电子系统数据驱动诊断与预测算法的研究与实现

发布时间：2018-04-05 18:59

本文选题：模拟电路　切入点：故障诊断　出处：《电子科技大学》2014年硕士论文

【摘要】：据资料统计,电子设备中80%的故障都是由模拟电路导致的。因此,模拟电路故障诊断技术是电子系统健康管理的重点和难点。另外,随着电动汽车领域的发展,锂离子电池得到了广泛的应用,锂离子电池的寿命成为保障电动汽车性能和安全的关键。因此,锂离子电池寿命预测是电子系统健康管理的另一个重要研究课题。基于上述原因,本论文主要完成以下的工作:1.模拟电路故障诊断方法的研究。根据支持向量机进行模式分类时的本质,提出了采用统计特征量:均值、方差、标准偏差、熵、峭度、偏斜度和形心来组成故障特征向量。同时,针对目前支持向量机进行模式分类时的缺陷,即采用同一个特征向量组合来训练支持向量机所有的二分类器,然而支持向量机的每个二分类器对于不同的特征向量组合有不同的分类精度,提出了基于粒子群算法的特征优选方法。实验结果表明提出的方法提高了模拟电路故障诊断的精度。2.锂离子电池寿命预测方法的研究。根据锂离子电池寿命预测的原理,提出了锂离子电池寿命预测的整体框架。首先,本文提出了采用Verhulst模型作为锂离子电池寿命退化模型。由于传统的Verhulst模型预测精度不高,提出了采用粒子群算法对传统的模型进行优化,提高了预测精度。其次,估计Verhulst模型的参数,本文提出了采用粒子群算法搜索模型的参数。最后,为了降低噪声对预测结果的影响,采用粒子滤波对模型参数进行更新。实验结果表明提出的方法可以以较小的误差预测出锂离子电池的剩余寿命。3.诊断与预测系统的软件设计。为了满足电子系统故障诊断与故障预测的需求,分别设计了电子系统故障诊断与预测系统的软件。诊断系统集成了支持向量机分类算法,对外部读入的数据自动进行故障特征的计算,并迅速给出故障诊断的结果。预测系统集成了多个常用的预测模型,如Verhulst模型、GM(1,1)模型、AR模型,对于外部读入的数据序列,可以自动优选出合适的预测算法,并给出预测结果。经实验数据验证,诊断与预测系统都可以高效地给出精确的结果。
[Abstract]:According to data statistics, 80% of failures in electronic equipment are caused by analog circuits.Therefore, analog circuit fault diagnosis technology is the focus and difficulty of electronic system health management.In addition, with the development of electric vehicle field, lithium ion battery has been widely used. The life of lithium ion battery has become the key to ensure the performance and safety of electric vehicle.Therefore, Li-ion battery life prediction is another important research topic in electronic system health management.Based on the above reasons, this paper mainly completes the following work: 1.Research on Fault diagnosis of Analog Circuits.According to the nature of pattern classification using support vector machines (SVM), a fault feature vector is proposed, which is composed of statistical features: mean, variance, standard deviation, entropy, kurtosis, skew and centroid.At the same time, aiming at the shortcoming of pattern classification of support vector machine, that is, using the same feature vector combination to train all the two classifiers of support vector machine,However, each two-classifier of support vector machine has different classification accuracy for different eigenvector combinations. A particle swarm optimization (PSO) based feature optimization method is proposed.Experimental results show that the proposed method improves the accuracy of analog circuit fault diagnosis.Study on Lithium Ion Battery Life Prediction method.According to the principle of Li-ion battery life prediction, the whole frame of Li-ion battery life prediction is put forward.Firstly, Verhulst model is used as the life degradation model of lithium ion battery.Because the prediction accuracy of traditional Verhulst model is not high, particle swarm optimization (PSO) algorithm is proposed to optimize the traditional model and improve the prediction accuracy.Secondly, the parameters of the Verhulst model are estimated, and the particle swarm optimization algorithm is proposed to search the parameters of the model.Finally, in order to reduce the influence of noise on the prediction results, particle filter is used to update the model parameters.The experimental results show that the proposed method can predict the residual life of lithium ion battery with small error.Software design of diagnosis and prediction system.In order to meet the need of fault diagnosis and prediction of electronic system, the software of fault diagnosis and prediction system of electronic system is designed.Support vector machine (SVM) classification algorithm is integrated in the diagnosis system to calculate the fault features of the external data automatically, and the results of fault diagnosis are given quickly.The prediction system integrates several commonly used prediction models, such as the Verhulst model and the AR model. The prediction system can automatically select the appropriate prediction algorithm for the external reading data series, and give the prediction results.The experimental data show that the diagnosis and prediction system can give accurate results efficiently.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN710;TM912

【共引文献】