Functional State Modelling: How it works
Process modelling using the functional state concept is an approach
where models are built using a two-level hierarchy. In the upper
level, the process is treated like a finite state automaton, where the
states are so called functional states. The lower level models the
process locally using conventional approaches, such as state or
input--output models. The functional state models are designed using
process knowledge either using knowledge of the process structure or
following an expert system approach. In addition to the functional
state structure, the transition controls must be defined. Normally
there are no uniquely defined transition controls between functional
states, which makes a great difference between this and the
conventional automaton approach. A basic task of this approach is to
identify on-line transitions between functional states. This chapter
provides an introduction to the functional state concept, and
illustrates it with a simple `wire model', where the methods used to
detect changes in process dynamics and transitions between states are
discussed. A neural network classifier using the Wiener model and
Laguerre representation of the input signals is introduced. An example
of the use of a Kalman filter for the representation of multiple
dynamics is given. The methods are applied to a practical biotechnical
process control task.
R Murray-Smith
Last modified: Tue Mar 25 14:55:59 GMT