Use of BP Neural Network to Predict Hydrogen Content in Coal

Abstract

Neural network is a popular method for predicting unknown process variables from measured process data. Many learning algorithms have been proposed in the literature to improve model prediction. In this paper, we introduce the concept of sample study risk in neural network (NN) to improve the prediction of hydrogen content in coal using Back Propagation (BP) NN. Targeting the problem of training convergence quality impaired by the interfering infonnation of some samples in BP NN, the validity of the concept of sample study in NN and the feasibility of analysing the chemical element in coal using NN are discussed.