论文标题:集成化声发射信号处理平台的研究
Studies on Integrated Platform for Acoustic Emission Signal Processing
论文作者 张平
论文导师 施克仁；刘时风,论文学位 博士,论文专业 材料加工工程
论文单位 清华大学,点击次数 2,论文页数 132页File Size1016k
2002-04-01论文免费下载 http://paper.dic123.com/lunwen_106391052/ 无损检测，声发射，集成化信号处理平台，盲目反卷积
Non-destructive evaluation, Acoustic emission, Integration platform for signal processing, Blind deconvolution
对声发射信号进行分析与处理是目前获取声发射源信息的唯一有效途径，因此集成化的信号处理平台是声发射检测技术工程应用的重要组成部分。针对目前国外已推出的信号处理平台存在的不足以及国内在这方面研究空白的现状，本文提出及开发了一种以小波分析及人工神经网络为主要信号处理和模式识别方法的集成化声发射信号处理平台。论文对小波分析方法在声发射信号处理中的应用进行了全面深入的研究。首先根据声发射信号的特点，提出一套声发射信号小波分析的小波基选取规则方法，并指出Daubechies小波、Symlets小波和Coiflets小波适合于声发射信号处理；其次深入剖析基于Mallat算法的小波分解的分频概念，推导出分解频带的带宽公式，并由此推导出小波分解的最大分解尺度公式，对声发射信号的小波分析具有重要的指导作用；提出了三种基于小波分析的声发射信号特征分析方法：小波特征频谱分析法、小波特征能谱系数法、小波分解系数分析法。实际的工程应用结果表明：提出的三种方法能够有效地提取声发射信号的特征。论文对BP神经网络在声发射信号模式识别应用中的共性问题进行了研究。提出了改进的BP算法、加噪声循环训练法、小波分析与BP神经网络有机结合等三种提高神经网络性能的方法。实验结果表明，基于以上方法的BP网络在声发射模式识别应用中取得了较好的效果。基于以上研究成果，本文开发了国内首个基于波形分析的集成化声发射信号处理平台。该平台已成为国内开发的首台多通道全波形声发射检测仪的核心模块，目前已经在多个声发射技术工程领域得到应用，并取得了良好的效果。本文首次采用盲目反卷积法研究声发射源信号的问题，实现了同时对声发射源信号和传播路径的冲击响应函数进行估计。实验结果表明，该方法能较好地恢复模拟声发射源信号，特别是能够在一定程度上恢复饱和限幅失真的声发射信号的源信息；为研究由声发射信号恢复声发射源信号探索了一个新方向。论文的研究成果对于推动我国声发射检测技术的发展，提高声发射源特征信息的获取量具有重要意义和实用价值。集成化处理平台的研究为我国第一台全波形数字化声发射检测仪的研制成功奠定了坚实的基础。
Acoustic emission signal processing is the only effective method to acquire the acoustic emission source information by now and the integrated platform for signal processing became more important for the engineering application of acoustic emission technique. Aiming at the shortcoming of the products for signal processing platform developed by foreign companies and the domestic status of related research, an integrated platform for acoustic emission signal processing based on wavelet analysis and neural network has been developed in the thesis. The application of wavelet analysis on acoustic emission signal processing was studied in the thesis. At first, the rules of how to select the suitable wavelets for acoustic emission signal processing were proposed based on the features of acoustic emission signal. In terms of the rules, Daubechies wavelet, Symlets wavelet and Coiflets wavelet were regarded to be suitable for acoustic emission. Secondly, the frequency decomposition band of wavelet analysis was formulated and the maximum decomposition level of wavelet analysis was also formulated. The research results above were important to use wavelet analysis for acoustic emission signal processing. Three feature extraction methods for acoustic emission signal based on wavelet analysis were proposed: wavelet feature frequency analyzing method, wavelet feature energy frequency coefficient method and wavelet decomposition coefficient method. The results of practical engineering application showed that the three methods can efficiently extract the features of acoustic emission signal.The common problems confronted in the application of BP neural network on the pattern recognition of acoustic emission signal were studied. To improve the neural network performance, the thesis purposed such three methods as improved BP algorithm, noise-added repetitious training method, combination of wavelet analysis and neural network. Experimental results showed that BP neural network based on the three methods above can make better results for the pattern recognition of acoustic emission signal.Based on the research results above, the thesis developed the first integrated platform for acoustic emission signal processing in China. As the core module, the platform has been integrated into the first multi-channel waveform digital acoustic emission instrumentation in China. By now, the platform has been applied in many engineering fields related to acoustic emission technique and made better results.For the first time, blind deconvolution was used to research acoustic emission source signal in the thesis. The acoustic emission source signal and the impulse response function of transmission path were estimated at the same time. Experimental results showed that blind deconvolution can recover the simulated acoustic emission source signal and especially can recover the source information of acoustic emission distorted by amplitude saturation to some extent. The research work explored one new orientation for the research of how to recover the acoustic emission source signal from the acoustic emission signal.The research results of the thesis have important significance and practical value to promote the development of acoustic emission technique and to improve the feature information extraction of acoustic emission source. The research work of the integrated platform for acoustic emission signal processing established the stable basis of successful development of the first waveform digital acoustic emission instrumentation in China.
|
