Dynamic Modeling of Gross Errors via Probabilistic Slow Feature Analysis Applied to a Mining Slurry Preparation Process*

Dynamic data reconciliation and gross error detection ask for an accurate physical model, e.g. a state-space model, based on which measurement noise and gross errors can be quantitatively assessed. The model can be established based on either first-principle knowledge or process operation data. This work considers a case with limited first-principle knowledge and imperfect operation data, which is inspired by a real industrial process. We seek to develop a dynamic model using operation data contaminated by not only measurement noise but also gross errors, which conforms to known static constraints such as mass balance. Probabilistic slow feature analysis (PSFA) is adopted to describe dynamics of both nominal variations and gross errors, and model parameters are estimated by means of the expectation maximization (EM) algorithm. Data from an industrial slurry preparation process are used to demonstrate the usefulness of the proposed method.