Abstract: For such attribute redundancy in earthquake-damaged risk system for underground engineering, a reduction algorithm is proposed for Rough set based on improved genetic algorithm with the prerequisite condition about classification accuracy and key elements. Following the basic principles of the earthquake-damaged risk system design for underground engineering, classification requirements and analysis mechanisms are advanced for earthquake-damaged risk system, and the attribute dependence and importance are described for earthquake-damaged risk system with Rough set method; then, nonlinear optimization model is erected on minimum attribute reduction operations about earthquake-damaged risk system for underground engineering. It translates such model into the searching problem of the best fitness individual in the genetic evolution process, and improves the selection, crossover, mutation and elitist mechanism of genetic populations, which keeps the overall optimization features and faster convergence rate. The simulation result shows that the method can quickly and efficiently determine the minimum reduction about the earthquake-damaged risk system for underground engineering, which establishes a good foundation for the deep application of Rough set method into earthquake-damaged problems for underground engineering.

Key words: earthquake, underground engineering, risk system, Rough set, genetic algorithm, attribute reduction

摘要： 针对地下工程震害风险体系中属性冗余问题，提出了一种保证分类识别精度和关键要素为前提条件的地下工程震害风险体系粗集约简方法。遵循地下工程震害风险体系设计的基本原理，提出了震害风险体系约简的分类需求和分析机制，应用粗糙集方法描述地下工程震害风险体系中的属性依赖度和重要度，构建地下工程震害风险体系最小属性约简的非线性优化模型，并将其转化为遗传进化过程中最佳适应度个体的搜索问题，改进遗传种群的选择、交叉、变异、精英保留机制，使得算法既保持整体优化特性，又具有较快的收敛速度。实验仿真表明，该方法能快速有效地求出地下工程震害风险体系的最小约简，为粗集方法在地下工程震害问题中的深入应用建立了良好基础。

关键词: 地震, 地下工程, 风险体系, 粗集, 遗传算法, 属性约简