关键词: 地铁火灾, 关键测点, 温度状态识别, 层次分析法, Kohonen自组织特征映射, 牵引电机

Abstract: In order to realize intelligent recognition of temperature state and related analysis for devices surface temperature states, an improved Analytic Hierarchy Process (AHP) model was introduced, which could dynamically analyze the relevance among several measuring points of temperature and selected key measuring point which could reflect the temperature state for devices. At the same time, Kohonen SelfOrganizing Feature Map (SOFM) neural network was established，which could update and follow and recognize temperature serials value of key measuring points during some time, so that show device temperature status. Take traction motor for example, Matlab software simulation analysis show its recognition rate is 89%,which effectively reduces the false positive rate of fire.

Key words: subway fire, key measuring point, temperature state recognition, Analytic Hierarchy Process (AHP), Kohonen Self-Organizing Feature Map (SOFM), traction motor