利用超声衍射时差法（Time of Flight Diffraction，TOFD）对合金钢焊缝中的小尺寸裂纹实施检测时，受脉冲宽度等因素制约，裂纹上、下端点的衍射波会发生混叠，而导致裂纹高度定量困难。采用自回归谱外推（Autoregressive Spectral Extrapolation，ARSE）技术对混叠信号进行处理，扩展有效频带范围，达到压缩时域信号脉宽和提高时间分辨率的目的。结果表明，结合ARSE技术可有效分离混叠信号，实现厚度100.0 mm合金钢中深度50.0 mm处高度1.0 mm裂纹的TOFD定量检测，且相对定量误差不超过5.3%。

When time of flight diffraction (TOFD) is applied to detect small size crack in alloy steel weld, the diffraction waves at upper and lower points of crack will overlap and cause difficulty in quantifying the crack height due to restriction in the pulse width. In this paper, autoregressive spectrum extrapolation (ARSE) technique was applied to deal with the overlapped signals and extend effective band range to compress time domain pulse width and improve time resolution. The results showed that ARSE technique could effectively separate overlapped signals. Focusing on 100 mm thickness alloy steel of having a crack of 1.0 mm in height and 50 mm in depth, the quantitative error of crack height was no more than 5.3%.