成统计模型、诊断模型、预测模型和验证模型四个联系的整体,开展研究工作。主要工作内容包括基于统计的发动机状态识别、基于动力学模型的发动机故障辨识和发动机振动预测方法研究。基于统计的状态识别研究了四种统计影响因素与振动的关系,在此基础上得到了振动限制值定限方法并提出了发动机正常状态和故障状态的识别方法。基于动力学模型的故障辨识从故障机理、特征频率和振动特征研究了三种典型故障——叶片结垢故障、双源拍振故障和叶片掉块故障,并提出了这三种典型故障的振动趋势模型和故障辨识准则。整机振动预测方法的建立是在前二者的基础之上,统计影响因素与振动的关系、故障辨析准则和趋势模型共同指导形成了振动预测策略,得到了工况平稳条件下的五种预测模型和过渡态预测模型。最后,以一组发动机实测数据验证了本文的方法和流程,考核算法,从侧面反映了本文的工程价值。通过上述研究,主要得到了以下结论:(1)一次试车中,相同工况下,正常状态下振动幅值符合正态分布。多次经历相同转速,振动幅值均值不同,属于不同的工况,在诊断和预测过程中应该区别讨论;同次装配不同次试车,同测点同工况下,振动幅值方差无显著差异,可为幅值波动范围的选择和限制值定限提供参考。(2)多源耦合拍振、振幅缓增和振幅突变故障模式是航空发动机最常见的几种异常振动现象。文章建立的模式辨别算法对于周期摆动案例、转子热弯曲案例和双稳态振动案例的识别值分别为rt=3.06、rt=87.01和rt=4.97,均在模型辨别的接受域范围内,说明了本文提出的故障辨别准则的有效性。(3)基于模型的振动趋势预测方法能够正确预测,双源拍振与线性缓增案例预测的平均相对误差均小于10%,振幅突增案例的平均相对误差均小于20%,预测效果良好。标准仪器验证和实测数据验证结果表明,本文所提出的方法正确,结果精度在工程可接受的范围内,对发动机故障诊断和振动预测提供了有力的支持。
[Abstract]:With the increasingly complex structure and improved performance of aero-engine, its vibration faults become more and more diverse, and the harm becomes more and more serious, which directly leads to the reduction of engine reliability. Aeroengine Fault Prediction and Health Management system (PHM) is one of the key technologies of advanced engine. Since 1950s, the United States has studied the technology of aero-engine predictive health management, and installed PHM system on F119 engine, which plays an important role. Therefore, the research of aeroengine fault diagnosis and vibration prediction is of great significance to improve the reliability of the engine and promote the improvement of the comprehensive design level of the engine in China. In this paper, the methods of mathematical statistics, dynamic analysis and vibration prediction of engine historical data are organically combined to form a whole of statistical model, diagnostic model, prediction model and verification model. The main work includes engine state recognition based on statistics, engine fault identification based on dynamic model and engine vibration prediction method. The relationship between four kinds of statistical factors and vibration is studied based on statistical state recognition. Based on this, the vibration limit method is obtained and the identification method of engine normal state and fault state is proposed. The fault identification based on dynamic model is studied from fault mechanism, characteristic frequency and vibration characteristics. Three kinds of typical fail-blade scaling fault, double source beat fault and blade block falling fault are studied. The vibration trend model and fault identification criterion of these three typical faults are presented. The whole machine vibration prediction method is established on the basis of the former two methods, the relationship between the influencing factors and the vibration, the fault discrimination criterion and the trend model together guide the vibration prediction strategy. Five prediction models and transition prediction models are obtained. Finally, the method and flow of this paper are verified by a group of engine measured data. The evaluation algorithm reflects the engineering value of this paper from the side. The main conclusions are as follows: (1) in a test run, the amplitude of vibration under the same working conditions conforms to the normal distribution. It should be discussed differently in the process of diagnosis and prediction, the vibration amplitude variance has no significant difference in the same test run and the same measuring point under the same working condition, the vibration amplitude and mean value are different from each other in the process of diagnosis and prediction, the vibration amplitude and mean value are different from each other in the same rotating speed and at the same measuring point. It can provide a reference for the selection of amplitude fluctuation range and the limit of limiting value. (2) the fault modes of multi-source coupled beat vibration, slow amplitude increase and amplitude abrupt change are the most common abnormal vibration phenomena of aero-engine. For periodic swing cases, rotor thermal bending cases and bistable vibration cases, the recognition values of the proposed pattern discrimination algorithm are RTD 3.06 and RT 4.97, respectively, which are within the range of the model discrimination acceptance region. The effectiveness of the fault discrimination criterion proposed in this paper is illustrated. (3) the vibration trend prediction method based on the model can predict correctly. The average relative error of case prediction of double source beat vibration and linear slow increase is less than 10 and the average relative error of amplitude surge case is less than 20. The prediction effect is good. The verification results of standard instruments and measured data show that the method presented in this paper is correct and the accuracy of the results is within the acceptable range of engineering. It provides a powerful support for engine fault diagnosis and vibration prediction.
【学位授予单位】：西北工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V263.6

【相似文献】