中文摘要：

鉴于工业过程的时变特性以及现场采集的数据通常具有非线性特性且包含离群点,利用最小二乘支持向量机回归（least squares support vector regression,LSSVR）建模易受离群点的影响.针对这一问题,结合鲁棒学习算法（robust learning algorithm,RLA）,本文提出了一种在线鲁棒最小二乘支持向量机回归建模方法.该方法首先利用LSSVR模型对过程输出进行预测,与真实输出相比较得到预测误差;然后利用RLA方法训练LSSVR模型的权值,建立鲁棒LSSVR模型;最后应用增量学习方法在线更新鲁棒LSSVR模型,从而得到在线鲁棒LSSVR模型.仿真研究验证了所提方法的有效性.

英文摘要：

Industrial processes possess time-varying feature,and data from industrial field usually possess nonlinear feature and contain outliers.Modeling with least-squares-support-vector regression（LSSVR） method may suffer from these outliers.To deal with this problem,we develop an online robust LSSVR method by combining with the robust learning algorithm（RLA）.The LSSVR model is used to predict process outputs,and the residuals are formed from real outputs and predicted outputs.The RLA trains the weights of LSSVR model iteratively.The trained robust LSSVR model is then updated by means of incremental updating algorithm.An online robust LSSVR model is also developed.Simulation results show the effectiveness of the proposed approach.