| With the widespread use of remote belt conveyors in the field of bulk transport,roller failure has caused more and more accidents.The traditional manual inspection can no longer meet the need,and the existing contact acceleration signal detection method has the problems of large demand for sensors and difficulty in data collection,so this dissertation proposes to use intelligent inspection robot equipped with a sound picker way to carry out roller fault inspection.The operation environment of the roller is noisy,and the noise in the signal needs to be eliminated,so the Single Channel Blind Source Separation method(SCBSS)based on CEEMDAN-PCA-RobustICA is proposed.First,Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)is used to the decompose signal into Intrinsic Mode Function(IMF),and then Principal Component Analysis(PCA)is applied to determine the number of source signals.Then the signal is reconstructed.And it is separated by Robust Independent Component Analysis(RobustICA)that has faster convergence speed and stability.Through correlation calculation,the roller signal that finally has noise elimination is determined.The roller signal has the features of non-stationary and non-linearity,and the characteristic parameter extraction method of Mel-Frequency Coefficient(MFCC)can not perfectly characterize the signal.So,this dissertation proposes a feature parameter extraction method based on CEEMDAN-PCA-MFCC.First,CEEMDAN is used to decompose the signal into IMF components which have stationary characteristics,and then PCA is adopted to eliminate IMF components that have small contributions so as to reduce the amount of calculation,and then MFCC calculation is performed.Finally,the MFCC first-order difference coefficient and Delta value are used to show the dynamic characteristics of the signal.A support vector machine(SVM)is used as a classifier for fault feature recognition.The inspection scheme is combined with the roller fault identification algorithm,and build an experimental platform for verification.When the inspection robot inspects and stores the roller signal on the experimental platform,it uses RFID electronic tag to associate the signal file with the roller position.Through RFID tag to find out different types of fault roller files for algorithm verification,the average recognition rate of 95%,which confirms the effectiveness of the overall scheme. |
