地铁深基坑施工安全专项方案知识重用建模及优化研究
张永成
摘 要
城市轨道交通工程建设是一项复杂的高风险活动。安全事故的发生既有施工环境复杂、风险突发等等客观因素，也有施工方案不合理等主观原因。作为保障施工安全的重要依据文件，安全专项方案编制是一项复杂的知识密集型活动，具有涉及面广、技术体系内容庞杂、富含经验性知识等特点，而以往的工程方案蕴藏了宝贵的知识信息，使得在新方案编制过程中对以往方案价值知识信息的需求尤为强烈。本文从知识管理的视角，将安全专项方案知识信息与工程信息模型有机结合，运用案例推理、贝叶斯网络以及多目标优化等方法，进行安全专项方案知识信息的表达、重用和优化研究，以支持新建工程安全专项方案的制定与运用。主要研究内容如下：
（1）施工安全专项方案知识信息表达建模研究。地铁车站深基坑工程安全专项方案是一个复杂的多维多场景知识信息集合。首先，通过对知识信息表达的需求分析，提出了面向重用的知识信息建模要求和建模原则。然后，分析了深基坑工程安全专项方案知识信息构成内容，基于知识元模型，研究了专项方案知识信息内容元素的组织与表达。最后，结合知识信息建模要求和原则，构建了基于知识元的安全专项方案框架表达模型，为专项方案重用及优化提供基础。
（2）施工前安全专项方案知识信息重用推理研究。专项方案知识信息重用的核心是方案知识信息的检索推理。针对方案推理中影响方案检索效果的三个重要参数：属性取值、相似度函数和案例参考权值，并行优化以支持最优相似方案检索。引入区间二型模糊集法以考虑表征专项方案属性的不确定性特点；提出相似度集成计算方法以避免传统数值统一转化而选择单一函数计算相似度造成信息损失的缺点；提出基于前馈调整权重的k近邻法以考虑检索案例对新工程案例影响度的差异。然后，分析了基于相似度的知识信息相似融合方法，以进一步精炼方案知识信息。最后，基于规则对检索得到的案例方案进行调整，以支持新建工程安全专项方案的制定。
（3）施工中基于现场信息的安全专项方案深化优化研究。首先，以地铁车站深基坑为对象，构建施工风险因素与风险事件贝叶斯网络模型，系统地辨识影响风险发生的关键因素，基于工程现场获取风险因素信息以开展项目安全抗风险能力评价。其次，系统分析了专项方案现场运用中安全风险控制措施选择的决策过程，得到影响安全措施选择决策的三个主要因素：风险控制措施效果、措施费用和措施工期影响。基于此，构建了满足施工现场安全抗风险能力要求的安全措施多目标优化模型，并引入遗传算法求取该模型的最优解，支持安全专项方案的动态深化优化。
本研究将知识管理中的相关分析模型与方法贯穿应用于地铁车站深基坑工程安全专项方案制定与优化的过程中。本研究期望为安全专项方案制定和优化提供良好的方法，而高质量的安全专项方案有助于更好地保障地铁车站深基坑工程安全建设。
关键词：地铁深基坑工程；安全专项方案；知识信息重用；案例推理；多目标优化模型
Abstract
Subway construction has become a complex high-risk activity. The occurrence of accidents has both objective factors such as complex construction environment, and subjective ones such as lack of construction plans or irrational plans. The special safety program is an important textual document that guarantees construction safety. The special program preparation for a new project is a complex knowledge-intensive activity with the features of wide coverage, complicated technical system, and higher requirements for experience-related knowledge. The demand for past and valuable program knowledge in the special program preparation is particularly strong. Therefore, from the perspective of knowledge management, this paper uses Case-based reasoning, Bayesian networks, and multi-objective optimization method to express, reuse and optimize based on the safety special program knowledge to support the formulation and optimization of special safety programs for new construction projects. The research content is as follows:
(1) Knowledge expression method for safety special plan. The special scheme of deep foundation pit in subway station is complex multi-dimension information. Firstly, it proposed the requirements of reuse-oriented knowledge information modeling to build the principle and criteria based on the demand analysis. Then, special schema is organized and expresed by using knowledge elements. Finally, a framework model based on knowledge elements was developed under the purpose and principles of knowledge information modeling, which provides a foundation for the next step of the special program reasoning.
(2) Knowledge information reusing method of special schema for risk control. Several important parameters affecting case retrieval were optimized. Firstly, considering the uncertainty of the special plan, the interval type-2 fuzzy set method was adopted to express the indefinite characteristics of the project attributes more realistically. A similarity integrated calculation method is proposed, which avoids the traditional numerical unified transformation and the information loss and inaccurate results caused by the using of a single function to calculate the similarity. Then, based on the different impacts of search cases on new engineering cases, this study proposed an improved k-nearest neighbor case retrieval method in terms of the idea of feed-forward kernel function. By defining the program's security index, the effect on weights of the feed-forward retrieved case on the new case were presented, which support the retrieval of optimal and most similar case searches. The retrieved case will support the development of the safety special program.
(3) Dynamic optimization method of safety measures in the specific security plan based on construction site. Safety measures, as an important part of a special security program, have significant time dimension characteristics. A Bayesian network model for risk factors and risk events is systematically constructed, which explains the mechanism of safety accidents and identifies the key factors. The causality matrix of risk measures and risk factors is proposed and risk assessment can be conducted. The decision-making process in the implementation of safety measures is systematically analyzed, and four factors influencing the decision-making are obtained: risk control, inherent risks, cost of measures, and duration influence of measures. A multi-objective optimization model that satisfies the requirements of the safety and anti-risk ability in the construction site is proposed. Genetic algorithm is introduced to obtain the optimal solution of the model to support the dynamic optimization decision-making of the special security program.
Relevant decision-making models and methods in knowledge management are applied throughout the process of formulating and optimizing specific security programs for deep foundation excavation, It hopes to provide good methods for the formulation and optimization of specific security plan. More high-quality specific security programs help to better guarantee the safety construction of deep foundation pit at subway stations.
Key words: Deep foundation excavation of subway; Construction functional plan for safety; Knowledge information reusing; Case-based reasoning; Multi-objective optimization modeling
国家数字建造技术
创新中心
