Abstract: In the petroleum industry, carbon steel structure is affected by corrosion and stress for a long time, complex cracks are easily introduced in the surface of the carbon steel structures. The detection and analysis of complex cracks require strong data processing ability, which poses many challenges to the functional architecture of detection instruments. Based on the principle of alternating current field measurement (ACFM), the finite element model of different types of cracks was established in the paper. The induced current distribution on the surface of the structure was analyzed. The mapping relationship between defect morphology-current perturbation-magnetic field distortion was explored. An intelligent defect identification algorithm based on the characteristic signal B_z was proposed. Based on FPGA (Field Programmable Gate Array) platform, a ACFM instrument for defects intelligent detection was built. The instrument realized the excitation signal generation and the detection signal collection, processing and display. Artificial crack identification experiments were carried out. The experimental results showed that the ACFM instrument for defects intelligent detection can realize the reconstruction and recognition of linear cracks and complex cracks at different angles.