Analysis of Passive Magnetic Inspection Signals Using the Haar Wavelet and Asymmetric Gaussian Chirplet Model (AGCM)

Abstract

Nowadays, Non-Destructive Testing (NDT) techniques are an essential foundation of infrastructure retrofit and rehabilitation plans, mainly due to the huge amount of construction, as well as the high cost of demolition and reconstruction. Modern NDT methods are moving toward automated detection methods to increase the speed and probability of detection, which enlarges the size of inspection data and raises the demand for new data analysis methods. NDT methods are divided into two main groups; active and passive. The external potentials are discharged into an object in an active method, and then the reflection wave is recorded. However, the passive methods use the self-created magnetic field of the object. Therefore, the magnetic value of ferromagnetic material in a passive method is less than the magnetic value of an active method, and defects and anomalies detection needs more variety of functional signal processing methods. The Passive Magnetic Inspection (PMI) method, as an NDT-passive technology, is used in this thesis for ferromagnetic materials quantitative assessment. The success of the PMI depends on the detection of anomalies of the passive magnetic signals, which is different for every single test. This research aims to develop appropriate signal processing methods to enhance the PMI quality of defect detection in ferromagnetic materials. This thesis has two main parts and presents two computer-based inspection data analysis methods based on the Haar wavelet and the Asymmetric Gaussian Chriplet Model (AGCM). The Passive Magnetic Inspection method (PMI) is used to scan ferromagnetic materials and produce the raw magnetic data analyzed by the Haar wavelet and AGCM. The first part of this study describes the Haar wavelet method for rebar defect detection. The Haar wavelet is used to analyze the PMI magnetic data of the embedded reinforcement steel rebar. The corrugated surface of reinforcing steel makes the detection of defects harder than in flat plates. The up and down shape of the Haar wavelet function can filter the repeating corrugations effect of steel rebars on the PMI signal and thereby better identify the defects. Toogood Pond Dam piers’ rebar defects, as a case study, were detected using the Haar wavelet analysis and verified by the Absolute Gradient (AG) method using visual comparison of the resultant signals and the correlation coefficient. The predicted number of points with a rebar area loss higher than 4% is generally the same with the AG and the Haar wavelet methods. The mean correlation coefficient between the signals analyzed using the AG and the Haar wavelet for all rebars is 0.8. In the second part of this study the use of the AGCM to simulate PMI signals is investigated. Three rail samples were scanned to extract a three-dimensional magnetic field along specific PMI transit lines of each sample for the AGCM simulations. Errors, defined as the absolute value of the difference between signal and simulation, were considered as a measure of simulation accuracy in each direction. The samples’ lengths differed, therefore error values were normalized with respect to the length to scale data for the three samples. The Simulation Error Factor (SEF) was used to measure the error and sample 3 showed the lower value. Finally, statistical properties of the samples' SEF, such as standard deviation and covariance, were evaluated, and the best distribution was fitted to each of the data sets based on the Probability Paper Plot (PPP) method. The Log-Normal probability distribution demonstrated the best compatibility with SEF values. These distributions and statistical properties help to detect outlier data for future data sets and to identify defects.