Research On The Methods Of Rotating Machinery Fault Feature Extraction Under Complex Conditions

Rotating machinery is the most common mechanical equipment in the industrial sector.With the rapid development of modern industry and science and technology,the application demands of rotating machinery become increasingly complicated and diversified.Under the impact of adverse running environment,complex and unsteady operating condition,alternating stress and other random factors,it may inevitably suffer from some faults.In this way,its normal operation may be impaired,and even a major safety accident may happen.Therefore,the study on state monitoring and fault diagnosis technique of rotating machinery is attracting more and more attention,and the feature extraction is among the technical difficulties and the most important and crucial aspects as well.As the vibration signals of the rotating machinery contain a wealth of fault information,it is of great significance to study the rotating machinery feature extraction based on vibration signal analysis.Unlike the fault feature extraction of rotating machinery under the steady and no interference ideal operating conditions,the wideband random noises caused by long-path signal transmission during test and noises from other vibration sources of the same frequency in the mechanical system bring about more serious interference,and the vibration signals representing the running status of the rotating machinery may be easily swallowed up by strong noises,so the fault features are so weak to extract,under such adverse operating conditions as high speed and heavy load;under the variable operating conditions like variable load and variable speed,the rotating machinery signals are more complex unsteady signals,and the reflection of changes in running conditions at the signal feature level is very similar to changes in signal features caused by faults,so some difficulty exists in fault feature extraction;under some special operating conditions,which are subject to such restrictions as enclosed environment,rotating operation and spatial distance,the rotating machinery needs to be placed under wireless status monitoring.However,the wireless channel bandwidth is limited and cannot satisfy the demands of real-time transmission and online monitoring of massive vibration data.Transmission and monitoring signal feature is a feasible alternative method.In order to realize online real-time monitoring of rotating machinery by signal feature,greater rapidity of feature extraction on the chip of wireless sensor node is always required.Thus,it has a great challenge to extract fault features of rotating machinery under complex operating conditions based on vibration signal analysis.To solve the fault feature extraction difficulty of the rotating machinery under the complex operating conditions above,the thesis carried out studies on the weak fault feature extraction of the rotating machinery based on the double-window spectrogram fusion enhancement,the fault feature extraction of the rotating machinery under variable operating conditions based on the pulse adaptive time-frequency transform and the on chip feature extraction for wireless sensor node based on the feature group short-time shifted discrete Fourier transform(DFT),as follows:(1)To solve the difficulty in extracting the weak fault features of the rotating machinery in the presence of strong noises,the thesis proposes the weak fault feature extraction method of the rotating machinery based on the double-window spectrogram fusion enhanced.The method first selects the optimal resonance demodulation frequency band to exclude noise signals other than those of selected frequency band,and for the noise interference in signals within the pass band,constructs the long-time and short-time window STFTs to acquire the high frequency resolution time frequency spectrum and high time resolution time frequency spectrum respectively.It uses the spectrum pre-processing method to preliminarily reduce noises in signals within the pass band and highlight the signal feature frequencies,the spectrum-related analysis and noise reduction method to further inhibit random noises within the pass bank,and the spectrum amplitude related enhancement method to further enhance the signal feature frequency amplitude,so as to extract the weak fault features of the rotating machinery in the presence of strong noises.(2)To solve the difficulty in extracting the fault features of the rotating machinery under the variable operating conditions,the thesis proposes the fault feature extraction method of the rotating machinery under the variable operating conditions based on the pulse adaptive time frequency transform.The method first uses the shock pulse method to extract the shock pulse sequence containing fault information,and through the proposed pulse adaptive time frequency transform method,transform the shock pulse of one-dimension time sequence into two-dimension time frequency domain,and extract the effective instantaneous frequencies with the sudden load shock and other noise interference being effectively inhibited;and then it employs the proposed pulse order tracking technology to transform the time-frequency domain into the time-order domain under the angular domain free resampling distortion,succeed in eliminating the impact from variable rotating speed,and achieve the fault feature extraction of the rotating machinery under the variable operating conditions.(3)To solve the difficulty in realizing the real-time data transmission and online monitoring of massive vibration data acquired from rotating machinery under wireless status monitoring,the signal feature transmission is a feasible alternative method.To ensure on-line real-time monitoring of the status of rotating machinery by means of signal feature transmission,the thesis proposes the on chip quick feature extraction method for wireless sensor node based on the feature group short-time shifted DFT.The method focuses on extraction of the discrete feature frequency points containing running status information of the rotating machinery,realizes synchronous sampling and complete cycle truncation analysis of the multi-component vibration signals of the rotating machinery by constructing the sampling frequency and sampling length for the synchronous whole cycle sample processing,avoids such problems as frequency spectrum leakage and barrier effect existing in the signal analysis method based on Fourier transform,and thus improves the signal feature extraction accuracy;it resamples the sampling data by constructing the resampling frequency for shift operation,proposes the shifted DFT and short-time shifted DFT,replaces the multiply operation with the shift operation,and realizes quick vibration signal feature extraction of the rotating machinery;at last,it constructs the helical array diagram,which will serve as an intuitive graphical method for realizing shifted DFT operation.(4)To commercialize the research findings above,and overcome the deficiencies in the present status monitoring and fault diagnosis system of the rotating machinery in terms of generality,derivability etc.,the thesis constructs a general fault feature extraction software platform for the rotating machinery.By focusing on realization of the above-mentioned three feature extraction methods,taking the software generality as the design guide,adopting the object-oriented software design philosophy,and starting from software demand analysis,the top-down principle is followed for software architecture design and modular design,and a general fault feature extraction software platform for the rotating machinery is finally developed.By modifying parameter configuration alone,the software platform may be quickly adapted to data acquisition and fault feature extraction of different objects and signals.At last,the thesis gives a summary of the three methods as discussed herein,and looks into the study directions in the future.