A BIC-based refined piecewise-linear Hammerstein identification for valve stiction pilot plant

Complicated friction characteristics of control valves, such as the asymmetric valve property, restrict the applicability of some existing data-driven models for efficiently replicating the valve stiction phenomena in practice. The authors have previously proposed a modelling strategy for asymmetric valve stiction. The strategy is based on a piecewise linear Hammerstein (PWL-HMM) identification approach with BIC-based knot refinement for simplifying the model. In this work, we validate the practical applicability and significance of this technique on data from a laboratory pilot plant. Description and illustration are outlined for the pilot plant setup. The control valve is modified such that various friction levels can be introduced to the valve stem to simulate different features of the asymmetric stiction property. The results suggest that the proposed algorithm seems to be a valid technique for identifying asymmetric behaviours of sticky valves.