The two-model approach

According to the EICAS methodology, the control system design of a plant is carried out in two steps, starting from the requirement specifications:

For what concerns the control system design, the dynamic behaviour of the plant has to be described by a mathematical model belonging to the class of the dynamic systems.

The plant description through such mathematical model is always only an approximation of the reality, which can neither perfectly nor completely be described through a mathematical model.

Therefore, because of the unavoidable approximation the model must be oriented toward the design of the specific control system one intends to implement, with the aim of grasping those aspects of the plant dynamics which are significant in view of the design to be carried out, and the knowledge of which is essential for reaching the established requirements.

As a consequence the mathematical model to be used for describing the dynamic behaviour of the plant will depend not only on the plant itself, but also on the performance requirements of the plant to be designed.

Given a plant and the control requirement specifications it does not and it can not exist (for what will be said later) any general rule defining how to build the corresponding mathematical model and enabling (even only in theory) to build it automatically. In fact any rule should have as a prerequisite the knowledge of the plant to be modelled, therefore it could be only approximate, so in order to be useful it should be oriented toward the model objectives, that is toward the design of the specified control system. Consequently it should be necessary to start from a plant model which must – as an hypothesis – include with sufficient accuracy all information needed for the control design and for that reason gives an exhaustive description of all above mentioned aspects of interest. In other words for building the model required by the control design one should start from a model given a priori which - by hypothesis – should be fit for the control design: to prove the thesis one should assume the hypothesis that the thesis be true! Then the above general rule can not exist.

In conclusion, it is the designer who – on the grounds of his experience and knowledge of the plant, which is not necessarily transferred into a mathematical model – decides at his own risk that a given dynamic model defined by a finite set of parameters (whose values might still have to be estimated) is fit for being taken as the foundation of a control system design aiming at the pre-established performance.

After having chosen the approximate model, the control design follows the automatic control theory, taking into account the uncertainty of the plant representation through the model.

The control design in presence of uncertainty is tackled through two different methodological approaches.

The most widespread one requires that the control be set up in-field. In this approach the model is used for a first attempt of the design.

The final control design is carried out experimentally: in this way the problem of the inherent model uncertainty is overcome.

The other one, the EICAS approach, follows a methodology, which is theoretically founded and was tested over years through several projects: it implements a control system with guaranteed performance within the uncertainty range of the model.

As a consequence what is performed is a true design according to the engineering significance of the word.

If the design has been carried out in the proper way and the control system is implemented in compliance with the project, the control system must operate fulfilling the design requirements without requiring any experimental adaptation.