Application of Improved Fuzzy RBF Neural Network Algorithm in the Prediction of Wood Dyeing Recipes

Based on the fuzzy RBF neural network, this paper proposes an algorithm of triangular membership function on the basis of minimum and maxim values of the parameter, according to wood anisotropy and difficult collection of its physical and anatomy characteristic. The new algorithm does not need to use various parameters of our study objects, and only needs to consult related literature for the scope of the material parameters, which can restrict the model and distinguish different tree species. The error of the algorithm used for the prediction of wood dyeing recipes is 0.62%. As it is only needs several adjustments to get an accurate color model, the normalization capability has been greatly improved. The operation time is 347s, which can be applied to the industry on-line prediction.