Research On Non-stationary Signal Analysis Method Based On LMS-STFT-SPWVD Fusion Algorithm

Because vibration signal data generated by construction vehicles are mostly non-stationary signals,traditional methods have many problems in analyzing such signal data,and it is difficult to fully excavate its local characteristics.Therefore,according to the characteristics of vibration time domain signal waveform,this paper constructed a non-stationary signal analysis method based on the LMS-STFT-SPWVD fusion algorithm,and carried out a series of research work.The main research contents are as follows:Firstly,the basic principles of STFT and WVD are introduced.According to the shortcomings of these two algorithms,time-frequency resolution and time-frequency aggregation cannot be achieved simultaneously,SPWVD time-frequency analysis method is proposed.By comparing the results of STFT,WVD and time window parameter analysis,the time window parameters are determined and applied to strong noise signals.The results show that SPWVD overcomes the defects of STFT and WVD algorithms,but the inhibitory effects of mode aliasing and cross frequency interference signals need to be improved.The main defects of SPWVD are mode aliasing and poor anti-noise performance,while LMS filter has better interference signal and mode aliasing suppression effect.Therefore,this paper combines the two algorithms and proposes the LMS-SPWVD time-frequency analysis method.The time-frequency analysis results of SPWVD,LMS-STFT and LMS-WVD were compared to verify the effectiveness of the algorithm,and the noise signals of different LMS parameters and the effect of mode aliasing suppression were compared to determine the parameters of the LMS filter algorithm,and then it was applied to the strong noise environment.The results show that when the order of weight matrix is 128 and the step size is 10e^-4,the LMS-SPWVD time-frequency analysis effect is better.In this paper,according to the characteristics of the consistency of the dimension of the time-frequency image data of LMS-STFT and LMS-SPWVD,aiming at the disadvantages of LMS-SPWVD,such as weak anti-interference ability,divergence of the signal center frequency energy,and serious mode mixing,a binary fusion algorithm of LMS-STFT-SPWVD is proposed by using the image binarization pretreatment technology.Firstly,the fusion effects of LMS-STFT,LMS-SPWVD,STFT-WVD and STFT-SPWVD were compared to verify their effectiveness and the combined fusion effects of the three threshold parameters were compared to determine the range of threshold parameters.Then,the fusion results were applied to time-frequency analysis images with strong noise.The results show that when the value range of LMS-STFT threshold parameter is（0.72,0.76）and the value range of LMS-SPWVD threshold parameter is（0.52,0.56）,the center frequency energy recognition effect and anti-noise performance are better,and the image fusion effect is better.