摘要
针对空分设备运行过程中产生的数据噪声大、特征提取困难等问题,采用深度信念网络(Deep Belief Network,DBN)方法实现空分设备的故障诊断。该方法采用对比散度(Contrastive Divergence,CD)方法和误差反向传播(Back Propagation,BP)算法分别进行模型的预训练和调优,使得参数θ达到全局最优,最后经Softmax分类器进行故障模式识别。将上述方法在某空分设备上进行实验,实验结果表明,该方法在空分设备故障诊断方面的准确性优于BP神经网络,对空分设备的维修、维护等工作具有很好的指导作用。
Aiming at the problems of large data noise and difficult feature extraction during the operation of air separation equipment,the Deep Belief Network(DBN)method is used to solve the fault diagnosis of air separation equipment.The method uses the Contrastive Divergence(CD)method and the Back Propagation(BP)algorithm to pre-train and tune the model respectively,so that the parameterθreaches the global optimum,and finally the fault is distinguished by the Softmax classifier.The method is carried out on an air separation plant.The simulation results show that the method is better than BP neural network in the fault diagnosis of air separation equipment,and has a good guiding effect on the maintenance of the equipment.
作者
钟永彦
吴亚
陈娟
邱爱兵
ZHONG Yong-yan;WU Ya;CHEN Juan;QIU Ai-bin(School of Electric Engineering,Nantong University,Nantong Jiangsu 226019,China)
出处
《计算机仿真》
北大核心
2020年第7期493-497,共5页
Computer Simulation
基金
国家自然科学基金(61473159)。