基于物元信息熵的生产线健康度评估及预测  

Evaluation and prediction of production line health index based on matter element information entropy

作　　者：牛国成[1,2] 胡贞[1] 胡冬梅[2] NIU Guocheng;HU Zhen;HU Dongmei(College of Electronic Information Engineering, Changchun University of Science and Technology, Changchun 130022, China;School of Electrical and Information Engineering, Beihua University, Jilin 132021, China)

机构地区：[1]长春理工大学电子信息工程学院,吉林长春130022 [2]北华大学电气信息工程学院,吉林吉林132021

出　　处：《计算机集成制造系统》2019年第7期1639-1646,共8页Computer Integrated Manufacturing Systems

基　　金：国家973计划资助项目(613225);国家自然科学基金资助项目(91338116);吉林省教育厅“十三五”科学技术资助项目(JJKH20180338KJ)~~

摘　　要：为解决复杂生产线健康度难以评估、量化和预测的问题,提出一种将物元信息熵和支持向量机相结合的生产线健康度评估及预测方法。在评估方法中,设计了由设备运行状态、能源消耗、生产速率和生产效率为一体的立体交叉复合物元。采用层次分析法确定评价指标的理论权重,熵值法确定评价指标的客观权重,最终权重为理论权重和客观权重的联合权重。运用复合物元关联熵计算生产线的健康度;采用网格搜索法、遗传算法和粒子群对支持向量机进行参数寻优,对生产线历史健康度进行机器学习,建立健康度预测模型。实验证明,基于高斯核函数运用粒子群进行参数优化的支持向量机对健康度的预测效果突出。本文所提方法为生产决策提供数据支持和理论依据。To evaluate and predict the health index of product line, the physical element information entropy and Support Vector Machine (SVM) were proposed. The interchange complex matter element which composed of running state, energy consumption, production rate and production efficiency was designed. The theoretical weights were determined with Analytic Hierarchy Process (AHP), the objective weights were determined with entropy method, and the joint weights were combined theoretical weights and objective weights. The health index of production line was calculated by using the compound matter element correlation entropy. SVM was optimized by grid search, genetic algorithm and Particle Swam Optimization (PSO) to predict the health index. The result shows that SVM based on PSO had prominent prediction effect on the health index. It provided data support and theoretical basis for the production of Beer filling production line.

关 键 词：生产线 健康度 层次分析法 物元信息熵 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TS26[自动化与计算机技术—计算机科学与技术]