Research On Time Synchronous Averaging Technique With Application To Condition Monitoring Of Rotating Machinery

With the rapid development of technology and modern industry,the structure of machinery is becoming increasingly large,high-speed,automation and integration.As there are tight connections inside and between the machineries,once the key compo-nents fail,the machine condition will be affected or even collapse,which will result in huge economic losses and even casualties.In order to prevent such serious conse-quences,anomaly monitoring of machine condition has been proposed and continu-ously developed.Since the machinery usually operates under complex working conditions,rotational speed will somewhat randomly fluctuated due to changes of load and/or environment,which results in the collected condition signals to be non-stationary,i.e.quasi-periodic signal.The traditional monitoring method is presented under the assumption of con-stant rotational speed,i.e.,the strict first-order cyclostationary,and thus they are pow-erless to the modulation of such kind of signal.Therefore,it is of great engineering application value to investigate the abnormal condition monitoring method suitable for complex working conditions.Based on the vibration response of machinery,this study carried out a series of research with respect to signal processing,modeling analysis and state recognition.1)This study first elaborated the background and significance of abnormal condi-tion monitoring for machinery,and outlined the related main steps and technologies.Among them,the domestic and foreign research status of non-stationary signal pro-cessing technology in the step of health information extraction was mainly analyzed and summarized.2)Based on the physical structure and kinematic properties of machinery,its vibra-tion response was analyzed.Then,from the perspective of cyclostationary,this study determined the average time waveform based framework for abnormal condition mon-itoring of machinery.Moreover,the modulation process of first-order cyclostationary signals under complex operating condition and its property pseudo-periodic are an-alyzed.On this basis,a novel time synchronous averaging method based on phase alignment is proposed for the extraction of average time waveform of non-stationary signals.3)The technique of dynamic time warping is introduced,and the time synchronous averaging technique suitable for non-stationary signals is realized based on the prop-erty of feature alignment of optimal warping path.The extraction of average time waveform is completed and its effectiveness is verified by simulation and real engi-neering application.Based on the above work,the anomaly measurement and decision making were further designed to form a completed approach for abnormal condition monitoring of rotating machinery.4)The proposed method is applied to real engineering applications of typical ro-tating machinery,and compared with four common abnormal monitoring methods.The monitoring results are analyzed and the commonly used indicators(recall rate,precision rate,F?)in classification task of machine learning are employed to compre-hensively evaluate the algorithm performance.The experimental results demonstrated the effectiveness and superiority of the proposed method,and show that the proposed method:(1)pure data-driven and does not require specific hardware support;(2)does not require reference models and prior knowledge;(3)can monitoring in online and real-time manner,so the method can be applied in the real applications.