Sulfuric acid corrosion poses a significant threat to urban ‘deep tunnel’ systems, exacerbating flood disasters and causing economic losses. Therefore, it is crucial to predict the deterioration process of steel fiber-reinforced concrete under sulfuric acid corrosion. This study developed four predictive models for the degradation process of steel fiber-reinforced concrete of ‘deep tunnel’ lining during service characterized by the corrosion depth and strength loss under sulfuric acid corrosion. The database consisted of various numerical data obtained from a validated numerical model. Based on correlation and grey relational analysis, the concrete’s historical maximum load, initial corrosion depth, and load for 10 time points were selected as input data, while the output parameters are the corrosion depth and strength loss at these 10 time points. It was found that the Transformer-LSTM model not only exhibited the highest prediction accuracy but also demonstrated robust performance in the continuous prediction mode. Furthermore, after incorporating additional training data, the Transformer-LSTM model was able to predict the variation in corrosion depth and strength loss of steel fiber-reinforced concrete under sulfuric acid corrosion over a 120-year period under different loading conditions, which gives the maximum corrosion depth of 8.00 mm and strength loss of 15.06 MPa, with maximum prediction errors of 0.37 mm and 0.63 MPa, respectively. This prediction model can effectively capture the deterioration process of steel fiber-reinforced concrete under sulfuric acid corrosion during the service life of ‘deep tunnel’ structures, thereby, offering valuable guidance on scheduled maintenance and corrosion area repairs for the long-term serviceability of ‘deep tunnel’ structures.

中文翻译：

---

硫酸腐蚀下“深层隧道”钢纤维混凝土衬砌劣化过程预测

硫酸腐蚀对城市“深隧道”系统构成重大威胁，加剧洪水灾害并造成经济损失。因此，预测硫酸腐蚀下钢纤维混凝土的劣化过程至关重要。本研究开发了 4 个预测模型，用于“深层隧道”衬砌钢纤维增强混凝土在服役过程中的降解过程，其特征是硫酸腐蚀下的腐蚀深度和强度损失。该数据库由从经过验证的数值模型获得的各种数值数据组成。基于相关性和灰色关联分析，选取混凝土的历史最大荷载、初始腐蚀深度和 10 个时间点的荷载作为输入数据，输出参数为这 10 个时间点的腐蚀深度和强度损失。结果发现，Transformer-LSTM 模型不仅表现出最高的预测精度，而且在连续预测模式下也表现出稳健的性能。此外，在结合额外的训练数据后，Transformer-LSTM 模型能够预测不同加载条件下 120 年硫酸腐蚀下钢纤维混凝土腐蚀深度和强度损失的变化，其中最大腐蚀深度为 8.00 mm，强度损失为 15.06 MPa，最大预测误差为 0.37 mm 和 0.63 MPa， 分别。 该预测模型可以有效捕捉“深层隧道”结构在“深层隧道”结构使用寿命期间在硫酸腐蚀下的劣化过程，从而为“深层隧道”结构的长期适用性提供有价值的定期维护和腐蚀区域修复指导。