Novel adaptive nonlinear dynamic data reconciliation and gross error detection method

Data reconciliation is a well-known method in online process control engineering aimed at estimating the true values of corrupted measurements under constraints. Most nonlinear dynamic data reconciliation methods have studied cases where the input variables are constant over relatively long periods of time separated by simple step changes (e.g., set-point changes). While this scenario is not uncommon in process control, it imposes strong limitations on a method's applicability. In this paper a novel adaptive nonlinear dynamic data reconciliation algorithm is presented that extends the method presented by Laylabadi and Taylor [1] to the cases where the input variables are ramps or slow sinusoidal functions or, for that matter, any slow, smooth variation.