Computation Method of Processing Time Based on BP Neural Network and Genetic Algorithm

摘要

Looking-up standard processing time table is a commonly used and important determination method of processing time. However, the large error in nonstandard nodes brings adverse effect on its accuracy. In view of the problem, a computation method of processing time based on back propagation neural network (BPNN) and genetic algorithm (GA) is proposed. Several key technologies of BPNN based on Matlab, including computation of the number of neurons in hidden layer, determination of training algorithm, and affecting factors of generalization ability, are researched in depth. In order to improve the training efficiency of BPNN, GA is used to optimize its connection weights and thresholds. The encoding method, selection operation, crossover, and mutation operation of GA are discussed in detail. The higher computation precision and faster operation speed of the proposed method is demonstrated through application cases.