Magnetic particle inspection is typically used to detect the magnetic leakage caused by defects. This method is mainly used to detect the surface and subsurface defects of ferromagnetic materials. The conventional detection method involves inspectors performing visual inspection under high-power ultraviolet light. However, the intense ultraviolet light can easily damage the eyes of the inspectors. Furthermore, the aforementioned process is not only time consuming but also susceptible to human errors. Therefore, this study developed an automated optical inspection system to perform magnetic particle inspection. Analysis of several image features revealed that a contour compactness between four and five can be used to distinguish defective and non-defective features effectively. The defect identification ability obtained with several input combinations of image features for neural networks was analyzed. The results revealed that a high identification ability can be achieved for defective features when the input combination of area, mean width, and compactness is used.
Development of modeling and simulation for magnetic particle inspection using finite elements