摘要
针对模拟电路故障信号的容差性、非线性、非平稳性等检测困难问题,提出适合处理这类信号的希尔伯特黄变换算法(Hilbert Huang Transform,HHT),但信号特征提取产生虚假分量和模态混叠等不足。基于此,文中分别提出相关系数法和集合经验模式分解法进行改进。核函数和内核参数决定不同性能的支持向量数据描述(Support Vector Data Describe,SVDD),寻找最优内核参数,选择合适的核函数并构造多核函数优化SVDD算法。文中首先用改进HHT提取联合故障特征向量,然后训练优化后的SVDD分类器,最后将数据输入SVDD中进行检测,能有效地诊断电路故障,并具有较高准确率。
In order to solve the problem of nonlinearity ,non-stationary and poor component tolerances in analog circuit fault detection, Hilbert Huang Transform （HHT） is proposed which is suitable to process this type of signal. Aiming at the problems of illusive compo- nent and mode mixing caused by HHT in the process of signal feature extraction, correlation coefficient method and ensemble empirical mode decomposition algorithm are presented to improve this appearance. The different kernel functions and kernel parameters decide dif- ferent property of Support Vector Daha Description （SVDD）. Improve the SVDD algorithm by optimizing kernel parameters and choosing the suitable kernel functions to construct a new function. The improved HI-IT is used to extract the joint fault characteristic vector. Then train optimized SVDD. Finally test the fault characteristic vector to effectively diagnose the circuit faults with higher precision.
出处
《计算机技术与发展》
2015年第7期179-183,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(GZ212015)
