基于文本挖掘和变权AHP法的隧道施工期安全状态评估

Safety Assessment of Tunnel Construction Based on Text Mining Technology and Variable Weight AHP Method

安全检查表法及其衍生方法作为工程领域中广泛、详尽的安全评价方法之一,其所得评价结果十分详细.然而,安全检查表法在数据获取方面存在明显的局限性,该方法虽然能够根据国家或地方法律法规、企业事故经验、科研所得事故及事故树的特点进行预先编制,但数据输入工作依然需要进行繁复的人工操作,使得整个安全评价流程的效率大幅降低.为了解决这一问题,本文拟保留安全检查表法的核心分析构架,对不同等级的安全隐患进行量化评分;通过调研提炼出与隧道施工期安全隐患相关的众多关键词,依靠文本挖掘技术从隧道建设相关责任方汇集的各类文本资料中提取与工程安全相关的关键词,并构建层次分析指标体系对关键词进行归类和定量评价;引入惩罚型变权模型对层次分析法的量化评价过程进行优化.依托工程实例证明:(1)整体而言,变权分析做的评价结论相较于常权层次分析所得分析结果更为保守,变权综合评价值一般小于等于常权综合评价值,当评价指标状态值趋于满分状态时,变权与常权层次分析所得评价值基本相同;(2)引入惩罚型变权模型后,惩罚区间内指标状态值越差对应的变权权重越大,状态值越好对应的权重则会小幅降低,这一原理可有效改善优秀指标状态值对不良指标状态值的中和作用,使极端条件下不良指标的状态值在整个系统的综合评价中具有更大的影响力.

As one of the most widely used and detailed safety evaluation methods in engineering field, the safety checklist method and its derivation methods produce evaluation results at length.However, this method is limited in the data acquisition. Although the safety checklist method can be pre-programmed according to national or local laws and regulations, corporate accident experience, accidents and accident tree characteristics from scientific researches, complicated manual operations are required for data entry, which reduces the efficiency of the safety assessment. In order to solve this problem, this paper intended to retain the core analysis framework of the safety checklist method and quantify the security risks at different levels. Through research, many key words related to the security risks during tunnel construction were extracted, and those related to project safety wereextracted from various text data collected from the relevant responsible parties in tunnel construction by Text Mining technology. An index system of AHP was constructedto classify and quantitatively evaluate the keywords, and the penalty Variable Weight AHP model wasintroducedto optimize the process of quantitative evaluation of AHP. Relying on engineering examples, it is proved that:(1)Generally, the evaluation conclusion of variable weight AHP analysis is more conservative than that of constant weight analytic hierarchy process, and the comprehensive evaluation value of variable weight is generally less than or equal to corresponding value of constant weight. When the state value of evaluation index tends to be full score, the evaluation value of variable weight and constant weight analytic hierarchy process is basically the same;(2)After introducing the penalty variable weight model, the worse the state value of the index in the penalty interval, the bigger the weight of the variable weight, the better the state value, the smaller the corresponding weight. This principle can effectively improve the neutralization effect of the state value of the excellent index on the state value of the bad index, and make the state value of the bad index under extreme conditions have greater influence on the comprehensive evaluation of the whole system.