An Interpretable Deep Learning Model for Wood Chip Moisture Content Prediction

Abstract: Wood chips are one of the most prominent biofuel energy sources and an essential raw material for the pulp and paper industries. The quality of the biofuel or pulp is highly dependent on the moisture content of the wood chips. When manufacturers have accurate moisture content information at their disposal, they can adjust their manufacturing processes to ensure the highest quality output with the least amount of waste. Therefore, industrial applications of wood chips necessitate rapid and accurate measurements of moisture content. Existing methods of determining moisture content are either time-consuming or do not offer global moisture content values. In this study, we proposed an image-based deep-learning model for wood chip moisture content measurement. Due to the inherent feature extraction capability of the proposed method, it can operate as an end-to-end process to predict moisture levels from images. The proposed model achieved up to 82.22 percent accuracy and an AUC score of 0.98. We also performed the model interpretability analysis to explain how the model makes predictions. The eventual purpose of this preliminary study is to develop an online tool that can predict the overall quality of wood chips, including the amount of ash and the size of the particles.

 
Have trouble accessing an IISE presentation? Login to your account at IISE Conference Archives and click "Search Archives"

If you are interested in the Sustainable Development division of the Institute of Industrial and Systems Engineers (IISE), become a member today!

 

Leave a Reply

Your email address will not be published. Required fields are marked *