New Signal Processing Methods for Surface Shape Restoration in Eddy Current
兼本 茂 Shigeru KANEMOTO 程 衛英 Weiying CHENG 志波 光晴 Mitsuharu SHIWA 古村 一朗 Ichiro KOMURA
Deconvolution FilterEddy Current TestLine Spread FunctionLinear Discriminant AnalysisPoint Spread FunctionSCC
This study proposes new signal processing methods to restore surface shapes of flaws from eddy current testing（ECT）signals. A linear discriminant analysis method is applied to ECT signals embedded in multi-dimensional state space to identify flaw signals form background noises in prior. A new concept of equivalent point spread function is introduced, and the real flaw shape is restored by utilizing a deconvolution filter based on Richardson-Lucy algorithm. An EDM notch covered by large welding noises is identified and the length of this notch is obtained from flaw shape restoration. Furthermore, this method is applied to the surface shape restoration of stress corrosion cracks（SCC）and the surface shape of a compli- cated SCC is restored. By the way, the optimum scanning pitch in flaw inspection can be determined from the quantitative evaluation of spatial response character- istics of the equivalent point spread functions of probes in usage.