大规模数据驱动的运营隧道结构变形安全时空耦合建模与风险感知研究

Structural deformation damages in operational shield tunnels are impacted by various factors, such as geological and hydrological conditions, exterior loads, and human factors, resulting in complexities in analyzing the structural damage mechanism. The empirical knowledge of domain experts are heavily relied on in on-site damage inspection. With growing applications of structural health monitoring (SHM) and sensor network technologies, the conflict problem between rich data and poor knowledge is becoming more and more serious. The traditional manual approach has challenges in dealing with large-scale and complex data, leading to inaccuracy and ineffectiveness on perceiving the structural safety risk. In this project, the granular-based clustering and reduction methods, based on the knowledge discovery theory, are used to identify the dominant failure modes and their patterns, in order to have a better understanding of large-scale and complex data in the SHM domain. A hybrid approach that integrates the Copula theory and the data-driven modeling technique is proposed to investigate the time-varying correlation among various failure modes from a huge dataset. A time-space-coupling model is built for simulating the structural behaviors in operational tunnels, which further reveals the mechanism of structural failures over time. On this basis, a novel dynamic Bayesian network based safety risk perception method, that is capable of self-updating under given observations, is developed, which can provide real-time decision support for the prevention of structural deformation damages in operational tunnels.

运营盾构隧道长期服役诱发结构变形破坏受地质、水文、外部荷载及人因管理等多种因素共同作用，病变演化规律复杂，现场病害排查对经验知识依赖大。随着监测与传感网络技术的发展，丰富的数据与贫乏的知识间矛盾日见突出。传统人工处理方式对大规模复杂数据的处理与分析存在较大不足，影响了盾构隧道结构安全风险感知的时效性与精准性。本项目以知识发现为理论基础，结合粒计算聚类和约简处理方法，辨识盾构隧道结构变形破坏的主要失效模式及其特征变量，提高对大规模复杂数据和知识粒度结构的认知水平；融合Copula理论与数据驱动建模，提出面向大规模复杂数据的失效模式时变相关性分析方法，建立隧道结构变形破坏的时空耦合模型，进一步揭示运营盾构隧道结构病变的时空演化规律；在此基础上集成动态贝叶斯网络，提出具有模型自适应智能更新机制的盾构隧道结构变形安全风险感知评价方法，为盾构隧道结构安全病害事前、事中及事后管理提供实时动态决策支持。

运营盾构隧道长期服役诱发结构变形破坏受地质、水文、外部荷载及人因管理等多种因素共同作用，病变演化规律复杂，现场病害排查对经验知识依赖大。随着监测与传感网络技术的发展，丰富的数据与贫乏的知识间矛盾日见突出。传统人工处理方式对大规模复杂数据的处理与分析存在较大不足，影响了盾构隧道结构安全风险感知的时效性与精准性。本项目引入对海量数据挖掘和知识发现具有较强适应能力的粒计算理论，结合云模型和粗糙集等一系列信息粒化方法，辨识盾构隧道结构变形破坏的主要失效模式及其特征变量，提高了对大规模复杂数据和知识粒度结构的认知水平；融合Copula理论与数据驱动建模，提出面向大规模复杂数据的失效模式时变相关性分析方法，建立隧道结构变形破坏的时空耦合模型，进一步揭示了运营盾构隧道结构病变的时空演化规律；在此基础上集成动态贝叶斯网络，提出了具有模型自适应智能更新机制的盾构隧道结构变形安全风险感知评价方法，能够为盾构隧道结构安全病害事前、事中及事后管理提供实时动态决策支持。