Magnesium oxychloride cement mortar (MOCM) coating has broad application prospects in engineering due to its excellent corrosion resistance. Based on orthogonal experimental design, the author systematically investigated the effects of matrix strength, MgO/MgCl₂ molar ratio, solution concentration, type and dosage of additives, and curing conditions on the corrosion resistance of coatings. 108 sets of measured data were obtained, including corrosion current density (Icorr), mass loss rate (Dm), and chloride ion diffusion coefficient (Dcoef). Unlike existing research that focuses on comparing the accuracy of a single algorithm, the innovation of this study lies in: ① For the first time, the grey wolf optimization algorithm is combined with BP neural network to construct a multi model comparative prediction framework for the corrosion resistance of magnesium oxychloride cement mortar coatings; ② Not only does it focus on improving prediction accuracy, but it also identifies molar ratio, concentration, and maintenance as stable dominant factors from both statistical and model dimensions through cross validation of sensitivity analysis and ablation experiments, enhancing the interpretability of the model; ③ Developed a supporting decision support software prototype, which extends from “data-driven prediction” to “engineering assisted decision-making”. On this basis, four prediction models including BPNN, RF, LSTM, and GWO-BPNN were constructed and compared. The results showed that GWO-BPNN performed the best, with RMSE and MAE predicted by Icorr being 0.029 and 0.023, respectively, and R²=0.962; RMSE = 0.040, R²=0.962; The RMSE = 1.02 × 10⁻¹³ of Dcoef, and R²=0.984. Sensitivity analysis showed that Ratio, Conc, and Curing were key factors, and ablation experiments further validated the importance of GWO optimization strategy and major variables. This study provides a reliable method for efficient prediction and optimization design of corrosion resistance of MOCM coatings.
Lv et al. (Fri,) studied this question.