摘要
针对信号稀疏分解中常用匹配追踪分解不够准确的问题,提出基于K-SVD奇异值分解的超声渡越时间获取方法。利用K-SVD训练得到超声回波信号的过完备字典,结合正交匹配追踪进行局部搜索适配原子,以提高信号稀疏分解的速度和准确度。基于Comsol Multipysics仿真软件建立充液污垢管道三维有限元模型,研究了超声回波传播特性规律。将K-SVD算法应用于超声回波仿真信号和换热污垢管道回波检测信号的处理,并与原始小波训练字典进行对比。结果表明,改进的K-SVD字典学习算法能够在提高信号稀疏分解的同时,获得较好的降噪结果和污垢特征信息提取,对超声检测信号的处理具有实际意义。
For the non-accurate decomposition of common matching pursuit in the signal sparse decomposition, this paper proposed to adopt an acquisition method of ultrasonic transit time based on the decomposition of K-SVD singular values, through K-SVD training and locally search adaptive atoms by combining the orthogonal matching pursuit, obtained the redundant dictionary of ultrasonic echo signals to increase the speed and accuracy of signal sparse decomposition. With the 3D finite element model of filling-liquid fouling pipelines based on the Comsol Multipysics simulation software, this paper researched characteris- tic rules of ultrasonic echo spread and achieved the echo extraction of fouling pipelines. Besides, it applied the K-SVD algo- rithm into the processing of ultrasonic-echo simulation signals and echo detection signals of heat-exchanger fouling pipelines, compared it with the original training of wavelet training dictionary. And results indicate that: with the K-SVD dictionary learning algorithm, it can achieve the better noise reduction result and extraction of fouling feature information while improving the signal sparse decomposition, which is of actual significance for the processing of ultrasonic detection signals.
出处
《计算机应用研究》
CSCD
北大核心
2017年第6期1740-1744,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(51176028)
吉林省科技发展计划资助项目(20140204030SF)