SVD-based Identification Algorithm for Hammerstein-typed Nonlinear Systems

Hammerstein-typed nonlinear models can be used to represent nonlinear systems in the areas of chemical processes, biological processes, signal processing, etc. Firstly a novel multichannel algorithm for the identification of Hammerstein-typed nonlinear systems is presented, in which the coefficient parameters of the dynamic linear block and memoryless nonlinear block are identified by least squares estimation (LSE) combined with singular value decomposition (SVD). This identification algorithm can eliminate any needs for the mechanism or prior knowledge of the nonlinear or linear block. Furthermore, in comparison with traditional single-channel identification algorithm, this multi-channel one can increase the approximate accuracy remarkably. In addition, under weak assumptions on the persistency of excitation (PE) of the inputs, the algorithm provides consistent estimates in the presence of white output noise, moreover, its convergence can also be theoretically proved. At last, the performances of the identification algorithm are illustrated through simulations on a benchmark problem, a pH neutralization process, which validate the feasibility and superiority of these proposed algorithms