Abstract: Based on the dataset of the hydrolysis process for preparing pseudo-boehmite and γ-Al₂O₃ from aluminum alkoxide, segmented linear regression models were constructed to link the preparation conditions with the pore volume and specific surface area of pseudo-boehmite and γ-Al₂O₃, and the optimal model structures were determined. The adjusted coefficient of determination(R²) of all models were satisfactory, and all models passed the Davies significancetest for segmented variables. This indicates that, while maintaining model simplicity, segmented linear regression models can provide high fitting accuracy, reasonable complexity, and strong stability, effectively revealing the complex intrinsic relationships between preparation parameters and the physicochemical properties of the products. Validation results showed that the maximum relative error in predicting the pore volume and specific surface area of pseudo-boehmite and γ-Al₂O₃ under known preparation conditions was no more than 4.80%. The models developed in this study have high precision and can be used to guide the design of preparation conditions and predict the physicochemical properties of pseudo-boehmite and γ-Al₂O₃, providing significant guidance for the precise control of their physicochemical properties.

Key words: pseudo-boehmite, γ-Al₂O₃, preparation process, segmented linear regression model