基于孤立森林的电梯异常状态非侵入式检测  

Non-invasive Detection of Elevator Abnormal State Based on Isolation Forest

作　　者：郭道靖 孙笠文 张月 唐靓 GUO Daojing;SUN Liwen;ZHANG Yue;TANG Liang(Zhenjiang Power Supply Branch of State Grid Jiangsu Electric Power Co.,Ltd.,Zhenjiang 212000,China;Zhenjiang Dazhao Electric Power Construction Co.,Ltd.,Zhenjiang 212000,China)

机构地区：[1]国网江苏省电力有限公司镇江供电分公司,江苏镇江212000 [2]镇江大照电力建设有限公司,江苏镇江212000

出　　处：《微型电脑应用》2025年第1期272-275,共4页Microcomputer Applications

摘　　要：电梯自动化运行效率是保障民生安全不可或缺的重要因素之一。为了进一步提高电梯异常状态入侵检测能力,设计一种基于孤立森林的电梯异常状态非侵入式检测方法。该方法根据基线判断振动信号状态,通过内部特征尺度分解(ICD)信号特征,并采用孤立森林算法检测异常数据,具有实时效果。结果表明,振动分析结果位于2σ区间的数据比例超过95%,满足统计过程控制要求,能够获得可靠的结果;异常值检测结果均高于99.6%,选择正常值比例作为电梯系统状态判断指标,可以利用孤立森林模型对电梯运行状态与正常状态之间的差异性进行判断。The efficiency of elevator automatic operation is one of the indispensable factors to ensure the safety of people’s livelihood.In order to further improve the ability of elevator abnormal state invasive detection,a non-invasive detection method based on isolation forest is designed.The method determines the vibration signal state according to the baseline,analyzes the signal characteristics through internal characteristic-scale decomposition(ICD),and uses isolation forest algorithm to detect the abnormal data,which has real-time effect.The results show that the proportion of vibration analysis results in the 2σinterval is more than 95%,which meets the requirements of statistical process control and can obtain reliable results.The abnormal values are all higher than 99.6%.The proportion of normal values is selected as the indicator to judge the elevator system status.The isolation forest model can be used to judge the difference between the elevator running state and the normal state.

关 键 词：异常状态 非侵入式检测 孤立森林 差异性 

分 类 号：TP12[自动化与计算机技术—控制理论与控制工程]