Machine-Learning-Assisted Evaluation of Ethylene Removal by Torrefied Sawdust for Mushroom Storage Applications

초록

Ethylene control is important for maintaining the post-harvest quality of fresh mushrooms during refrigerated storage and distribution. In response to the need for renewable alternatives to conventional potassium permanganate-based scavengers, this study evaluated heat-treated sawdust as a biomass-derived ethylene removal material. Ethylene removal was examined using an adsorbent pouch system containing torrefied sawdust alone or mixed with a commercial ethylene remover. The experimental design focused on three practical operating variables: torrefied sawdust ratio, particle size, and exposure time. Ethylene removal was evaluated over the tested exposure periods. Data-driven regression models were subsequently used as supplementary tools to capture nonlinear relationships among formulation and operating variables within the investigated range. Among the evaluated models, XGBoost showed the most stable prediction performance (R² = 0.866; RMSE = 6.89%), while SHAP analysis indicated that exposure time and adsorbent dosage were the most influential variables. The graphical interface developed from the model is presented as a decision-support tool rather than as independent evidence of material performance. Overall, the results indicate that torrefied sawdust has potential as a biomass-based ethylene-removal material under the tested closed-chamber conditions. However, its effectiveness in preserving mushroom quality requires further validation through direct measurements of mass loss, firmness, color, and microbial quality.

키워드