Abstract: The quality of process data in a chemical plantsignificantly affects the performance and benefits gained fromactivities like performance monitoring, online optimization, andcontrol. Since many chemical processes often show nonlineardynamics, techniques like extended Kalman filter (EKF) and nonlineardynamic data reconciliation (NDDR) have been developed to improvethe data quality. Recently, the recursive nonlinear dynamic datareconciliation (RNDDR) technique has been proposed, which combinesthe merits of EKF and NDDR techniques. However, the RNDDR techniquecannot handle measurements with gross errors. In this paper, asupport vector (SV) regression approach for recursive simultaneousdata reconciliation and gross error detection in nonlinear dynamicalsystems is proposed. SV regression is a compromise between theempirical risk and the model complexity, and for data reconciliationit is robust to random and gross errors. By minimizing theregularized risk instead of the maximum likelihood in the RNDDR, ourapproach could achieve not only recursive nonlinear dynamic datareconciliation but also gross error detection simultaneously. Thenonlinear dynamic system simulation results in this paper show thatthe proposed approach is robust, efficient, stable, and accurate forsimultaneous data reconciliation and gross error detection innonlinear dynamic systems within a recursive real-time estimation framework. It can also give better performance of control.