Abstract: Objective Welds are critical components in crane structures, they directly influence manufacturing quality and in-service performance, therefore, detection of weld defects is of paramount importance. Alternating current field measurement (ACFM) applies a uniform alternating current field to the component under test and measures magnetic field perturbations, enabling effective identification of surface and near-surface defects in metallic materials. Consequently, it is widely used in weld inspection. However, surface microcracks and surface irregularities in crane welds often lead to missed detections and false positives in ACFM, this paper aims to address this issue and improve detection accuracy of weld defects. Methods Firstly, finite element simulations are conducted to analyze the interference effects of surface irregularities on electromagnetic field responses. Subsequently, a signal processing method based on wavelet threshold is proposed, which suppresses interference through a four-step process of detrending, filtering, wavelet basis selection, and threshold processing. Finally, experimental validation is performed with an array instrument. Results Simulation results reveal that the lift-off effect caused by weld irregularities significantly impacts B_x signal, whereas B_z signal exhibits superior stability. Experimental results demonstrate that the proposed method achieves a noise reduction ratio of 7.43 dB and a signal-to-noise ratio of 14.07 dB, both of which outperform existing methods. Conclusion This approach effectively eliminates noise interference and clearly extracts defect features, providing reliable technical support for the real-time monitoring of weld quality.