Search

The traditional identification method of surface crack has some shortcomings such as detection equipment complex, high requirements for operator and not suitable for large-scale structure inspection, vibration method is applied to identify surface crack in steel rods in this paper, using the method of calculating modal analysis combined with time-frequency wavelet analysis. The research on damage identification is performed on steel rods with different deep crack. The research results show that the accurate damage location can be judged by wavelet modulus maximum and the quantitative analysis of single crack identification can be achieved by constructing mathematical expressions of a single crack damage degree and damage index Lipschitz index. The research would province some guidance for engineering applications.

During bars manufacturing and using process, the use of bars with cracks influenced the service life and sustainable work, on the premise of certain detection system, the choose of signal de-noising method plays an important part in magnetic flux leakage testing process, for which determining the global performance of testing equipment. This paper studies on magnetic flux leakage signals characteristic of cracks on the bar, analyzed the advantages and disadvantages of wavelet analysis and wavelet packet threshold de-noising, and on this basis, in order to give consideration to both the low and high frequency interference signal removing, adopted an integration method that combines wavelet packet threshold de-noising, wavelet packet decomposition and wavelet reconstruction, which greatly improved signal processing results.

The steel rod is an important part for project fields , and it is large-scale to be used. it is apt to crack, corrosion and so on in the poor working conditions. In order to recognize correctly the type of defects, a method was presented to extract frequency band energy feature by using wavelet package decomposition. In the meantime, to extract the peak-peak value in the time-domain and make the mixed feature vector. With the way of pattern recognition, the best recognition way was got by comparing the BP artificial neural network(ANN), PNN(probability neural network) artificial neural network and "one-versus-one" support vector machine(SVM).The result showed that the recognition rate of SVM was more suitable for defects’ identification in steel rod.

Sucker rod is the hey component for the rod pumping system. In the working process, there are mainly four kind of defects: cracks, corrosion, partial wear and artificial damage, which will cause downhole accidents, so feature extraction is a premise to distinguish different defects. In the paper, by using wavelet energy spectrum, sucker rod defects signals that came from magnetic flux leakage (MFL) detection method are decomposed to the different frequency bands by arbitrary time-frequency resolution, the frequency signal energy is an eigenvectors to judgment sucker rod defects. In practice, the detection method is effective and feasible, and there is a new way for the sucker rod defects pattern recognition.