In the Gobi region, concrete structures frequently suffer erosion from wind gravel flow. This erosion notably impairs their longevity. Therefore, creating a predictive model for wind gravel flow-related concrete damage is crucial to proactively address and manage this problem. Traditional theoretical models often fail to predict the erosion rate of concrete (CER) structures accurately. This issue arises from oversimplified assumptions and the failure to account for environmental variations and complex nonlinear relationships between parameters. Consequently, a single traditional model is inadequate for predicting the CER under wind gravel flow conditions in this region. To address this, the study utilized a machine learning (ML) model for a more precise prediction and evaluation of CER. The support vector machine (SVM) model demonstrates superior predictive performance, evidenced by its R² value nearing one and a notable reduction in RMSE 1.123 and 1.573 less than the long short-term memory network (LSTM) and BP neural network (BPNN) models, respectively. Ensuring that the training set comprises at least 80% of the total data volume is crucial for the SVM model’s prediction accuracy. Moreover, erosion time is identified as the most significant factor affecting the CER. An enhanced theoretical erosion model, derived from the Bitter and Oka framework and integrating concrete strength and erosion parameters, was formulated. It showed average relative errors of 22% and 31.6% for the Bitter and Oka models, respectively. The SVM model, however, recorded a minimal average relative error of just −0.5%, markedly surpassing these improved theoretical models in terms of prediction accuracy. Theoretical models often rely on simplifying assumptions, such as linear relationships and homogeneous material properties. In practice, however, factors like concrete materials, wind gravel flow, and climate change are nonlinear and non-homogeneous. This significantly limits the applicability of these models in real-world environments. Ultimately, the SVM algorithm is highly effective in developing a reliable prediction model for CER. This model is crucial for safeguarding concrete structures in wind gravel flow environments.

The Qinghai–Tibet Plateau is the main permafrost area in China. Concrete structures constructed on permafrost are affected by the early negative-temperature environment. In particular, the negative-temperature environment seriously affects the strength growth process and the frost resistance of concrete (FRC). Therefore, this study considered the influence of the gas content, water–binder ratio (w/b), age, and other factors on the strength variation law and FRC under −3 °C curing conditions. Nuclear magnetic resonance (NMR) was used to analyze the pore structure of concrete before and after freeze–thaw cycles (FTCs). The results showed that the compressive strength of the concrete (CSC) under −3 °C curing was only 57.8–86.4% of that cured under standard conditions. The CSC under −3 °C curing showed an obvious age-lag phenomenon. The FRC under −3 °C curing was much lower than that under standard curing. The porosity of the concrete under −3 °C curing was greater, with a higher percentage of harmful and multi-harmful pores than that under standard curing. The concrete properties deteriorated primarily because curing at −3 °C hindered the hydration reaction compared with standard methods. This hindrance resulted in diminished hydration development, weakening the concrete’s structural integrity. Under both curing conditions, when the gas content was between 3.2% and 3.8%, the frost resistance was the best. This is because a gas content within this range effectively enhances the internal pore structure, therefore relieving the swelling pressure caused by FTCs. Based on the freeze–thaw damage (FTD) model proposed by previous authors, a new model for the CSC under −3 °C curing reaching that of the concrete under standard curing for 28 d was established in this study. This advanced model was capable of accurately assessing the FTD of concrete structures in permafrost regions. Finally, the life expectancy of concrete in Northwest China was predicted. The life of the concrete reached 46.9 years under standard curing, while the longest life of the concrete under −3 °C curing was only 12.9 years. Therefore, attention should be paid to constructing and curing concrete structures in cold environments.

In high-elevation or high-latitude permafrost areas, persistent subzero temperatures significantly impact the freeze–thaw durability of concrete structures. Traditional methods for studying the frost resistance of concrete in permafrost regions do not provide a complete picture for predicting properties, and new approaches are needed using, for example, machine learning algorithms. This study utilizes four machine learning models—Support Vector Machine (SVM), extreme learning machine (ELM), long short-term memory (LSTM), and radial basis function neural network (RBFNN)—to predict freeze–thaw damage factors in concrete under low and subzero temperature conservation conditions. Building on the prediction results, the optimal model is refined to develop a new machine learning model: the Sparrow Search Algorithm-optimized Extreme Learning Machine (SSA-ELM). Furthermore, the SHapley Additive exPlanations (SHAP) value analysis method is employed to interpret this model, clarifying the relationship between factors affecting the freezing resistance of concrete and freeze–thaw damage factors. In conclusion, the empirical formula for concrete freeze–thaw damage is compared and validated against the prediction results from the SSA-ELM model. The study results indicate that the SSA-ELM model offers the most accurate predictions for concrete freeze–thaw resistance compared to the SVM, ELM, LSTM, and RBFNN models. SHAP value analysis quantitatively confirms that the number of freeze–thaw cycles is the most significant input parameter affecting the freeze–thaw damage coefficient of concrete. Comparative analysis shows that the accuracy of the SSA-ELMDE prediction set is improved by 15.46%, 9.19%, 21.79%, and 11.76%, respectively, compared with the prediction results of SVM, ELM, LSTM, and RBF. This parameter positively influences the prediction results for the freeze–thaw damage coefficient. Curing humidity has the least influence on the freeze–thaw damage factor of concrete. Comparing the prediction results with empirical formulas shows that the machine learning model provides more accurate predictions. This introduces a new approach for predicting the extent of freeze–thaw damage to concrete under low and subzero temperature conservation conditions.