Utilization of Multisensor Data Fusion for Magnetic Nondestructive Evaluation of Defects in Steel Elements under Various Operation Strategies

Increasing the number of inspection sources creates an opportunity to combine information in order to properly set the operation of the entire system, not only in terms of such factors as reliability, confidence, or accuracy, but inspection time as well. In this paper, a magnetic sensor-array-based nondestructive system was applied to inspect defects inside circular-shaped steel elements. The experiments were carried out for various sensor network strategies, followed by the fusion of multisensor data for each case. In order to combine the measurements, first data registration and then four algorithms based on spatial and transformed representations of sensor signals were applied. In the case of spatial representation, the data were combined using an algorithm operating directly on input signals, allowing pooling of information. To build the transformed representation, a multiresolution analysis based on the Laplacian pyramid was used. Finally, the quality of the obtained results was assessed. The details of algorithms are given and the results are presented and discussed. It is shown that the application of data fusion rules for magnetic multisensor inspection systems can result in the growth of reliability of proper identification and classification of defects in steel elements depending on the utilized configuration of the sensor network.