Open Collections

Estimating the final moisture content variation in kiln-dried Pacific coast hemlock

University of British Columbia

Kiln drying is indisputably a significant value-adding step in timber processing where the importance of predicting moisture within a dried batch cannot be overemphasized. Among timber drying quality indices, the uniformity of final moisture content within a drying timber batch is crucial. Lack of such uniformity leads to undesirable moisture ranges, thus producing large percentages of over-dried and/or under-dried timber sub-populations which in turn, result is significant degradation and value downgrade. Because it is a cumbersome task to collectively correlate various pre-dry timber factors to moisture variation of the post-dry timber population, predicting the value and variability of final moisture is still a great challenge. Part I of this study is dedicated to predicting and characterizing that moisture variation based on the initial and target moisture values using polynomial models. Four polynomial models (PM) are used to correlate initial and final moisture characteristics. While one model failed due to discontinuity three models successfully characterized final moisture variation with the best one showing an R2 > 96% for the goodness of fit. The robustness of the three best models is analyzed and a closed formula is proposed to evaluate the final moisture coefficient of variation based on the target moisture (setpoint) and initial moisture coefficient of variation. In part II, five artificial intelligence (AI) approaches including multilayer perceptron (MLP) and radial basis function (RBF), group method of data handling (GMDH), adaptive neuro-fuzzy inference system (ANFIS), and support vector regression (SVR) are employed to create a comprehensive predictive model that connects selected initial wood attributes (basic density, initial weight, initial moisture, and target moisture) to the final moisture. Seven configurations are constructed, i.e., A (4 attributes); B, C, and D (3 attributes); E, F, and G (2 attributes). As a result, GMDH showed the best performance in predicting final moisture followed by SVR and MLP, especially when it came to configuration A. Neural networks and polynomial models have their own advantages and limitations. Neural networks are capable of predicting final moisture for every single dried timber using numerous factors while polynomial models provide a closed formula for estimating the final moisture variation.

Text

2021-03-31

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/77686

Forestry

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/