Features Extraction For Sensorarray Based Pmfl Technology

Pulsed magnetic flux leakage (PMFL) is a new kind of non-destructive testing (NDT) technology which combines the conventional magnetic flux leakage (MFL) technology with the theory of pulsed current. The sensor array may scan large areas with a higher speed and lower lapsus than single sensor testing. With the development of sensor integration technology, sensor array has being more and more applied to the NDT. This paper focuses on features extractions based on sensor array in time-domain, which may detect the defects by PMFL.First of all, FEM is used to get the ideal signal of MFL with defects,including signal of horizon and signal of vertical.According to the analysis of diferent signal of MFL with defects,the relation between magnetic signal and defect size is received.Then a full set of PMFL system based on sensor array is designed according to the lab's detect environment,such as exicited structure,sensor array of probe,signal conditonal circuit,software of IBM-PC.Later, the platform is used to dectect the standard specimens. In order to eliminate the noise and extract the feature, the detecting signal is analyzed and processed by some pre-treatments,such as wavelet, fitting curve and so on. In the process of time-domain analysis,every peak value of sensor signal is extracted as analysing signal,then peak values are lined as output signal of the sensor array. By analysing different characters of curve under different defect sizes, and getting corresponding characteristic quantities,the reasons of the errors is discussed. According to these characteristic quantities,the steps of detecting these unknown surface defect is proposed as follows: first,defect of symmetry is detcted, then the defect width and depth of symmetry defects, and identification the incline angle of non-symmetry defect. At the mean time, a software that is used to image of the defect has also been designed.The features extraction in frequncy-domain is also been carryed out both in FEM and experiments, but there's less useful signal characteristic been gained. And further research work has been recommended.