Modeling and Simulation of Soft measurement Based on Improved BP Neural Network

Abstract

Keywords: Back Propagation (BP), Principal Component Analysis (PCA), Genetic Algorithms (GA), Soft measurement. Abstract. To solve the problem of Back Propagation (BP) neural network easy to get in local least value and the initial weight is chosen randomly, Principal Component Analysis (PCA) and Genetic Algorithms (GA) were introduced to the BP Neural Network to achieve their complementary advantages. Based on the BP neural Network a GA-BP neural Network improved network based on PCA is built and practically applied. The simulation results show that, the improved network could improve the generalization ability of the model and the ability to predict dynamic measurement data, which make the BP neural Network can be used even more widely.