东北大学学报(自然科学版) ›› 2024, Vol. 45 ›› Issue (6): 761-768.DOI: 10.12068/j.issn.1005-3026.2024.06.001

• 信息与控制 •    

电梯门机异常检测和故障定位方法

范玉川1,2,3(), 范波3, 陈卓3, 张孝顺1,2   

  1. 1.东北大学 佛山研究生创新学院, 广东 佛山 528311
    2.东北大学 信息科学与工程学院, 辽宁 沈阳 110819
    3.广东美的暖通设备有限公司, 广东 佛山 528311
  • 收稿日期:2023-01-20 出版日期:2024-06-15 发布日期:2024-09-18
  • 通讯作者: 范玉川
  • 作者简介:范玉川(1988-),男,河南新乡人,东北大学博士后研究人员.
  • 基金资助:
    国家自然科学基金青年基金资助项目(51907112)

A Method for Anomaly Detection and Fault Diagnosis of Elevator Door Machine

Yu-chuan FAN1,2,3(), Bo FAN3, Zhuo CHEN3, Xiao-shun ZHANG1,2   

  1. 1.Foshan Graduate School of Innovation,Northeastern University,Foshan 528311,China
    2.College of Information Science and Engineering,Northeastern University,Shenyang 110819,China
    3.Guangdong Midea Heating and Ventilating Equipment Co. ,Ltd. ,Foshan 528311,China.
  • Received:2023-01-20 Online:2024-06-15 Published:2024-09-18
  • Contact: Yu-chuan FAN
  • About author:FAN Yu-chuan E-mail:fycworking@163.com

摘要:

提出一种电梯门机运行异常检测和故障定位方法.首先,从电梯门机的运行数据中分离出开关门曲线并将其划分为10个运行段,提取各段的数据特征.其次,提出基于箱线图的异常检测方法并利用各运行段的累积数据特征进行异常诊断;为防止数据不满足正态分布而造成的诊断误差,加入了门机特征数据的正态性检验方法,并对不满足正态分布的数据进行Box-Cox变换,使其满足正态分布.最后,提取电梯门机分段特征数据,采用极限学习机(ELM)对门刀卡阻故障、整体阻力增大故障和同步带松脱故障3种故障进行分类模型训练.经实验验证,提出的异常检测方法和故障定位方法准确度高,具有较高的应用推广价值.

关键词: 电梯, 异常检测, 箱线图, Box-Cox变换, 故障定位, 极限学习机

Abstract:

A method for anomaly detection and fault diagnosis of elevator door machine operation is proposed. Firstly, the opening and closing door curve is spearated from the operation data of the elevator door machine and divided it into 10 operating segments, and the data characteristics of each section is extracted. Secondly, an anomaly detection method based on boxplot is proposed, and the accumulated data characteristics of each running section are used for anomaly diagnosis. In order to prevent the diagnostic error caused by the data not satisfying the normal distribution, the normality test method of the feature data of the door machine is added, and the Box-Cox transformation is performed on the data that does not meet the normal distribution. Finally, the segmented feature data of the elevator door is extracted, and the extreme learning machine(ELM) is used to train the classification model for three faults: door knife jamming fault, overall resistance increase fault and synchronous belt loosening fault. Experiments have verified that the proposed anomaly detection method and fault diagnosis method have high accuracy and value of application and promotion.

Key words: elevator, anomaly detection, boxplot, Box-Cox transformation, fault diagnosis, ELM(extrame learning machine)

中图分类号: