Analog Circuit Fault Diagnosis Based on Genetic Algorithm Optimized DELM

Abstract:
In view of the small feature difference between fault classes and difficult diagnosis in analog circuit fault diagnosis, and the artificial selection of hidden layer nodes will miss the effective feature information of the fault itself, resulting in large training errors. This paper proposes an analog circuit fault diagnosis model based on genetic algorithm (GA) optimized DELM. Because the number of hidden layer nodes in deep network is difficult to select, selecting appropriate hidden nodes can effectively reduce the time complexity, The network has good generalization ability. However, the genetic algorithm can better avoid the model falling into local optimization, has high convergence ability, and effectively improves the classification accuracy of large samples. Therefore, the genetic algorithm is used to optimize the hidden layer nodes of DELM, and the testable nodes of the circuit are studied. The best test nodes are selected by comparing the included angles between different fault categories. Finally, it is verified by the nonlinear rectifier circuit. The diagnosis results of hard fault and soft fault show the feasibility of the algorithm. Compared with other algorithms, it is proved that the algorithm can adaptively search the global optimal solution.