Performance evaluation of instrumentation sensor network design using a data reconciliation technique based on the unscented Kalman filter

Safe and reliable operation of industrial chemical plants necessitates proper design and performance of instrumentation sensor networks. In this paper, a data reconciliation technique based on the unscented Kalman filter (UKF) is proposed to extend an instrumentation sensor network design approach to non-linear dynamic processes. Moreover, an efficient performance measure based on the root mean squared error (RMSE) of the estimated variables has been presented to evaluate each candidate instrumentation sensor network design. A simulated nonlinear continuous stirred tank reactor (CSTR) benchmark plant has been utilized to illustrate the effective capabilities of the proposed approach.