【摘要】 针对超声波长时间自动检测复杂构件缺陷时，界面与缺陷回波相互干涉使回波信号呈现随机多峰性，难以实时准确地解析缺陷信息，且事后解读也将带来海量数据存储的技术瓶颈，提出了一种随机信息自由度超声检测信号的压缩及参数重构方法。通过先验能量分布，自适应匹配采样核的频域支撑区间及参数，将随机多峰回波脉冲流通过自适应采样核调制，再以低速率对其进行等间隔稀疏采样，最后通过零化滤波器重构多峰回波信号幅值与时延特征。以钢轨缺陷超声检测为例，相较于奈奎斯特采样法，本文所提方法在保留轨头、轨腰、轨底、人工缺陷及二次回波信息的同时，数据量减少了90%。更多还原

【Abstract】 During prolonged ultrasonic automatic defect detection in complex components, the mutual interference between interface and defect echoes causes echo signals to exhibit random multi-peak characteristics. This not only makes real-time accurate defect information extraction challenging but also creates a technical bottleneck for massive data storage during post-processing interpretation. To address this, we propose a compression and parameter reconstruction method for ultrasonic detection signals with random information degrees of freedom. By leveraging prior energy distribution knowledge, this method adaptively matches the frequency-domain support interval and parameters of the sampling kernel. The random multi-peak echo pulse stream is then modulated by this adaptive sampling kernel, followed by a low-rate uniform sparse sampling. Finally, the amplitude and time-delay features of the multi-peak echo signal are reconstructed using an annihilating filter. When applied to rail defect ultrasonic testing, the proposed approach achieved a 90% reduction in data volume compared to Nyquist sampling, while preserving information regarding the rail head, web, and base as well as artificial defects and secondary echoes.更多还原