Test for global identifiability for nonlinear model structures

In the present work an algorithm for testing global identifiability for a wide class of nonlinear model structures with bounded signals is presented. The objective functional applied for parameter estimation is assumed to be nonconvex and analytic in a broad region of the parameter space. The supporting theory is given in lemmas which concern the existence and uniqueness of straight slots of nonconvex functionals and the convergence of polynomial mu-approaches. The method is exhaustive and hence the detection of the global minimum (if there is anyone) is practically sure. It is shown that the algorithm can be implemented with a few amount of system information.