Intelligent system supporting defect identification and sizing using eddy current testing signals in nuclear power plants

Two premier issues in eddy current testing, namely the automatic identification of cracks from their noisy backgrounds and the sizing of cracks with complex morphologies, are dealt with in this paper. A signal processing method based on statistical pattern recognition is established to realise automatic crack identification, and a method for estimating crack depth is proposed, using a new concept, the depth sizing index, which is analytically constructed from raw measurement signals based on system optimisation theory. In addition, a surface crack shape reconstruction algorithm, using a deconvolution filter, is proposed to evaluate the crack length.

Keywords: crack detection, crack sizing, deconvolution filter, eddy current testing, nondestructive evaluation, NDE, intelligent systems, defect identification, nuclear power plants, statistical pattern recognition, crack depth estimation, nuclear energy