Features Extraction for Sensorarray Based Pmfl Technology
|School||Nanjing University of Aeronautics and Astronautics|
|Course||Precision instruments and machinery|
|Keywords||Pulsed magnetic flux leakage testing Sensor array Finite Element Time domain analysis Feature Extraction|
Pulsed magnetic flux leakage technology is a new non-destructive testing techniques for traditional magnetic flux leakage detection technology combined with pulse theory, the principle is simple, high detection efficiency, rich frequency. The sensor array structure for rapid measurement of large areas has the advantages of reducing detect blind spots, improve the detection efficiency. In recent years, with the rapid development of sensor technology, the sensor array is more and more used in the field of signal detection. This topic will be combined with the advantages of both, the establishment of the set pulse magnetic flux leakage detection system based on the sensor array, by extracting the feature amount of the time domain signal of the sensor array, the different geometric characteristics of the defect identification. Paper, the first to use the means of finite element simulation of defects in the ideal case of leakage magnetic field signal, including the horizontal magnetic field component and the vertical magnetic field component. Through the magnetic flux leakage signal analysis of different defect model, the contact between the magnetic field component defect size. Then, according to the experimental testing environment designed a complete array of pulsed magnetic flux leakage detection system, including the incentive structure, the sensor array transducers, signal conditioning circuitry, PC software, etc., and then use the platform to the more standard specimen detection. The array output signal detected by the need for the necessary signal preprocessing, including: denoising, the average period, and fitting proposed in order to solve the sensor spatial resolution limit reductive means, etc.. As an analytical signal in the time-domain analysis of the process of extracting the peak value of each sensor, the output characteristic curve of the peak point of connection as the array. By this characteristic curve is characterized in the analysis of the geometric dimensions of the different defects, and extracts the corresponding feature quantity analysis of the causes of error. Then based on these feature amount for identifying the unknown surface defects general steps: First symmetry defect identification, followed symmetric defective width, depth, and the non-symmetry defect recognition of the inclination angle (inclination). Meanwhile, according to the design of this step of identifying a defect contour simple imaging analysis software. Simulation and experimental analysis of frequency domain characteristics of the signal, but less effective, this part of the work summarized in this paper. Finally, some suggestions for further research.