Multiple Instantaneous Frequency Ridge Extraction And Bearing Fault Diagnosis Under Variable Speed Operations

Rotating machinery,as an important power equipment,is widely used in all industrial fields,such as aerospace,ship transportation,railway transportation,energy industry,etc.Rolling bearings,as one of the most significant elements of a rotating machinery,whose damage can directly or indirectly lead to the failure of the whole system performance.Therefore,the health management and fault diagnosis of rolling bearings are indispensable to ensure the safe operation of the entire rotating mechanical system.In fact,the time-varying speed condition and variable load are the actual working state of bearing operations.In this kind of working environment rather than fixed-speed and load conditions,bearings will subject to more complex stresses,which is more likely to cause failures.However,the bearing fault diagnosis approaches under a constant speed would be ineffective for time-varying speed operations due to the blurred spectrum of fault characteristies.The key for bearing fault diagnosis under variable speed operations is the instantaneous frequency extraction,which can be directly measured by the tachometer.But this method is subject to the space and the installation position.Therefore,it is more meaningful to extract instantaneous frequencies from the signal itself for bearing fault diagnosis under time-varying speed operations.With the support of the National Natural Science Foundation of China "Bearing Fault Characteristic Extraction and Diagnosis under Time-varying Operation" and "Ridge optimization and multiscale sparse fiusion for bearing fault diagnosis under time-varying speed condition with large fluctuations"(No.51605319 and 51705349),the bearing fault diagnosis under time-varying speed operations is studied in this thesis.Aiming at bearing local faults diagnosis,a multiple instantaneous frequency ridge extraction strategy and a mean value of fault characteristic coefficients based fault diagnosis strategy are proposed in this paper.A specific theoretical introduction is given as well as both the simulated and experimental analysis.Details are as follows(1)Firstly,common fault types of rolling bearings and corresponding vibration signal characteristics are analyzed.The extraction of signal characteristics and multiple instantaneous frequencies under different fault types and the fault diagnosis strategy are the keys to bearing fault diagnosis under time-varying speed operations.Therefore,a multiple instantaneous frequency ridge extraction approach based on the frequency band separation is proposed in this paper to obtain fault characteristics and carry out bearing fault diagnosis under time-varying speed operations.Basic theories of the frequency band separation and instantaneous frequency ridge extraction are expounded.What's more,limitations of the traditional integration approach are discussed.(2)To overcome limitations of the traditional integration approach,a novel probability density distribution based instantaneous frequency ridge integration strategy is proposed.Then,a fault characteristie coefficient mean value based diagnosis strategy is proposed for bearing fault diagnosis under time-varying speed operations.The key to the instantaneous frequency ridge integration is the determination of integration intervals.Thus,this paper studies multiple initial frequency values extracted at each time bin and proposes an integration interval determination criterion based on the standard deviation.Then,the mean-based integration strategy is applied to normal intervals,while a novel integration strategy based on the probability density distribution is utilized in abnormal intervals,i.e.integration intervals.A fault diagnosis strategy based on the mean value of fault characteristic coefficients is proposed to realize the bearing fault diagnosis.The analysis of both simulated and experimental signals verifies the effectiveness of the proposed instantaneous frequency ridge integration strategy and fault diagnosis strategy.(3)Obstacles in the instantaneous frequency ridge extraction lie in the ridge extraction algorithm and the time-frequency representation aggregation,which have great impact on the ridge extraction accuracy.In order to overcome the limitation of the traditional regional peak search algorithm(RPSA),a fast path optimization algorithm is applied to the accurate extraction of instantaneous frequencies.Then,with the more accurate extracted instantaneous frequency ridge,a dynamic base angle strategy is applied to overcome two obstacles in traditional short time Fourier transform(STFT),i.e.it is hard to match the fundamental function frequencies with variable instantaneous frequencies and to select the optimal window length.The dynamic base angle strategy is applied to match the angle frequency of the fundamental function in STFT with the instantaneous frequency of the signal component,aiming at improving the aggregation of time-frequency representations and the accuracy of multiple instantaneous frequency ridge extractions.Finally,the analysis of the simulated and experimental bearing fault signals effectively validates the effectiveness of the fast path optimization algorithm and the dynamic base angle strategy for the accuracy of the instantaneous frequency ridge extraction and bearing fault diagnosis.In summary,based on multiple time-frequency ridges extracted from the lower and resonance frequency band,this paper proposes a probability density distribution based integration strategy for instantaneous frequency estimation and a mean value of fault characteristic coefficients based diagnosis strategy to realize the diagnosis of bearing fault types.In addition,the fast path optimization algorithm has helped to improve the instantaneous frequency ridge extraction accuracy.The dynamic base angle strategy helps to solve the problem that the fundamental function frequencies in linear transform do not match the time-frequency ridges of the target signal and enhance the time-frequency representation aggregation,which finally contributes to improving the accurate identification of bearing local faults transient characteristics.The bearing fault diagnosis approach based on the time-frequency analysis of the vibration signal and multiple time-frequency ridge extraction proposed in this paper makes the bearing fault diagnosis more accurate with non-tachometer and greatly contributes to the wide practical applicability of this approach.