基于改进EM算法的混凝土泵车数据治理  

An improved expectation maximization algorithm for missing data management of concrete pump truck

作　　者：邓子畏[1] 唐朝晖[1] 朱红求[1] 赵于前[1,2] DENG Ziwei;TANG Zhaohui;ZHU Hongqiu;ZHAO Yuqian(School of Automation,Central South University,Changsha 410083,China;Hunan Engineering Research Center of High Strength Fastener Intelligent Manufacturing,Changde 415701,China)

机构地区：[1]中南大学自动化学院,湖南长沙410083 [2]湖南省高强度紧固件智能制造工程技术研究中心,湖南常德415701

出　　处：《中南大学学报（自然科学版）》2021年第2期443-449,共7页Journal of Central South University:Science and Technology

基　　金：国家工信部重点资助项目(TC19083WB)。

摘　　要：针对混凝土泵车远程监控中存在数据缺失的问题,提出一种基于随机期望最大化算法的缺失数据治理算法,通过将马尔科夫链蒙特卡洛方法(MCMC)与随机变量相结合改进期望最大化(EM)算法。首先,在期望步中,在MCMC矩阵中采样生成缺失值,并将该值代入进行随机近似模拟以更新估计值。然后,在最大化步中,通过反复迭代得到最大化估计值作为重构值来填充缺失数据。最后,以混凝土泵车实际运行数据对本文提出的算法、均值填充法和EM算法的缺失数据填充效果进行比较。研究结果表明:所提算法有效地解决了EM算法依赖初始值设定的问题,提高了数据填充的准确率。To solve the recovery problem of data missing in remote monitoring of concrete pump truck,an improved EM missing data filling model based on the stochastic process was proposed,which combines the Markov Chain Monte Carlo(MCMC)and random variables to improve the EM algorithm.Firstly,in the expectation steps,the missing values were sampled in the MCMC matrix,and were applied in the stochastic approximation to update the estimation.Secondly,the maximization steps in EM algorithm were applied iteratively to find the most possible estimated value as the reconstruction value.Finally,compared with the mean filling method and EM algorithm,the proposed method was verified by using the remote monitoring data of pump truck data.The results show that the improved EM algorithm effectively reduces the dependence on the initial setting value and improves the accuracy of data filling.

关 键 词：数据缺失 随机过程 改进EM算法 数据治理 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]