Conclusion - Advice
Modélisation et identification d’un processus sidérurgique

Identification is not just a matter of using a "miracle" algorithm on data collected by applying any test protocol (pseudo-random binary signals). Representing a real physical process using a mathematical operator is always a hazardous operation (non-linearities, non-stationarity, disturbance, noise, variability of operating conditions, etc.).

Remember that all models are "wrong", but that doesn't mean that some can't be "useful", if we know how to adapt their establishment conditions to their subsequent use.

Identification must proceed through a series of steps, of equal importance in terms of their potential harm, if not carried out correctly. The most critical part, however, both methodologically and operationally, is clearly in the optimal choice of test protocol, which will ask the process the relevant questions. If there is no information...