A Neural Network Model for Wood Chip Thickness Distributions

Emily B. Schultz, Thomas G. Matney, Jerry L. Koger

Abstract


Wood chip thickness is an important factor in pulp quality and yield. An artificial neural network model was developed and incorporated into a growth and yield simulator to predict wood chip thickness distributions from stand and tree characteristics. Models based on direct parameter estimation and parameter recovery were also developed for comparison to the neural network. Data were derived from 11,771 individual loblolly pine chip thickness measurements. Four stand ages, five dbh (diameter at breast height) classes, and three stem positions were used to predict the cumulative proportion of chip weight per chip thickness class. Results showed that the neural network model was superior to the two deterministic models on the basis of bias, root mean square error, and index of fit. Sensitivity analyses for the neural network model demonstrated that thicker chips were produced by younger stands and lower stem positions. The neural network was combined with a growth and yield simulator to demonstrate its use as a tool for procurement foresters and mill managers in predicting yields from stands of given characteristics.

Keywords


Neural network;wood chips;wood chip thickness;wood chip thickness distributions

Full Text:

PDF

Refbacks

  • There are currently no refbacks.