The Way Of Artificial Neural Networks Is Applied To Larch Wood Drying

This paper studies the larch wood, a dry wood was used as raw material form Northeast Forestry University laboratory. The data of Larch wood dried as a training and test samples, neural network model of larch wood drying were studied by MATLAB software operation and BP artificial neural network in the application of modeling techniques.In a study of larch wood dried model, the dry bulb temperature, humidity, medium circulation wind speed and equilibrium moisture content were selected as the input matrix, the moisture content is selected as the output matrix, building structure4:S:1multi-input multi-output1BP neural network model, the function of log-sigmoid function (logsig) in Matlab as the role of neuronal function. With80sets of data as training samples,the model does repeated training comparison test and gets appropriate model structure is4:10:1. Model training results showed that the mean square error mse=0.0017. Analysing fitting accuracy of the model, the result shows that, overall fitting accuracy of larch wood drying model is96.12%, the fitted theoretical value and the actual approximation value are close. Using40sets of data as the test samples test the model, and the overall fitting accuracy is94.32%, and graph the contrasting chart of actual and theoretical values of the moisture content of dried larch wood the in three-dimensional space. Now it shows that in this paper the predicted theoretical value of neural network model of larch wood drying is close to the measured values better, further explain the predictive ability of the model is very strong.On the basis of above, applying multiple regression analysis model do further research on drying wood, using statistical software Statistical Package for the Social Sciences(abbreviation SPSS) creates multiple multivariate regression equations of larch wood drying, regression coefficient was0.89.