Fault diagnosis method based on improved genetic algorithm and neural network

Abstract

In order to overcome the shortcomings such as slow convergence rate and prone to sink into small locality in BP neural network, adaptive genetic algorithm and BP algorithm are combined to take shape a hybrid algorithm to train artificial neural network. In a specific implementation, firstly, an adaptive genetic algorithm is used to perform multi-point genetic optimization on the initial weight space of the neural network, and better search space is located in the solution space. On this basis, local exact search is performed using BP algorithm, ultimately the global optimum is achieved. This algorithm is simulated based on the fault diagnosis of one certain helicopter's airborne electrical control box and one certain flight control box of aircraft autopilot. The simulation conclusions indicate that the algorithm has faster convergence rate and higher diagnostic accuracy.