Search

This paper presents a technique for solving inverse problems in magnetostatic nondestructive evaluation (NDE), using finite element models. In-line inspection of ferromagnetic gas pipelines containing pipe-wall defects, is chosen as the candidate NDE process. The signal inversion technique consists of iteratively solving the forward problem by updating the finite element mesh, rather than the material properties of the finite elements. Preliminary simulation results obtained using a 2D finite element model are presented.

The magnetic flux leakage method, used for nondestructive evaluation of ferromagnetic objects, generates greyscale images that are representative of the integrity of the specimen. Defective areas typically appear as bright regions in the image. Unfortunately, the task of defect characterization becomes more challenging due to the effects of variations in the test parameters associated with the experiment. One such test parameter is the permeability of the test object. Conventional invariant pattern recognition algorithms are not capable of performing invariance transformations to compensate for such variations. This paper describes a novel technique that uses wavelet basis functions to provide selective invariant features and eliminate image intensity variations from undesirable changes in operational variables. The performance of the invariance transformation is demonstrated by applying the method to magnetic flux leakage images obtained using a finite element simulation of in-line inspection of natural gas transmission pipelines.

Magnetic flux leakage (MFL) methods are used extensively for inspecting ferromagnetic materials. The analysis of the MFL signal is however fraught with problems associated with the sensitivity of the signal to a number of factors such as the MFL sensor velocity and variations in the permeability of the test specimen. The interpretation can be simplified if the signals can be processed to attain invariance to these conditions. This paper presents novel methods for obtaining permeability invariant and velocity invariant MFL signals.

In this paper, a novel method for modeling the B-H characteristics of ferromagnetic materials is discussed. The method finds application in numerical finite element models for computing the leakage fields in magnetostatic nondestructive testing. The method is computationally simple and employs a neural network approach using Radial Basis Functions for representing the hysteresis loop as a closed curve and expands a periodic contour function in a Fourier series. The second technique, though parsimonious in terms of data storage requirements, is computationally more intensive.

This paper presents a technique that can be used to fuse data from multiple sensors that are employed in nondestructive evaluation (NDE) applications, specifically for the in-line inspection of gas transmission pipelines. A radial basis function artificial neural network is used to perform geometric transformations on data obtained from multiple sources. The technique allows the user to define the redundant and complementary information present in the data sets. The efficacy of the algorithm is demonstrated using experimental images obtained from the NDE of a test specimen suite using magnetic flux leakage (MFL), ultrasonic (UT) and thermal imaging methods. The results presented in this paper indicate that neural network based geometric transformation algorithms show considerable promise in multi-sensor data fusion applications.

Several measurement modalities have been developed over the years for various nondestructive testing and evaluation (NDT and E) applications, such as ultrasonic, magnetic flux leakage, and eddy current testing, all of which have been used extensively in pipeline defect identification. While it is generally believed that different testing modalities provide complementary information, only a single testing modality is typically used for a given application. This is part due to lack of effective, computationally feasible data fusion algorithms that are applicable to NDT and E signals. Such an algorithm capable of data fusion can combine information from two or more different sources of data, giving more insight and confidence to the data analysis than a decision that would otherwise be based on either of the sources alone. Learn++, previously introduced as an incremental learning algorithm, was applied to a NDT and E data fusion application. Specifically, we generated two ensembles of classifiers, one trained on ultrasonic signals, and the other on corresponding magnetic flux leakage signals obtained from stainless steel samples that contained five classes of discontinuities: crack, pitting, weld, mechanical damage, and no discontinuity. We have observed that the prediction ability of the automated classification system, as measured by the accuracy and reliability of the classification performance on validation data, was significantly improved when the two data sources were combined using Learn++.

The objectives of this research project are: (1) To design sensor data fusion algorithms that can synergistically combine defect related information from heterogeneous sensors used in gas pipeline inspection for reliably and accurately predicting the condition of the pipe-wall. (2) To develop efficient data management techniques for signals obtained during multisensor interrogation of a gas pipeline. During this reporting period, Rowan University fabricated test specimens with simulated defects for nondestructive evaluation (NDE); designed and developed two versions of a test platform for performing multi-sensor interrogation of test specimens under loaded conditions simulating pressurized gas pipelines; and performed magnetic flux leakage (MFL), ultrasonic testing (UT), thermal imaging and acoustic emission (AE) NDE on the test specimens. The data resulting from this work will be employed for designing multi-sensor data fusion algorithms.

This paper presents a novel approach for training a multiresolution, hierarchical wavelet basis function neural network. Such a network can be employed for characterizing defects in gas pipelines which are inspected using the magnetic flux leakage method of nondestructive testing. The results indicate that significant advantages over other neural network based defect characterization schemes could be obtained, in that the accuracy of the predicted defect profile can be controlled by the resolution of the network. The centers of the basis functions are calculated using a dyadic expansion scheme and a hybrid learning method. The performance of the network is demonstrated by predicting defect profiles from experimental magnetic flux leakage signals.

The four engineering departments (electrical, mechanical, chemical and civil) at Rowan University (USA) rely extensively on semester-long project modules called "Clinics" to provide practical applications for coursework taught within the respective departments. The engineering clinics are typically team-taught by faculty from more than one discipline; the student project teams are also multidisciplinary. The focus is on product design and development, students and faculty bring skills acquired from their respective disciplines to bear on the design process. This paper describes a sophomore level clinic experience wherein students and faculty from the Electrical and Mechanical Engineering departments collaborated to design and develop a portable device that could be used to nondestructively inspect aircraft skin for cracks.

A new class of invariant pattern recognition algorithms is required for interpreting nondestructive evaluation signals that occur during in-line inspection of components with varying material properties. This paper presents the theoretical development of these invariance algorithms and provides experimental validation of these algorithms using applications in magnetic flux leakage NDE and ultrasound NDE.

Signals that are obtained in a variety of nondestructive evaluation (NDE) processes capture information not only about the characteristics of the flaw, but also reflect variations in the specimen's material properties. Such signal changes may be viewed as anomalies that could obscure defect related information. An example of this situation occurs during in-line inspection of gas transmission pipelines. The magnetic flux leakage (MFL) method is used to conduct noninvasive measurements of the integrity of the pipe-wall. The MFL signals contain information both about the permeability of the pipe-wall and the dimensions of the flaw. Similar operational effects can be found in other NDE processes. This paper presents algorithms to render NDE signals invariant to selected test parameters, while retaining defect related information. Wavelet transform based neural network techniques are employed to develop the invariance algorithms. The invariance transformation is shown to be a necessary pre-processing step for subsequent defect characterization and visualization schemes. Results demonstrating the successful application of the method are presented.© (1996) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Natural gas, which is one of the nation’s cheapest forms of energy, is transported to consumer sites via a vast transmission pipeline network .Safety considerations and a desire to assure uninterrupted energy supply require that the pipelines be inspected periodically. The effective detection of defects in the pipeline is vital to assure the integrity of the transmission systems. A variety of nondestructive evaluation techniques (NDE), such as ultrasonic, eddy current, and magnetic flux leakage (MFL) methods have been employed to detect defects in gas pipelines [1]. Among these methods, the MFL method represents the commonly used technique.

Gas transmission pipelines are often inspected and monitored using the magnetic flux leakage method [1]. An inspection vehicle known as a “pig” is launched into the pipeline and conveyed along the pipe by the pressure of natural gas. The pig contains a magnetizer, an array of sensors and a microprocessor-based data acquisition system for logging data. The data is subsequently retrieved and analyzed offline. The pipeline inspection results in the generation of a vast amount of data — in excess of 4 GB, even in compressed form. It is important that these data are presented in a suitable manner for evaluation by trained operator. Virtual reality (VR) display techniques represent an attractive mechanism for presenting this huge amount of data effectively. The application of VR techniques enables the operator to explore the virtual environment generated by the computer. This technique can serve as an important bridge between human operator and the computer. In this paper, we present some preliminary efforts in achieving this interface.