主成分分析和反向传播神经网络模型在血液透析机预防维护中的应用  

Application of PCA and BP neural network model in the preventive maintenance of hemodialysis machine

作　　者：敬微微 韩倩 吴昊[1] 林建阳[1] JING Wei-wei;HAN Qian;WU Hao(Department of Asset Management,The First Hospital of China Medical University,Shenyang 110001,China.)

机构地区：[1]中国医科大学附属第一医院资产管理部,辽宁沈阳110001

出　　处：《中国医学装备》2020年第7期137-140,共4页China Medical Equipment

摘　　要：目的:应用主成分分析和反向传播(BP)神经网络组合模型,分析影响血液透析机维修成本因素,控制维修成本。方法:对影响血液透析机维修费用的设备型号、购置单价、设备使用年限、配件类型、配件单价和配件数量6大因素进行主成分分析,根据特征根值和累计贡献率进行主成分提取,将入选主成分作为输入变量,维修费用为输出变量,构建3层结构的BP神经网络模型,分析影响维修费用的主要因素。结果:主成分分析选取的4个主成分包含全部变量87.708%的信息,建立的BP神经网络模型预测值与设备维修费用呈正相关性。灵敏度由高到低依次为第4主成分(38.5%)、第1主成分(36.0%)、第2主成分(13.4%)和第3主成分(12.1%);维修成本贡献由大到小主要因素依次为配件单价、设备型号和配件类型,其次为配件数量、使用年限和购置单价。结论:基于主成分分析的BP神经网络模型能较好的预测血液透析机维修成本,做好特定型号透析机的预防性维护保养能减少重大故障发生,降低设备维修成本。Objective:To analyze the factors that affected the maintenance cost of hemodialysis machine and control the maintenance cost by using the combination model of principal component analysis(PCA)and back propagation(BP)neural network.Methods:The PCA was applied to analyze the six factors that affected the maintenance cost of hemodialysis machine included equipment model,unit price of purchase,service life of equipment,type of accessories,unit price of accessories and number of accessories.And the principal components extraction was implemented based on the values of characteristic root and cumulative contribution rate.The selected principal components were used as input variables and the maintenance costs were used as output variables to construct BP neural network model with three-layer structure,and to analyze the main factors that affected the maintenance costs.Results:The four principal components which were selected by PCA contained 87.708%of all variable information,and the predicted value of the established BP neural network model was positively correlated with the maintenance cost of equipment.The sensitivities of the four principal components were ranked from high to low as the fourth principal component(38.5%),the first principal component(36%),the second principal component(13.4%)and the third principal component(12.1%).The main factors contributing to the maintenance cost from large to small were the price of accessories,equipment model and type of accessories,followed by the number of accessories,service life and unit price of purchase.Conclusion:The BP neural network model based on PCA can better predict the maintenance cost of hemodialysis machines,and the realized preventive maintenance of certain types of hemodialysis machine can reduce the occurrence of major failures and decrease the maintenance cost of equipment.

关 键 词：血液透析机 维修成本控制 主成分分析(PCA) BP神经网络 预防性维护 

分 类 号：R197.39[医药卫生—卫生事业管理]