Research On Diagnosis Algorithm Of Mechanical Equipment For Pumps

Pump machinery and equipment are widely used in various industries.For the problem of fault diagnosis,the traditional human inspection-based approach is inefficient and wasteful of resources.After analyzing the working principle and fault characteristics of pumps,this project will design fault detection methods and establish fault diagnosis models based on the vibration signal data collected by sensors.For the situation that the actual sensor acquisition is seriously disturbed by environmental noise,the wavelet transform can analyze the signal in the time-frequency domain.In this paper,the wavelet transform is used to decompose the noise-containing signal,and the noise is denoised by the difference characteristics of the modal values of the effective signal and noise corresponding to the wavelet coefficients.Combining the advantages of different traditional thresholding functions for improvement,we propose a wavelet denoising method using segmental thresholding and a new threshold function.It is demonstrated on the simulated signal that the SNR of the denoised signal obtained by the method in this paper is improved by 26.0% and 12.1% compared with hard and soft thresholding,respectively,and the mean square error is reduced by 77.0% and 47.6%,respectively.The characteristics of the vibration signal are analyzed using three approaches: time domain,frequency domain and variable modal decomposition(VMD).Since it is difficult to detect fault information and warn in advance in the spectrogram,the method of extracting feature parameters is designed to be able to characterize the information of the signal and to realize the automatic determination of the operating status by the subsequent diagnostic algorithm.The characteristic parameters include the distribution of vibration amplitude and shock characteristics in the time domain,and the characteristics of periodic components in the frequency domain structure.The variational modal decomposition has good demodulation and noise immunity for complex signals.Combining the particle swarm optimization(PSO)and cuckoo search(CS),the CS-PSO algorithm is proposed to optimize the VMD parameters.It is experimentally demonstrated that compared with empirical mode decomposition(EMD)and default parameter VMD methods,the CS-PSO optimized VMD can achieve better division in the frequency domain and retain more fault features,which constitute the input vectors of the subsequent fault diagnosis classifier.For the problem of interconnection between different features of the input vector,principal component analysis(PCA)is used to convert it into a new feature vector composed of mutually unrelated parameters for the purpose of dimensionality reduction.The fault diagnosis model is designed based on least squares support vector machine(LSSVM),and the model regularization parameter C and kernel function parameter g are optimized by CS-PSO algorithm.On the measured signals,the average accuracy of extracting only time-domain features is 79.61%,and the average accuracy of the feature extraction method in this paper is 93.84%,which is improved by 14.23%.The accuracy of the CS-PSO optimized classifier is only 0.09% lower than that of the grid search method,but the running time is reduced by 67.7% and the training speed is significantly accelerated,and the final accuracy is 94.03%.It is proved that the method in this paper can achieve fault diagnosis and type identification for the whole life cycle data including normal state to late failure.