Abstract:

A method to predict the wood radial thermal conductivity based on back propagation(BP) neural network model which has non-linear relation highly was proposed. The generalization ability of the network was improved by regularization. The simulation results showed that the neural network model given in this paper is capable of forecasting the behavior of the wood radial thermal conductivity exactly and rapidly. Experiments presented that the model was of accuracy and was higher in respect of precision than the formula derived.

Key words: neural network, generalization ability of network, predict, thermal conductivity, regularization