摘 要
地铁工程由于其施工技术复杂、不可预见致险因素多,是一项高风险的建筑工程。同时,地铁工程施工安全事故案例散落各处,未对其做系统整理和有效利用。本文提出基于案例(CBR)和基于规则(RBR)混合模式设计安全事故案例库的结构化模型,实现事故案例在风险分析、险情预判、应急抢险中的借鉴作用,减少相似事故案例重复发生。
本文搜集国内外(国内为主)71 个地铁工程施工安全事故案例,分析典型案例的表示形式、内容特征及统计规律,并通过对特征的相似性、案例库的有效性和可操作性分析,论证历史案例重用的可行性,这也是进一步研究 CBR 技术的基础。CBR关键技术包括知识表示和检索,本文通过总结安全事故案例的多层次属性特征,提出以标识属性、分类属性、工程特征属性、事故过程属性、事故原因属性、事故后果属性、事故处理措施属性、案例应用属性等为基本属性的案例框架表示法,并设计分层检索和最近邻检索算法相结合的检索模型,提高案例库求解效率。
针对单独使用 CBR 的局限性,本文在 CBR 系统中集成 RBR 系统,利用 RBR 制定案例检索策略,并设计 RBR 控制器实现灵活调用推理规则,实现主动干预推理过程,有效解决单一的 CBR 检索时遗漏最优解、相似度计算工作量大、解决方案适用
性差等缺点,增强系统自学习能力。
要真正实现 CBR-RBR 混合系统应用还需要很长一段路要走,本文只是这方面的步研究。如何建立完善的案例积累和共享机制,是实现系统功能最为基础的部分,同时,还可考虑将险兆事件列入案例库中,使之成为预防事故发生的重要一环。
关键字:地铁工程 安全事故 案例推理 知识表示 检索 混合模式
Abstract
Subway construction is a high risk of the construction works due to its complexityand unpredictability. Meanwhile, because subway construction accident cases had beenscattered around, we cannot neither arrange them systematically nor utilize them effectively. The paper presents case-based (CBR) and rule-based (RBR) mixed-mode todesign structural model of accident cases, which implements the reference function of riskanalysis, predict danger, emergency rescue to reduce the recurrence of similar accident cases.
This article collects 71 subway project construction safety incident cases to analyzethe expression form, the content characteristic and the statistical rule. Then, the possibilityof reusing the old cases that is also the basis for further study of CBR has beendemonstrated by the similarity of characteristicthe validity and operability of caselibrary. CBR key technologies include knowledge representation and retrieval. This articlepresents the frame knowledge representation by summarized multilevel attributes, likeproposed to identify attributes, classified attributes, engineering characteristics attributes,process attributes, cause attributes, consequences attributes and treatment measures asthebasic attributes. The combinational retrieval proposed is hierarchical search algorithm andthe nearest neighbor algorithm, which improves solution efficiency as well as the representation.
In the light of limitations of CBR alone, the CBR system has integrated RBR whichcan formulate the case retrieval strategy,at the same time, the RBR controller may transferthe inference rule nimbly. The entire process may solve the shortcomings like omittingoptimal solution, increasing similarity computation work, bad solution serviceability andenhance the system’s capability of self-learning.
A lot of effort needs to be done to truly achieve CBRRBR hybrid system application,but this is only preliminary research in this area. How to create a perfect caseaccumulation and sharing mechanism is the most basic part of the system function. Furthermore,to add near-miss events into the case base is also a good way to make it a
risk trillion important part of accident prevention.
Keywords : Subway construction Accident Case-based reasoning Knowledge representation Retrieval Hybrid mode