Research On Data Processing Methods Based On Visible Eddy Current Testing

Visible eddy current testing is a rapid,non-contact and high sensitive non-destructive testing(NDT)technology with great development prospect.The NDT technique has been utilized to detect damages of conductive or composite materials in aerospace field,electrical power system,mechanical manufacturing etc..However,the current visible eddy current testing technology is limited by some difficulties,such as harsh testing environment decreasing sensor sensitivity,collected signals with low SNR(Signal to Noise Ratio),big data problem etc..Feature extraction of recorded data,which is an important part in visible eddy current testing has been widely researched.But the current feature extraction methods still need to be improved to satisfy the requirements of application.In this thesis,several feature extraction algorithms have been proposed to improve defect testing efficiency and detection accuracy.The main research content and innovation points of this thesis are as follows:(1)A fast feature extraction algorithm in eddy current pulsed thermography(ECPT)testing is developed to realize defect feature extraction rapidly.The proposed feature extraction algorithm includes a data block segmentation,a variable interval search,a correlation value classification and a between-class distance decision function.Data block segmentation and variable interval search are firstly combined to reduce the repetitive calculation in automatic defect identification.The classification method and between-class distance are used to select the typical features of thermographic sequence.This method is not only able to extract the main image information,but also can reduce the time of thermographic sequence processing to improve the detection efficiency.Experiments and comparisons are provided to demonstrate the capabilities and benefits(i.e.reducing the processing time)of the proposed algorithm in automatic defect identification.(2)A new feature extraction method based on thermography is proposed to enhance quantitative defect information.The proposed method included local(element-wise)sparse and low rank decomposition,image fusion etc.can increase the contrast of defect area and background and extract more useful defect features than other two common feature extraction algorithms in ECPT.The experiments including comparison results are provided to demonstrate the capabilities and benefits of the proposed algorithm.More meaningful defect information of the experimental specimens is reserved from raw ECPT data and the background area is suppressed severely compared with other feature extraction algorithms.(3)One thermal process separation strategy is developed to improve the performance of defect feature extraction method based on local sparse components and image fusion.This new proposed method combines the physical meanings of initial image sequence with the mathematics of the local sparse component evaluation.The defined fusion criterion integrates defect features into one feature image and abandon useless information.Experimental results show that this developed method can retain the thermal characteristics of defects at different stages and suppress background interference.(4)A novel deep learning model based on multi-frequencies eddy current testing for defect detection is proposed and investigated in this thesis.The proposed model based on a Convolutional Neural Network(CNN)is developed to improve defect detection performance with uncertainty information.The novelty of this work consists in combining characteristics of ECT data with general deep learning model to improve performance of deep learning in ECT field.This model is realized by a region of interest method based on robust principle component analysis,a CNN classification model with weighted loss function and measurement of uncertainties.Experimental dataset obtained from eddy current inspection of heat exchanger tubes is utilized to validate the detection performance improvement.As a result,both the classification accuracy and the percentage of defects correctly identified have been increased.