In this paper a deterministic suitability measure is introduced that q
uantifies the capability of a particular model class to capture the co
ntrol relevant I/O-behavior of a nonlinear system. This suitability me
asure can be used for the purpose of model structure selection prior t
o the actual parameter identification. Important properties of this su
itability measure are discussed and an attractive approach for its num
erical computation based on convex optimization is outlined. The suita
bility measure is exemplary used to select an appropriate model struct
ure for a simple continuous stirred tank reactor.