基于提升小波和支持向量机的模拟电路故障诊断

发布时间：2018-03-29 15:41

本文选题：故障诊断　切入点：模拟电路　出处：《南京邮电大学》2015年硕士论文

【摘要】：模拟电路故障诊断对电路设计、设备生产和仪器维护是至关重要的,故障诊断技术是目前专家学者们和测试工程师在电路与系统领域中具有挑战的重要课题。但是由于模拟电路元件参数的容差性、电路响应的连续性和非连续性以及环境因素造成电路故障的多样性和复杂性,使得传统的诊断方法无法更好地运用在实际模拟电路故障诊断中。本文基于现代诊断技术,构建以提升小波和支持向量机相结合的模拟电路故障诊断框架,深入研究了故障特征提取和分类器构建这两个模拟电路故障诊断中的重要环节。本文取得成果如下:(1)在总结已有的模拟电路故障特征提取方法基础之上,本文提出了将提升db5小波运用到电路故障特征提取当中。这是因为db5小波函数与故障输出信号具有较高的相似度,有利于提取信号的重要特征,而且分解速度快。与非提升方法相比能够更加准确地反映原始信号的特征。通过仿真实例得到的数据验证了该方法的优越性。(2)针对目前模拟电路故障识别中遇到的难题,本文构建了基于马氏距离的最小二乘支持向量机分类器。该方法通过运用最小二乘把复杂的求解问题简单化了,通过引入马氏距离,改善了最小二乘支持向量机的稀疏性,从而节省了分类器的训练时间。仿真实例表明,此方法可以有效地运用在模拟电路故障诊断之中。(3)为了提高支持向量机的泛化学习能力,本文通过粒子群算法优化最小二乘支持向量机的结构参数,但是由于粒子群容易陷入局部最优和早熟收敛问题,所以对标准PSO算法进行改进,提出了吸引-排斥控制粒子群优化方法,通过控制粒子的吸引运动和排斥运动提高粒子的多样性,避免了粒子在进化过程中陷入了局部最优和早熟收敛问题。
[Abstract]:Analog circuit fault diagnosis is very important for circuit design, equipment production and instrument maintenance. Fault diagnosis technology is an important subject for experts and test engineers in the field of circuits and systems. However, because of the tolerance of analog circuit components, The variety and complexity of circuit fault caused by the continuity and discontinuity of circuit response and environmental factors make it impossible for the traditional diagnosis method to be better used in the actual analog circuit fault diagnosis. An analog circuit fault diagnosis framework based on lifting wavelet and support vector machine is constructed. In this paper, the fault feature extraction and classifier construction are studied in detail. The main achievements of this paper are as follows: 1) on the basis of summarizing the existing methods of fault feature extraction in analog circuits, In this paper, the lifting db5 wavelet is applied to circuit fault feature extraction, which is due to the high similarity between the db5 wavelet function and the fault output signal, which is helpful to extract the important features of the signal. Moreover, the decomposition speed is fast. Compared with the non-lifting method, it can reflect the characteristics of the original signal more accurately. The data obtained from the simulation example verify the superiority of the method. In this paper, the least squares support vector machine classifier based on Markov distance is constructed. The method simplifies the complex problem by using least square, and improves the sparsity of least squares support vector machine by introducing Markov distance. Thus, the training time of classifier is saved. The simulation example shows that this method can be effectively used in analog circuit fault diagnosis. In this paper, the structure parameters of least squares support vector machine are optimized by particle swarm optimization, but the standard PSO algorithm is improved because the particle swarm is easy to fall into local optimal and premature convergence problems. An attractive repulsive control particle swarm optimization method is proposed to improve the diversity of particles by controlling the attractive and repellent motion of particles and to avoid the problem of local optimum and premature convergence of particles in the evolution process.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN710

【相似文献】