摘要
针对矿用压风机这类分布式系统的异常类别复杂、识别精度低等问题,提出了一种基于深度置信网络(DBN)和最小二乘支持向量机(LSSVM)的异常状态识别方法。首先,分析压风机组成系统及其运行机理,确定常见的异常状态类型;其次,采用DBN无监督学习方式充分挖掘监测数据中异常特征并快速提取;然后,利用秃鹰搜索算法(BES)优化LSSVM的超参数,构建最优的BES-LSSVM分类模型;最后,将DBN提取的异常特征作为BES-LSSVM模型的输入,对矿用压风机异常状态进行识别。试验验证与对比分析结果表明,相较于GA,PSO,GWO算法,BES算法的求解精度和收敛速度均有所提高,同时DBN-BES-LSSVM模型在测试集上平均识别精度达到94.65%,较PCA-LSSVM模型、DBN模型和DBN-LSSVM模型的识别精度分别提高了10.53%,5.84%和3.76%,验证了DBN-BES-LSSVM模型在矿用压风机异常特征提取以及特征识别方面的优越性。
For the problems of complex categories of abnormality and low recognition accuracy of distributed systems such as mining air compressors,an abnormal state recognition method based on deep belief network(DBN)and least squares support vector machine(LSSVM)was proposed.Firstly,the composition system of the air compressor and its operation mechanism were analyzed to determine the types of common abnormal states.Secondly,DBN unsupervised learning was used to fully mine the abnormal features in the monitoring data and quickly extract them.Then,the bald eagle search(BES)was used to optimize the hyperparameters of LSSVM to construct the optimal BES-LSSVM classification model.Finally,the abnormal features extracted by DBN were used as inputs to the BES-LSSVM model to identify the abnormal status of mining air compressor.The experimental verification and comparative analysis results show that compared to GA,PSO and GWO algorithms,the BES algorithm has improved solution accuracy and convergence speed.At the same time,the DBN-BES-LSSSVM model has an average recognition accuracy of 94.65%on the test set,which is 10.53%,5.84%and 3.76%higher than the PCA-LSSVM model,DBN model,and DBN-LSSVM model,respectively,which verifies the superiority of the DBN-BES-LSSVM model in extracting abnormal features and feature recognition of mining air compressor.
作者
李敬兆
王克定
王国锋
郑鑫
石晴
LI Jingzhao;WANG Keding;WANG Guofeng;ZHENG Xin;SHI Qing(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232001 China;Huainan Mining Group,Huainan 232001,China;Huaibei Hezhong Mechanical Equipment Co.,Ltd.,Huaibei 235000,China)
出处
《流体机械》
CSCD
北大核心
2024年第3期89-97,共9页
Fluid Machinery
基金
国家自然科学基金项目(51874010,61170060)
淮北市重大科技专项(Z2020004)
淮南市科技计划项目(2021A243)
物联网关键技术研究创新团队(201950ZX003)。
关键词
矿用压风机
深度置信网络
秃鹰搜索算法
最小二乘支持向量机
异常识别
mining air compressor
deep belief network
bald eagle search algorithm
least squares support vector machine
exception recognition
