Aiming at theSproblem of accuracySrate ofSa 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 is selecting the features of Wiener kernel of each state by intelligent optimization algorithm, considering the lumped Euclidean distance of vectorial constituted bySrepresentative features of each state to be the evaluation function, and acquiring the optimal solution by optimizing maximum values of the lumped Euclidean distance. Then distinguishing the distance of every pair of characteristic vectors, finding 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 foresaid method to this kind of fault, optimal feature vectors of each state of the next order will be got. By that analogy, a satisfactory resolution will be obtained. Experiments show that this method can effectively enhance the accuracy ofSthe nonlinear multiple soft fault diagnosis.
