Research On Fault Identification Of Drill Pipe Drill Based On EWT And Feature Fusion

As an important infrastructure engineering machinery equipment,drilling rigs are widely used in oil and gas mining,mining excavation,highways and many more fields.However the phenomenon that the drill is broken occur from time to time in the actual work process due to the destruction of the drill pipe and the loosening of the drill pipe connection.Therefore,it is important to diagnose the drill pipe drill pipe.In this paper,the vibration signal of drill pipe is taken as the research object,and the fault diagnosis of drill pipe is taken as the research topic.The fault diagnosis is carried out by empirical wavelet decomposition,fusion feature,BP neural network and SVM.The main research contents and work summary of the thesis are as follows:(1)According to the working principle of the rig and the actual fault problem,the rig experimental platform was designed and built.According to the normality of the drill pipe,the looseness of different degrees,the different degrees of failure conditions,the rotating speed and the sampling frequency that affects the vibration signal acquisition,the orthogonal experimental scheme of the drill pipe fault diagnosis is designed.The drilling program is simulated and the signal is collected according to this scheme;(2)The algorithm principles of empirical wavelet transform(EWT)and empirical mode(EMD)are studied.The results of two methods are compared to verify the superiority of the EWT method.The collected signal is denoised by wavelet threshold method,and the denoised signal is decomposed by EWT method.The time spectrum and Fourier spectrum analysis are performed on each modal component after decomposition;(3)EWT-array entropy method and EWT-energy method are proposed for the feature extraction of drill pipe vibration signal.Through these two methods,the characteristics of vibration signals under different speeds and working conditions are extracted and analyzed.The results reveal that the two methods can effectively obtain the fault characteristics of the drill pipe.The entropy and energy features of each modal component of the signal are fused and used as the classification features of fault identification;(4)The algorithm principle of BP neural network and support vector machine(SVM)is studied,and the parameter optimization of SVM kernel function based on particle swarm optimization is proposed.This paper uses these two pattern recognition methods to classify the working conditions of the drill pipe.The results show that SVM has a higher recognition rate for the fusion feature vector constructed in this paper,and it shows that the fusion feature vector constructed in this paper can be used for fault diagnosis of drill pipe.