Analog circuit fault diagnosis based UCISVM

Focusing on the issue of analog circuit performance online evaluation, the arithmetic speed and the evaluation reliability should be considered. Moreover, the data collected from industrial field has a lots of undesirable features, such as nonlinear feature, time varying feature and contained faults value. All of them should be taken into account. Therefore, two online evaluation strategies are proposed for an analog circuit performance evaluation. First, an analog circuit performance evaluation strategy based on improved support vector machine (ISVM) is presented for the purpose of deducing the training data number largely. This method can deduce the data training set largely as little as 10% of the initial training set and tackle the computational complexity. However, the ISVM is established on the basis of random selection of training set, and this blindness of data training set random selection would bring great impact on the performance of evaluation accuracy. Based on this, another analog circuit fault diagnosis strategy based on unsupervised clustering ISVM (UCISVM) is proposed. This method not only maintains the merit of small data set, but also overcomes the defect of training set selection randomly. The strong characteristic of the support vectors are the only concerns during the diagnosis processes. Corresponding, the unknown fault diagnosis also can be recognized via the UCISVM. The experiment takes a typical analog circuit as diagnosis object. In order to prove the effectiveness of the proposed two methods in this paper, the traditional fault diagnosis method based on standard support vector machine (SVM) is employed also. The diagnosis speed and accuracy are all proved via numerical simulation.