Wood moisture monitoring and classification in kiln‐dried timber

Different mass timber structures require a specific range of wood moisture. Thus, moisture monitoring is a crucial quality control task during timber drying that impacts the properties and quality of the subsequent timber structures. Variations in the average and distribution of the final moisture can result in significant variations in the timber properties and thus negatively affect performance in service. This necessitates rigorous moisture monitoring and timber grading during and after kiln drying. This study aims to estimate the kiln-dried population final moisture distribution based on green timber characteristics and machine learning. This will facilitate the timber grading process by reducing the variation in the mechanical properties of the kiln-dried timber, as timbers with a higher chance of moisture variation will be identified before the drying process. Timber initial moisture, initial weight, basic density, and target moisture were used to train machine learning models. Linear discriminant analysis and decision tree were used for moisture classification. Principal component and variable clustering analysis were performed to study the critical parameters affecting the timber overdrying and underdrying. The results indicated that initial moisture level and weight are the essential variables, while density had the least significant effect on the performance of classification models. The decision tree approach exhibited better performance than discriminant analysis with ~91% classification accuracy proving the effectiveness of using initial timber properties for quality control and grading of kiln-dried timber.