摘要
电梯制动器的制动力矩是影响电梯运行安全的关键参数,利用深度学习算法对其进行预测,能为电梯的安全使用和后期维保提供重要参考。基于门控神经网络(GRU)预测模型,结合生成对抗网络(GAN)的基本思想,以1D-CNN作为鉴别器,提高电梯制动力矩预测模型的泛化能力。利用实验数据进行训练,获得的预测结果方均根误差为1.024 4,并与常用的时间序列分析模型如GRU、LSTM等进行对比,结果表明:所提出的方法在电梯的制动力矩预测精度上具有明显的优势。
The braking torque of elevator brake is a key parameter affecting the safety of elevator operation.Deep learning algorithm is used to predict it,which can provide an important reference for the safe use and subsequent maintenance of the elevator.Based on the Gated Neural Network(GRU) prediction model,this paper combines it with the basic idea of Generative Adversarial Network(GAN),and uses 1D-CNN as the discriminator to enhance the generalization ability of the elevator braking torque prediction model.The experiment data is applied for training to abtain the prediction result with the root mean square error indicating as 1.024 4.Comparison is conducted with commonly used time series analysis models such as GRU and LSTM,and the results show that the proposed method has obvious advantages in the prediction accuracy of elevator braking torque.
作者
苏万斌
江叶峰
易灿灿
徐彪
SU Wanbin;JIANG Yefeng;YI Cancan;XU Biao(Jiaxing Special Equipment Inspection and Testing Institute,Jiaxing 314050,China;Wuhan University of Science and Technology,Wuhan 430081,China)
出处
《机械制造与自动化》
2024年第2期50-55,共6页
Machine Building & Automation
基金
国家自然科学基金项目(U1709210,51805382)
2019年浙江省省级市场监管科研计划项目(20190339)
2020年浙江省市场监管局质量技术基础建设项目(20200126)。
关键词
电梯
制动力矩
时间序列分析
生成对抗网络
门控循环神经网络
elevator
braking torque
time series analysis
generate adversarial network
gated recurrent neural networks
