非线性电路多软故障的智能优化递阶特征选择诊断方法

A hierarchical features selection and diagnosis method of intelligent optimization for multiple soft fault of nonlinear circuit

针对非线性模拟电路的多软故障诊断中由于部分状态特征近似而影响诊断准确率的问题,在不提高电路信息采集工作量的情况下,采用分层诊断的思路,提出了基于Wiener核的智能优化递阶特征选择方法。该方法在获得电路各种状态的Wiener核后,采用智能优化算法对各状态的核特征进行特征选择,以代表各状态的特征构成的矢量的集总欧氏距离为评价函数,对集总欧氏距离的最大值进行寻优得到最优解;再对各个特征矢量间的距离进行判别,找到相互距离小于设定阈值的各个状态,构成次阶故障状态类,对该类故障状态采用前述的方法进行智能优化故障特征选择,得到次阶各状态的最优特征矢量;以此类推,直到得到满意的分辨率为止。实验表明,该方法有效地提高了多软故障诊断的准确率。

Aiming at the problem of accuracy rate of a diagnosis,caused by the similar characteristics of states in the multiple soft fault diagnoses of the nonlinear analog circuits,in case of not increasing the workload of circuit information collection,using the idea of hierarchical diagnosis,a hierarchical features selection method of intelligent optimization based on Wiener kernel is proposed. Firstly,after obtaining the Wiener kernel of various states of the circuit,this method selects the features of Wiener kernel of each state by intelligent optimization algorithm,considering the lumped Euclidean distance of vectorial constituted by representative features of each state to be the evaluation function,and acquiring the optimal solution by optimizing maximum values of the lumped Euclidean distance. Then,this method distinguishes the distance of every pair of characteristic vectors to find those states where mutual distance is less than the set thresholds,constituting the fault condition class of the next order,and applying the intelligent optimization fault feature selection according to the fore-said method to this kind of fault,optimal feature vectors of each state of the next order will be obtained. By that analogy,a satisfactory resolution will be obtained. Experiments show that this method can effectively enhance the accuracy rate of the nonlinear multiple soft fault diagnosis.