针对电站管道不规则接头的超声相控阵自动检测时数据处理效率低、信号源位置识别困难等问题，建立了超声相控阵自动检测信号识别算法，设计制作了带有人工缺陷的模拟试件，并利用该试件对识别算法进行验证。结果表明：提出的超声相控阵自动检测信号识别算法能够提高信号的初判效率，直观显示信号沿扫查方向的分布情况，有效区分结构干扰信号及同一缺陷的多个显示，为管道不规则接头超声相控阵自动检测提供了技术支持。

Aiming at the automatic phased array ultrasonic inspection of the irregular structure of welds, an algorithm of signal recognition is proposed to solve the problem of inefficiency of data processing due to large amount and identification of signal source because of complexity of beam propagation with the appearance of tilt parts. The test block with artificial defects is designed and manufactured for validation of the algorithm. The results show that the proposed algorithm provides important support for the inspection as follows: the efficiency of data processing can be improved; the distribution of signals along the scanning direction can be displayed intuitively; the source of signals can also be acquired to distinguish pseudo signal from structure and multiple displays from one and the same defect.