A size inversion method for inner surface cracks on stainless steel tubes using horizontal ACFM signal

Accurate assessment of stress corrosion cracking (SCC) dimensions is essential for the integrity and safety of stainless steel (SS) tubes operating under harsh conditions. This paper presents a novel size inversion method for the SCC on the inner surface of SS tubes using only the horizontal component of the alternating current field measurement (ACFM) signal (B_x signal), replacing the conventional biaxial signals. A theoretical model is constructed to provide a basis for the proposed method. A finite element model is analysed to investigate the effect of crack size on the B_x signal. Two characteristics, the reduced amplitude of the B_x signal (ΔB_x) and the distance between the peak and trough in the differential of the B_x signal regarding x-axis (l_x), are selected to inverse the crack size. An extreme learning machine (ELM) neural network is employed to establish the mapping relationship between the two characteristics and the crack size. Experiments are conducted to validate the effectiveness of the ELM based on a developed ACFM system, designed with a uniaxial tunnelling magneto-resistive (TMR) sensor. The results indicate that the length and depth of cracks can be accurately predicted by the ELM, which is trained based on the B_x signal.