Quantitative detection of combined cracks based on artificial neural network and eddy current testing signals

Quantitative nondestructive testing with enough precision are the basis for studying crack propagation behaviour and the residual life of structural component. Eddy current testing (ECT) is a fast nondestructive testing technique with many testing objects. As the common effect of each crack in combined cracks on ECT signals, quantitative detection of combined cracks is a challenge. In this paper, quantitative detection of combined cracks using features of ECT signals and an artificial neural network (ANN) method is proposed. Firstly, a model of combined cracks containing a long crack and a short vertical crack is used to approximately calculate two-dimensional ECT signals of crack. Secondly, correlation between the parameters of combined cracks and the features of the two-dimensional ECT signals are investigated by numerical simulation. Finally, the crack parameters are evaluated from the simulation signals of combined cracks and the measured signals of stress corrosion cracking using the proposed strategy. Numerical results verify the effectiveness of the proposed strategy.