基于规则学习元胞自动机的桥梁钢结构点蚀涌现预测理论

点蚀在公路钢结构及组合结构桥梁中受到材料变异、荷载和作用随机性的影响，呈现散在发生、各异演化的特征。现有腐蚀管控的半经验养护策略和被动式技术手段难以满足实际预测需求。点蚀涌现机理成为钢结构精准管养的关键问题。目前材料领域的点蚀研究在实现工程尺度点蚀涌现预测方面仍有难度。本项目另辟蹊径，从材料和工程层面研究差异和联系入手，构建全新的试验方法、建模技术和模拟理论。首先，开展大尺寸点蚀涌现试验，揭示场域随机性的影响；同时，采用统计体积单元法将微观层面的钢材变异性转译到工程尺度；在此基础上，利用机器学习方法从图像数据中辨识影响点蚀演化的元胞自动机状态参数，主动学习局部更新规则，发展精确高效的点蚀涌现模拟方法；最后，揭示钢结构桥梁服役期间的点蚀涌现规律，为钢结构日常管养提供决策支持。研究成果将形成同时考虑材料变异性和场域随机性的点蚀涌现预测理论，解决材料性能退化和结构性能演化成果沟通的前沿问题。

The pitting corrosion in steel structure and composite structure bridges is affected by the variation of material, the randomness of loads as well as actions, showing the characteristics of scattered occurrence and differentiated evolution. The existing semi-empirical maintenance strategy and passive corrosion control techniques are struggling to meet the practical forecast demands. The emergence mechanism of pitting corrosion has become a vital issue in the precise maintenance of steel structures. The research on pitting corrosion in the field of materials is still struggling to enable the prediction of pitting corrosion at the engineering level. This project will take a different approach, which starts from the difference and correlation between the research on material and engineering levels. It endeavors to construct new test methods, modeling techniques, and simulation theories. First, it will carry out large-scale pitting erosion experiments, unveiling the influence of the field randomness. Meanwhile, the statistical volume unit method is used to translate the variability of the steel at the micro level to the engineering level. On this basis, machine learning methods are used to identify the cell state parameters of cellular automata that affect the evolution of pitting corrosion from image data. It aims to learn the local update rules of cellular automata actively and develop an accurate and efficient pitting corrosion simulation method. Finally, it will reveal the law of pitting corrosion during the service of steel bridges, which provides decision support for the daily maintenance of steel structures. The results will form a pitting corrosion prediction theory that considers both material variability and field randomness and solves the cutting-edge problem of how to transfer the mechanism of material performance degradation into the interpretation of structural performance evolution.