作者：姚懋欣1, 刘桂雄1, 梁敏健2
作者单位：1. 华南理工大学机械与汽车工程学院，广东 广州 510640;
2. 广东省特种设备检测研究院珠海检测院，广东 珠海 519002
关键词：电梯液压缓冲器;深度学习;长短时记忆网络;隐患识别
摘要：
作为电梯安全最后保护装置的电梯液压缓冲器，其质量检测非常重要，目前在位现场检测采用人工检测方法，准确度低，存在人身危险。该文提出一种电梯液压缓冲器隐患识别方法总体框架，重点研究基于长短时记忆(long short term memory, LSTM)网络的关键点识别、缓冲器压缩复位特征提取与模式识别技术，以及训练LSTM运动状态识别网络方法。最后，在昱奥GeN2乘客电梯及底坑安装的液压式缓冲器上搭建检测系统进行测试。结果表明：该方法对电梯液压缓冲器常见隐患识别率可达100%，有助于提高检测效率、准确度。
Elevator hydraulic buffers hidden hazard recognition method based on LSTM feature extraction
YAO Maoxin1, LIU Guixiong1, LIANG Minjian2
1. School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510640, China;
2. Guangdong Institute of Special Equipment Inspection and Research Zhuhai Branch, Zhuhai 519002, China
Abstract: The quality inspection of elevator buffers, which are the last protection device for elevator safety, is very important. Currently, manual inspection methods are used for in-situ inspection, which have low accuracy and personal danger. This paper proposes an overall architecture of an elevator hydraulic buffer hidden danger identification method, focusing on LSTM-based key point identification, buffer compression and reset feature extraction and pattern recognition techniques, and training LSTM operation state identification network methods. Finally, the detection system is built and tested on ZHYO GeN2 passenger elevators and hydraulic buffers installed in the pit, showing that the method can identify up to 100% of the common hidden hazards of elevator hydraulic buffers, which helps improve the efficiency and accuracy of detection.
Keywords: elevator hydraulic buffers;deep learning;long short term memory;hidden hazard recognition
2022, 48(9):36-40 收稿日期: 2021-08-04;收到修改稿日期: 2021-09-21
基金项目: 广东省市场监督管理局科技项目(2020CT03)
作者简介: 姚懋欣（1998-），男，广东汕头市人，硕士研究生，专业方向为智能化检测与仪器
参考文献
[1] 杨峰, 汤维峰, 虞杰. 检验液压缓冲器时常见缺陷的分析及对策[J]. 中国特种设备安全, 2017, 33(8): 78-80
[2] LI R T, CHEN G F, LIANG P. Modeling and finite element analysis of the hydraulic buffer in elevator systems[J]. Journal of Civil Engineering Research, 2017, 7(2): 63-66
[3] 周成, 付帅, 孟祥. 电梯液压缓冲器缓冲特性研究[J]. 机床与液压, 2018, 46(13): 148-151
[4] REN F C, JIANG T, LIANG X, et al. Experimental study on aging performance of elevator polyurethane buffer[C]//E3S Web of Conferences. EDP Sciences, 2021.
[5] 程大钊. 基于ARM9的电梯液压缓冲器复位时间测试仪研究[D]. 合肥: 中国科学技术大学, 2009.
[6] 刘光柱, 关胜晓, 邵锦锦. 电梯液压缓冲器复位时间测试仪的传感器电路设计[J]. 电子技术, 2010, 37(2): 29-30
[7] 曾健生. 在用电梯液压缓冲器动态测试系统的设计[D]. 天津: 天津大学, 2016.
[8] 董小龙. 电梯液压缓冲器复位时间测试仪校准方法的研究[J]. 质量技术监督研究, 2020, 4(6): 5-7
[9] FENG S C, WANG C, XUE J A. Research on elevator hydraulic buffer detection system based on intelligent image recognition[C]//Journal of Physics: Conference Series, 2020.
[10] SCHMIDHUBER J, HOCHREITER S. Long short-term memory[J]. Neural Computer, 1997, 9(8): 1735-1780
[11] WILLIAMS R J, ZIPSER D. Experimental analysis of the real-time recurrent learning algorithm[J]. Connection Science, 1989, 1(1): 87-111
[12] CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3): 273-297
