摘要
针对高速电梯制动器失效率及维护决策方面的研究目前仍存在明显的不足。为了解决目前高速电梯在制动器失效率预测上存在结果准确性和可靠性不足的问题,对高速电梯制动器失效模式和机理进行了分析,确定了影响制动器失效的主要原因和相关参数,提出了一种经贝叶斯超参数优化后的预测性生成对抗网络(ForGAN)模型。首先,采集了高速电梯制动器工作性能数据,并对其进行了归一化处理;然后,利用主成分分析法进行了理论失效率计算,并采用了基于BO+ForGAN的模型对制动器失效率进行了预测和分析;最后,将所得结果与SVM、BiLSTM等传统预测模型所得结果进行了分析对比,并选取绝对误差、均方根误差、决定系数(R2)对上述各个预测结果的精度进行了评估。研究结果表明:基于BO+ForGAN模型的制动器失效率预测效果最好,泛化能力最高,能适应不同的实验工况,且贝叶斯超参数寻优算法能够找到一组最优的超参数。评估结果显示,高速电梯制动器失效率预测值的准确率达到了98.1%,从而验证了基于BO+ForGAN模型(方法)的有效性。
At present,there are still obvious deficiencies in the research on brake failure rate and maintenance decision of high-speed elevator.In order to solve the problem of insufficient accuracy and reliability of the current brake failure rate prediction of high-speed elevators,the main reasons and related parameters affecting brake failure were determined by analyzing the failure mode and mechanism of high-speed elevator brakes,and a forecasting generative adversarial networks(ForGAN)model optimized by Bayesian hyperparameters was proposed.Firstly,the performance data of the brakes were collected and normalized.Then,the theoretical failure rate was calculated using the principal component analysis method,and the Bayesian optimization-forecasting generative adversarial networks(BO+ForGAN)model was used to predict the failure rate of the brakes.Finally,the results were compared with traditional prediction models such as support vector machines(SVM)and bi-directional long short-term memory(BiLSTM).Absolute error,root mean square error and determination coefficient(R 2)was selected to evaluate the accuracy of prediction results.The research results show that the BO+ForGAN-based model has the best prediction effect,the highest generalization ability,and can adapt to different experimental conditions,and the Bayesian hyperparameter optimization algorithm can find a set of optimal hyperparameters.The evaluation results show that the accuracy of the predicted value of the brake failure rate reaches 98.1%,which verifies the effectiveness of the BO+ForGAN-based model(method).
作者
苏万斌
陈伟刚
易灿灿
陈启锐
SU Wan-bin;CHEN Wei-gang;YI Can-can;CHEN Qi-rui(Jiaxing Special Equipment Inspection and Testing Institute,Jiaxing 314050,China;School of Mechanical Automation,Wuhan University of Science and Technology,Wuhan 430081,China)
出处
《机电工程》
CAS
北大核心
2023年第4期615-624,共10页
Journal of Mechanical & Electrical Engineering
基金
国家自然科学基金资助项目(U1709210,51805382)。
关键词
预测性生成对抗网络
贝叶斯超参数优化
传统预测模型
均方根误差
泛化能力
失效率
维护决策
forecasting generative adversarial networks(ForGAN)
Bayesian hyperparameter optimization(BO)
traditional prediction model
root mean square error
generalization ability
failure rate
maintenance decision
