Download Practical Grey-box Process Identification: Theory and by Torsten P. Bohlin PDF

Show description

The following are some consequences of the construction: : A component is not basically connected to another component, but to a parameter or constant. This means that if one would draw a block diagram of the system, then i) some of its output signals may connect to targets in several other components, if they share parameters, ii) a component may output several signals to the same or different components, iii) a component may receive signals from several source components. Hence, even if the model class in some respects grows like a tree, a block−diagram representation of it will not be a tree.

But bond graphs do not look like ordinary process engineer’s graphs either. It is also the reason for defining object−oriented modelling languages like ModelicaX (Tiller, 2001). In conclusion, a process engineer is used to seeing the upper graph. A control engineer is used to the lower graph. The software (IdKit) is based on functions implementing the lower graph. In addition, the three different reasons for modularizing a model, namely to allow convenient descriptions of process units, refinement or units, and control of units suggests several types of components.

31. 5. 35 instead of G(x), since this makes a better approximation, should the estimate ^ x drift far from the reference trajectory x. On the other hand, this will make the EKF more susceptible to large disturbances, and will thus increase the risk for instability. 36 are stable as long as the model is, while a negative value of ^ x1 g + ^ x 2 Ã, for instance caused by a spurious value in y(τ), would cause a run−time error in the evaluation of G(x^). Notice that most matrices that are needed to handle the consequences of having an unknown input depend on τ, which means that the calculations generally take much more time than with a white box.