A new soft sensor method for dynamic processes based on dynamic orthogonal forward regression

To cope with the issue of dynamic characteristic in industrial processes, a new soft sensor method based on dynamic orthogonal forward regression (dynamic OFR) is proposed in this paper. The proposed method applies OFR to the augmented matrix with time-delayed secondary variables. The meaningful time-delayed variables which can well explain primary variables are then selected automatically and a sparse soft sensor model is thus constructed. The simulation results on predicting butane concentration in the bottom of debutanizer column demonstrate the superiority of the proposed method in terms of prediction accuracy and the computational complexity.