An Inverse-Based Methodology for Disturbance Identification of Nonlinear MIMO Systems

Disturbance identification for chemical industrial processes is very important but challenging due to the process complexity. The disturbance can be recovered from the outputs by the system inverse constructed based on the linear time-invariant (LTI) state-space model of the real process. Existing system inverse algorithms are complicated and require output differentiation and a feed-though D matrix in the state space representation. This paper proposes an inverse-model-based methodology based on newly developed inverse algorithms that can detect the disturbances of nonlinear multivariable dynamic systems with high accuracy. Using the developed algorithms, system inverse can be obtained even if D matrix is absent. The methodology mainly involves process synthesis, linearization around the operating point, variable scaling, model reduction, and system inverse. A case study of a typical distillation column demonstrates the efficacy of the developed methodology.