基于提升框架和微粒群优化神经网络的超声信号消噪技术

Denoising Techniques for Ultrasonic Signals Based on Lifting Scheme and PSO Optimized NN

为消除超声无损检测中晶粒散射引起的相干噪声,提高超声无损检测(UNDT)与无损评价(UNDE)基础数据的信噪比(SNR),提出了一种基于微粒群(PSO)优化神经网络模式识别理论的超声信号消噪技术,利用提升框架对原始超声检测信号进行多分辨率分析,根据微粒群算法强大的全局最优化能力,确定径向基函数(RBF)神经网络的结构,并通过径向基函数神经网络所构成的信噪分离器对信号和噪声进行识别、分离来消除噪声。

To eliminate the coherent noise in ultrasonic nondestructive testing caused by the scattering improve the signal-to-noise ratio （SNR） of the basic data in ultrasonic nondestructive testing of grain, and （UNDT） and nondestructive evaluation （UNDE）, This paper put forward a method based on particle swarm optimization （PSO） neural network pattern recognition theory of ultrasonic signal denoising technology, multi-resolution analyze the original ultrasonic detection signal by lifting scheme, determine the structure of radial basis function （RBF） neural network according to the powerful global optimization capability of particle swarm algorithm, eliminate noise on the signal and noise identification and separation by using signal-to-noise radial basis function neural network separator, obtain the high SNR of the ultrasonic echo signal. The experimental results show that, compared with the traditional split spectrum analysis method, this one improves the stability of the de-noising performance and enhance the signal-to-noise ratio of ultrasonic nondestructive testing.