基于双权M-估计子的鲁棒数据校正新方法  

New Method for Robust Data Reconciliation Based on Biweight M-Estimator

作　　者：金思毅[1] 刘猛[1] 李宁[1] 陶少辉[1] 

机构地区：[1]青岛科技大学化工学院,山东青岛266042

出　　处：《高校化学工程学报》2010年第4期676-680,共5页Journal of Chemical Engineering of Chinese Universities

基　　金：重质油国家重点实验室开放课题基金资助项目(2008-08)

摘　　要：准确可靠的测量数据是实现装置过程控制、模拟、优化和生产管理的前提条件,而通过仪表测量获取的过程数据中可能存在过失误差,直接影响数据校正的准确性,现有的数据校正方法不能完全有效地避免过失误差的影响。根据双权M-估计的原理,今以相对残差为变量构造了一种新型的具有强鲁棒性的目标函数,使含有过失误差的变量对函数的贡献为一常数,从而避免了过失误差对数据校正过程的影响。选取了具有代表性的一个线性问题和一个非线性问题进行实例研究,并与现有的Huber法和Cauchy法进行了对比分析。计算结果表明,对线性系统和非线性系统,新方法的过失误差侦破性能均优于Huber法和Cauchy法,且其稳定性更高。因此,在进行数据校正时应首选新方法。Reliable and accurate process measurements are crucial for the control,simulation and management of the process.However,gross errors existing in the measured process data can severely bias the reconciled data.The existing robust methods are not very efficient to eliminate the influence of gross errors completely.Based on the mechanism of the Biweight M-Estimator,a new target function with strong robustness was proposed,which nullifies the effect of gross errors on data reconciliation by limiting the function values of variables with gross error to a constant.The performance of the proposed method and its comparison with Huber＇s method and the method based on the Cauchy distribution were illustrated through a linear problem and a nonlinear one.The new method gives promising results for data reconciliation and gross errors detection with more stable property.Therefore,the proposed method should be chosen with first priority for data reconciliation.

关 键 词：数据校正 过失误差侦破 双权M-估计子 鲁棒性 

分 类 号：TQ015.9[化学工程] TP274[自动化与计算机技术—检测技术与自动化装置]