Concepts and Algorithms for Non-Linear System Identifiability

摘要: Recently, commutative algebra and differential have come to use as mathematical tools for solving problems in automatic control. We will these answer questions regarding identifiability models given a set of polynomials. A constructive algorithm, Ritt's has been modified this specific situation. Furthermore, comparisons between algorithm Buchberger's the question when model structures are state space form, performed. The basic problem is that computational complexity rapidly increases with size. examine various ways simplify computations respect, but it must also be stressed increase inherent problem.In identification from deterministic point view an said robustly convergent if true system regained noise level tends zero. In thesis we introduce concept close performance measure; robust global identifiability. structure, i.e., smoothly parameterized models, globally identifiable there exist such parameters show implies considered structure characteristic This means any parameters, can uniquely estimated data robustness property.Finally, method estirnation residence time continuous flow systems varying dynamics discussed. By resampling, choosing instants different sampling instants, interpolation measured points, obtain fl.ow constant expressed new resampled vector. assume fiow patterns invariant. then used chosen structure. From identified model, readily calculated procedure briefly described. presented extended enable recursive identification. case improvement tracking ability compared ordinary routine.