Artificial intelligence approach to predict microfibril angle of cellulose in wood cell walls by wide-angle X-ray diffraction

Abstract

In the cell wall of cellulose-based fibers such as wood, the microfibril angle (MFA) in the S2 layer plays a crucial role in determining anisotropic properties. Current Wide-angle X-ray diffraction (WAXD) methods for MFA prediction rely on empirical equations, lacking clear predictive capabilities and remaining stagnant for decades. This study presents a novel approach to predict MFA and its variability using a generalized diffraction equation, Monte Carlo simulations of diffraction patterns, and Machine Learning models, including Random Forest (RF), k-Nearest Neighbors (kNN), and Artificial Neural Networks (ANNs). Results show that the commonly used Variance Approach generates inaccurate predictions (RMSE=2.61°, MAE=2.12°), while the proposed AI models demonstrate significantly higher accuracy (RF: RMSE=0.72°, MAE=0.29°; kNN: RMSE=0.87°, MAE=0.40°; ANN: RMSE=0.47°, MAE=0.24°). Furthermore, the AI models suggest that empirical cross-section shape data is not required for accurate MFA prediction. This innovative approach, leveraging advanced computational methods and AI, addresses long-standing challenges in MFA prediction using WAXD.