在超声无损检测中，超声检测信号的噪声消除对于缺陷的检测和识别具有重要的意义。由于超声信号具有非平稳性的特点，所以降噪方法的选择就显得尤为重要，直接影响着缺陷的定性、定量和定位分析。独立分量分析(Independent Component Analysis，ICA)是目前信号处理中新兴的技术之一，其目的是要寻求一个线性变换矩阵，使得变换后的分量尽可能相互独立，从而达到消除噪声信号的目的。 本文从超声仿真信号和实测超声检测信号两方面出发，结合相空间重构技术、粒子群算法，开展了超声检测信号的消噪的尝试性工作。主要的研究内容包括： （1） 介绍和分析了独立分量分析方法的原理和各种典型的实现算法，以及各自的特点； （2） 利用经典的FastICA算法、JADE算法和Infomax算法进行了实验仿真和超声信号的分离工作，验证了算法的有效性； （3） 利用超声信号和噪声信号的相空间特性，采用相空间重构技术建立相空间矩阵，对相空间矩阵进行独立分量分析，并通过实验选择出最优算法进行后续的信号处理研究； （4） 针对传统的优化算法收敛速度慢、易于陷入局部最优，影响了信号的分离性能这方面的缺点，将一种改进的粒子群算法与独立分量分析相结合，通过超声仿真信号和实测信号来验证算法的可行性和优越性。 实验结果表明，独立分量分析方法在超声检测的微弱信号提取中具有可行性和有效性。同时，文中对独立分量分析在超声检测信号处理中的应用进行了总结，并指出了进一步的研究方向。

In ultrasonic nondestructive testing, ultrasonic testing signal noise cancellation is of great significance for the detection and identification of defects. Ultrasonic signal has non-stationary characteristics, so the choice of noise reduction method is particularly important, which have a direct impact on the defects in quantitative and positioning analysis. Independent Component Analysis (ICA) is one of the emerging technologies of the current signal processing, and its purpose is to find a linear transformation matrix, the transformed components as soon as possible independent of each other, so as to achieve the goal of eliminating noise signals. This thesis puts forward its own solutions from the two aspects of the simulation of ultrasonic signal and the measured ultrasonic detection signal through the experiments to verify the feasibility and effectiveness of the method. The main research content and results are as follows: (1) This thesis has been analyzed and studied the principle of Independent Component Analysis and other algorithms’ implementation method, and described each algorithm’s nature. (2) The simulation model of the FastICA algorithm, JADE algorithm and Informax algorithm has been established, and their separation performance has been completed, and their validities have been verified. (3) By the characteristics of the ultrasonic signal and noise signal of the phase space, the phase space matrix has been established and independent component analysis has been exacted. The most optimal algorithm has been obtained by experiment. (4) Considering the defects of the traditional optimization algorithm, such as, the slow convergence, easy to fall into local optimum, affecting the signal separation performance, an improved particle swarm optimization algorithm and independent component analysis are combined through ultrasound simulation signal and measured signal to verify the feasibility and the superiority. The simulation results show that the ICA algorithm is a effectiveness and feasibility method. At the same time, the thesis has summarized independent component analysis of signal processing in ultrasonic testing on the basis of the study, and pointed out the feasibility of the weak signal extraction method in ultrasonic testing as well as the problems, and summarized the further research directions.