Fault Diagnosis Method Of Rotating Machinery Based On Time-Frequency Ridge Extraction

Rotating machinery plays an important role in the fields of manufacturing,electric power and railway.With the rapid development of our country,the application of rotating machinery is becoming more and more extensive.The service environment of rotating machinery is usually harsh,the operating conditions are unstable,and it often faced with extreme environments such as high temperature,high temperature,high speed,and heavy load,which make the rotating machinery parts inevitably fail,resulting in huge economic losses and casualties.Therefore,the condition monitoring and fault diagnosis of rotating machinery can effectively ensure the safe and reliable operation of the equipment and avoid economic losses and casualties.When the rotating machinery is running at a constant speed,the frequency domain analysis can used to analyze the vibration signal generated by the rotating machinery.When a rotating mechanical component fails,the associated fault characteristic frequencies appear in the frequency domain.However,the main components in rotating machinery such as bearings and gearboxes often operate under variable speed conditions.Under the influence of excitation such as variable speed,the vibration signals generated by key transmission components such as bearings and gearboxes usually accompanied by frequency modulation.The vibration signal produces spectral line aliasing in the frequency domain,which affects the diagnosis of the fault location and degree of mechanical components.The order tracking technology can make the spectral lines in the order spectrum no longer change with the rotational speed through resampling in the angular domain,so the order tracking technology can used for fault diagnosis of rotating machinery under variable rotational speed conditions.However,the order tracking technology requires the rotation speed of the rotating shaft for angular domain resampling,and in actual production and life,it is difficult to meet the needs of installing a tachometer on the rotating shaft due to space constraints and other reasons.Tachometer-less order tracking technology is an effective method for fault diagnosis of rotating machinery,which can identify the characteristic orders of rotating machinery from vibration signals without a tachometer.The key of the tachometer-less order tracking technology is to extract the rotational speed of the rotating machine from the vibration signal for order tracking,so a method that can estimate the rotational speed urgently needed.After the time-frequency representation(TFR)of the vibration signal obtained by using the time-frequency analysis method,the characteristic ridge of the signal component is extracted from the TFR,which be used to estimate the instantaneous frequency of the fault feature of the vibration signal,and then converted into the rotational speed for order tracking.This paper takes ball bearings and fixed-axis gearboxes as the research objects,and conducts in-depth research on their condition monitoring and fault diagnosis methods.The main research contents divided into the following aspects:1.Several existing time-frequency ridge extraction methods reviewed,and the applicable scope,advantages and disadvantages of various methods compared.2.An adaptive cost function ridge extraction(ACFRE)method is proposed.The method re-establishes the mathematical model of the cost function,and comprehensively considers the balance between the amplitude of the ridge and the continuity of the ridge.Redefines the penalty factor in the cost function,so that the algorithm can adaptively adjust the penalty for frequency hopping.The full band search carried out in the TFR of the signal to avoid unreasonable band constraints.The ACFRE method can automatically extract the time-frequency ridges related to the signal characteristic components from the TFR,and has strong adaptability.The algorithm improves the defects such as unreasonable penalty factor and no well-defined search bandwidth in the original cost function.The vibration signals collected by the mechanical fault simulation test-bed used to test the algorithm,and compared with the original CF method.The analysis results show the superiority of the proposed method.3.An improved dynamic path ridge extraction(IDPRE)method is proposed.The algorithm first extracts an initial time-frequency ridge from TFR,obtains the relevant frequency difference,and uses a box plot to remove outliers in the frequency difference.The definition of time-frequency support introduced to automatically obtain the upper and lower boundaries of the initial time-frequency ridge,and then calculate the search bandwidth of the time-frequency ridge from the upper and lower boundaries.Establish a penalty function for extracting the time-frequency ridge according to parameters such as frequency difference and search bandwidth,and comprehensively calculate all potential paths in the TFR and select the optimal path.The algorithm obtains parameters according to the characteristics of the signal itself,so it does not need to manually adjust other parameters,and can automatically and iteratively extract multiple time-frequency ridges from the TFR,which is highly adaptive.The algorithm improves the defects of the original dynamic path algorithm,such as the frequency median error constraint,the inability to extract the fast time-varying time-frequency ridge,and the vulnerability to noise interference.The algorithm used to test the simulated and experimental signals,and the results verify the effectiveness of the method.