基于不确定性理论和贝叶斯网络的地铁隧道施工环境变形安全实时预警控制研究

The surrounding environment deformation safety induced by tunnel excavation involves the interaction among various risk factors, such as geological conditions, construction methods, environmental parameters, organizational management and others. However, there exists much uncertainty of randomness and fuzziness, as well as the complex correlations between those various factors, which contribute to the high complexity in cataclysmic law of tunnel excavating disasters. As a sequence, the real-time analysis of the environment safety status is greatly influenced, resulting in challenges in accurate decision making for safety control in tunnel construction. It is therefore necessary and meaningful to define those rand fuzzy uncertainty, discover the complex and hidden correlations, and then establish safety forewarning models for real-time decision support analysis in tunnel construction. In this research, cloud model and cloud transform methods are first utilized to deal with the rand fuzzy uncertainty information based on uncertainty theory, providing efficient data for precisely analyzing the potential forewarning information with randomness and fuzziness fully considered. Then, a multi-dimension association rule mining algorithm, Improved Apriori Algorithm, is proposed to investigate the complex correlations among various risk factors, aiming to discover and reveal and the environment disaster mechanisms during the tunnel construction. Finally, Bayesian Networks (BNs) which are sensitive and precise to time-space change are employed to build up the forewarning decision model for real-time safety control in tunnel construction, attempting to realize the safety forewarning management in the entire life cycle with pre-accident, during-construction and post-accident control included. Furthermore, a safety forewarning system for environment safety control in tunnel construction is developed to perfect the safety monitoring and forewarning mechanism, making an effort to promote the development and progress of disaster prevention theory, as well as its application potential.

地铁隧道施工诱发环境变形安全涉及地质、施工方法、环境、管理等多源因素，这些多源致险因素存在大量的模糊随机不确定性及关联性，灾变规律复杂，影响了隧道施工环境变形安全状态实时分析与准确判断，有效分析这些多源不确定性因素，挖掘其存在的复杂关联性，建立实时安全预警决策模型一直是地铁施工安全预警控制的难题。本项目基于不确定性理论，利用云模型与云变换方法处理施工中多源致险因素的模糊随机不确定性，为全面准确分析多源数据中潜藏的安全征兆信息提供数据基础；运用多维关联规则挖掘改进Apriori算法研究环境变形安全及其多源致险因素之间的复杂关联性，加强认知和揭示隧道施工诱发环境变形灾变规律；在此基础上，利用对时空演化敏感的贝叶斯网络构建隧道施工环境变形安全实时决策预警模型，实现事前、事中及事后实时预警；研发隧道施工安全风险实时预警管理平台，完善施工安全监控预警机制，推动防灾减灾理论与应用技术的发展和进步。

本项目基于不确定性理论，利用云模型与云变换方法处理施工中多源致险因素的模糊随机不确定性，为全面准确分析多源数据中潜藏的安全征兆信息提供数据基础；运用多维关联规则挖掘改算法研究环境变形安全及其多源致险因素之间的复杂关联性，加强认知和揭示隧道施工诱发环境变形灾变规律；在此基础上，利用对时空演化敏感的贝叶斯网络构建隧道施工环境变形安全实时决策预警模型,研发隧道施工安全风险实时预警管理平台，完善施工安全监控预警机制，推动防灾减灾理论与应用技术的发展和进步。