溃坝事故中的人因失误表征与量化方法研究

The early dam failure accidents are mostly caused by catastrophe like flood or earthquake and the engineering quality problems. Progress in technique and improvement of management have greatly reduced these kind of accidents, while the proportion of accidents caused by human factors increased relatively. At present, less research has been done on the characterization methods of human factors in dam failure accidents. The lack of scientific cognition of error model and unestablished method of error measure affect dam risk prevention and control. In this research project, 3500 dam failure accidents so far in China are systematically teased out. The cause of human factors and the derivation process are analyzed and the influence factor set of human factors is construct categorically by data mining. Based on the theory of behavior decision, basic paths of human factors in the stage of observation, diagnosis and operation are revealed. Human factors model and key factors are present by introducing human reliability theory and typical map of human factors model is drawn by classification. By synthesizing cognition reliability model HCR and prediction technique of human factors THERP, human factors event tree of dam failure is established. Probability calculation models of observation error, diagnosis error and operation under various dam failure model are proposed. Calculation method of human total error probability is introduced. The project can promote the development of scientific cognition of human error effect in dam failure accidents and improve the deficiencies of subjective probability estimation of human error in dam failure probability calculation, which has certain theoretical value and application future.

早期溃坝事故多因洪水等灾变环境和工程质量问题所致，科技进步和管理水平提高使得这类溃坝事故显著减少，人因失误触发的溃坝事故比例反而相对上升。目前，溃坝事故中人因失误表征方法研究较少，失误模式缺乏科学认知，失误度量方法尚未建立，影响大坝风险防控。本项目系统梳理我国迄今3500余座溃坝资料，剖析人因失误原因及衍生过程，通过数据挖掘分类构建人因失误影响表征因子集；基于行为决策理论揭示察觉、诊断、操作三阶段的人因失误基本路径，引入人因可靠性理论提出人因失误模式及关键影响因子，分类创建人因失误模式典型图谱；综合认知可靠性模型HCR和人误率预测技术THERP，构建溃坝事故人因失误事件树，提出不同溃坝模式下的察觉失误、诊断失误、操作失误概率计算模型，建立人因总失误概率计算方法。本项目可促进溃坝事故中人因失误影响规律科学认知，改进溃坝概率计算中人因失误概率估算过于主观的不足，具有较高的理论意义和应用前景。

早期溃坝事故大多是由洪水等环境因素和工程质量问题引起，随着科技进步和管理水平的提高，人因失误在大坝溃坝事故中所占的比例上升，我国3500余座历史溃坝事故中，不少本可以避免或减少损失，但由于当时内外部多种不利因素，导致管理人员认知和行为失效。因此，大坝安全管理中对人的可靠性分析至关重要。本项目对我国历史上的典型溃坝案例按溃坝时间、坝型、溃坝原因分类，分析其中的人因失误根原因，建立人因失误影响因素集，运用人工神经网络技术，挖掘各溃坝情境下不同因素的重要性系数，得到不同溃坝模式下的主要人因失误影响因素。借鉴HFACS、CREAM等方法，研究溃坝事故中的人因失误基本路径，主要包括监视觉察、状态诊断、计划制定、操作执行四个阶段，归类分析得出各阶段的主要影响因素，与人因失误经典模式中显现型与潜在型失误模式对应，构建溃坝事故人因失误因果关系图，确定行为影响因子PSF并使用层次分析法找到重要影响因子，描绘溃坝事故中人因失误模式典型图谱。根据认知可靠性模型HCR与人误预测率技术THERP，将THERP+HCR引入溃坝事故人误概率分析，以湖北省崇阳县破冲口水库漫坝事故为例，计算觉察阶段、诊断阶段、操作阶段的人因失误率，建立人因总失误概率计算方法。