Eddy Current Microsensor and RBF Neural Networks for Detection and Characterization of Small Surface Defects

Abstract

The growing complexity of industrial processes and manufactured parts, the growing need for safety in service and the desire to optimize the life of parts, require the implementation of increasingly complex quality assessments. Among the various anomalies to consider, sub-millimeter surface defects must be the subject of particular care. These defects are extremely dangerous as they are often the starting point for larger defects such as fatigue cracks, which can lead to the destruction of the parts.
Penetrant testing is now widely used for this type of defect, due to its good performance. Nevertheless, it should be abandoned eventually due to environmental standards. Among the possible alternatives, the use of eddy currents (EC) for conductive materials is a reliable, fast, and inexpensive alternative.
The study concerns the design and modeling of eddy current probe structures comprising micro-sensors for non-destructive testing. The moving band finite element method is implemented for this purpose to take into account the movement of the sensor, experimental validations were conducted on a nickel-based alloy specimen. The real and imaginary parts of the impedance at every position of the sensor computed by experiments and simulations were in good agreement. The crack detection quality was quantified and the geometric characteristics of the defects were estimated using RBF NN (Radial Basis Function Neural Networks) that were designed and implemented on the acquired signals.