摘要
实现电梯制动器力矩的精确预测对确保电梯安全运行和实现预测性维护具有重要的意义。针对曳引式电梯在制动力矩预测方面存在准确性与可靠性不足的问题,以及现有Transformer存在计算复杂度高和训练时间长的局限性,提出了一种基于改进鲸沙虫群算法优化Transformer网络(ISSA-Transformer)的电梯制动力矩预测方法。首先,为了提高Transformer的预测精度,在Transformer模型中添加了特征融合门(FFG)以提高模型的特征提取能力,使其能够更有效地捕捉制动力矩的全局与局部特征;然后,利用拉普拉斯交叉算子、混合对立学习方法以及高斯扰动对鲸沙虫群算法(SSA)进行了改进,以增强SSA的搜索能力和全局最优收敛性。并采用ISSA算法优化了Transformer的迭代次数、批次大小和学习率,以提高模型的计算效率并减少训练时间,从而建立了电梯制动器制动力矩的预测模型;最后,对曳引式电梯制动器数据进行了分析,将所得结果与LSTM、Transformer和SSA-Transformer模型进行了比较。研究结果表明:ISSA-Transformer的均方根误差(RMSE)较LSTM、Transformer和SSA-Transformer模型分别降低了0.0318、0.0144和0.0133,用于电梯制动力矩预测的准确率达到了98.7%,相较传统方法具有更高的精度和稳定性。该方法可为电梯的安全评估和预测性维护提供更可靠的技术支持。
The precise prediction of elevator brake torque was of great significance for ensuring elevator safety operation and enabling predictive maintenance.In response to the issues of insufficient accuracy and reliability in torque prediction for traction elevators,as well as the limitations of existing Transformers in terms of high computational complexity and long training time,an elevator braking torque prediction method based on improved salp swarm algorithm optimization Transformer network(ISSA-Transformer)was proposed.Firstly,to improve the prediction accuracy of the Transformer,a feature fusion gate(FFG)was added to the Transformer model to enhance its feature extraction capability,allowing global and local torque features to be more effectively captured.Then,the salp swarm algorithm(SSA)was improved using a Laplacian crossover operator,a hybrid opposition-based learning approach,and Gaussian perturbation to strengthen its search capability and global optimal convergence.Furthermore,the ISSA was used to optimize the Transformer s number of iterations,batch size,and learning rate to improve computational efficiency and reduce training time,thereby establishing a predictive model for elevator brake torque.Finally,the data of traction elevator brakes were analyzed,and the obtained results were compared with those of LSTM,Transformer,and SSA-Transformer models.The research results show that the root mean square error(RMSE)of the ISSA-Transformer is respectively reduced by 0.0318,0.0144,and 0.0133 compared to the LSTM,Transformer,and SSA-Transformer models,and achieve an accuracy of 98.7%for elevator brake torque prediction.This demonstrates higher precision and stability compared to traditional methods,thereby providing more reliable technical support for elevator safety assessment and predictive maintenance.
作者
苏万斌
江叶峰
李科
周振超
易灿灿
SU Wanbin;JIANG Yefeng;LI Ke;ZHOU Zhenchao;YI Cancan(Jiaxing Special Equipment Inspection and Research Institute,Jiaxing 314000,China;Zhejiang Academy of Special Equipment Science,Hangzhou 310020,China;School of Mechanical Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)
出处
《机电工程》
北大核心
2025年第10期2027-2036,共10页
Journal of Mechanical & Electrical Engineering
基金
浙江省自然科学基金资助项目(LTGG24E050002)
嘉兴市科技计划项目(2023AD31084)。
