Improved Genetic Algorithm Neural Network Method and the Application in Valve Fault Diagnosis of Diesel Engine

摘要

The shortcomings of BP neural network is slow convergence rate, falling into local minimum easily and difticult to determine the number of hidden nodes accurately. In order to improve the diagnosis accuracy, in this paper, the number of hidden nodes, weights and threshold of BP neural network were optimized by using binary and real number hybrid coding based on genetic algorithms with global searching ability. Finally, the method tested with WD615 diesel engine valve fault diagnosis data. Experimental results showed that this algorithm has obvious advantages, it is able to overcome the deficiencies of BP neural network, and improves the network's learning and diagnosis ability.