Application of deep learning procedure to magnetic multi-sensor matrix transducer data for the need of defect characterization in steel elements

Review on condition monitoring techniques for water pipelines

“…Measurements are recorded from multiple sensors and data fusion is performed to improve the efficiency of defects characterisation [54]. Features of magnetic field are fused, and a machine learning algorithm is applied to evaluate the model [55,56,57,58]. A multi-sensor data fusion framework is proposed to better estimate the spatial distribution of magnetic field, and improved conditional Generative Adversarial Nets (icGANs) are designed to enhance MFL information [59].…”

Multi-Sensor Data Integration Using Deep Learning for Characterization of Defects in Steel Elements

“…Taking into consideration the fact that most of the current structures are made of conductive and magnetic materials (and changes in the structure during its lifetime cause changes in electrical and magnetic properties [ 1 , 4 ]), the utilization of electromagnetic testing methods becomes a natural solution. It is well known that by observation of vector magnetic properties it is possible to assess various aspects of conditions of an examined structure [ 4 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 ]. Therefore, this paper focuses on presenting the results of work on a local electromagnetic system for nondestructive (NDT) inspection of steel elements.…”

“…In addition to the rising requirements, they must present a wide range of sensitivity. This forces the design and construction of complex multidirectional and multi-sensor transducers [ 9 , 10 , 11 , 12 , 13 , 14 , 15 ]. The growth in the number of information sources, on the one hand, allows for an increase in the reliability and efficiency of measurement systems, on the other hand, it requires the use of sophisticated data processing and aggregation algorithms.…”

“…The axial-symmetric transducer presents similar sensitivity for any flaws (or in general any nonuniformity) affecting the changes of the examined materials’ magnetic properties, regardless their orientation in reference to the x-y axis. Previously, the details of the construction of the utilized transducer and the preliminary results of defect detection and evaluation were presented [ 12 , 13 ]. The magnetic field distribution sensed by the sensor matrix depends on a location of the defected area in the material with regard to the transducer’s circumference and differs with position [ 12 ].…”

“…Further, in order to obtain natural patterns of field distributions dependent only on given defect arrangement, a database was constructed using finite element methods (FEM) computation results. Following the rapid development of technology and computational power, utilization of numerical analysis to increase the efficiency of measuring systems and the reliability of the inspection methods is significantly growing [ 10 , 13 , 35 ]. The numerical modeling in industrial applications is a powerful tool.…”