Abstract:

Objective: Considering the current lack of clarity regarding the influence mechanism and the insufficient quantitative prediction of the physical and mechanical properties of bamboo scrimber resulting from the drying process parameters of phenolic resin impregnated heat-treated bamboo bundles (PHB), the Grey Wolf Optimization algorithm combined with a Back Propagation Neural Network (GWO-BPNN) was employed to predict and elucidate the response relationship between the drying process parameters of PHB and bamboo scrimber performance. This study aims to provide a theoretical foundation for achieving efficient PHB drying and precise control over the properties of bamboo scrimber. Method: With the drying temperature, drying time, and moisture content after drying of PHB as input variables, and the water absorption rate, thickness swelling rate, width swelling rate, modulus of rupture (MOR), modulus of elasticity (MOE), and horizontal shear strength (HSS) of bamboo scrimber as output variables, a dataset of physical and mechanical properties was systematically constructed. On this basis, a GWO-BPNN model was constructed and trained. Subsequently, the performance of the proposed model was comprehensively evaluated using five key indicators: Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R²). Finally, the developed GWO-BPNN model was applied to a set of independent new datasets for validating the reliability of the proposed model. Result: The GWO-BPNN model demonstrated excellent adaptability and prediction accuracy in estimating the water absorption rate, MOR, MOE, and HSS of bamboo scrimber. The R² was closed to or surpassed 0.9, and the MAE, MSE, RMSE, and MAPE were maintained at relatively low levels. However, the prediction performance for the water absorption thickness swelling rate was moderate (R² = 0.78), and the prediction for the width swelling rate was unsatisfactory (R² = 0.11). Verification results indicated that the overall predicted values of the model exhibited high consistency with the actual measured values. As the drying temperature increased from 50 °C to 80 °C, the water absorption rate and thickness swelling rate of bamboo scrimber rose, whereas the variation in the width swelling rate remained no significant change. The MOR, MOE, and HSS generally followed a trend of initially increasing and subsequently decreasing. When the drying temperature of PHB was set at 60 °C and the moisture content was 10%, the three mechanical property indicators achieved their optimal values. As the moisture content of PHB increased from 5% to 20%, the water absorption rate and width swelling rate of bamboo scrimber exhibited a downward trend. Notably, the thickness swelling rate under a 5% moisture content condition was significantly higher compared to other moisture content conditions, and the mechanical properties demonstrated a trend of first increasing and then decreasing. Conclusion: The GWO-BPNN model proves highly effective in predicting the relationship between the drying process parameters of PHB and the physical and mechanical properties of bamboo scrimber. Specifically, the model exhibits strong predictive accuracy for key performance indicators, including water absorption rate, MOR, MOE, and HSS, with correlation coefficients (R) surpassing 0.9 across all these parameters. The model demonstrates excellent overall fitting accuracy and predictive capability. However, its performance in predicting thickness and width swelling rates resulting from water absorption is relatively less accurate. Based on these findings, the model can provide a solid theoretical basis for optimizing critical drying parameters such as drying temperature, drying duration, and final moisture content of PHB, and thereby facilitates more efficient drying processes and precise control over the performance attributes of bamboo scrimber.

Key words: drying of PF resin impregnated heat-treated bamboo bundles, bamboo scrimber, physical and mechanical properties, grey wolf optimization (GWO), back propagation neural network (BPNN)