25.12.2012 Views

Design and Implementation of Object-Oriented ... - Automatic Control

Design and Implementation of Object-Oriented ... - Automatic Control

Design and Implementation of Object-Oriented ... - Automatic Control

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Chapter 2. Modeling Techniques<br />

• Send the modeler back to the step 1 to resolve the problem with<br />

pen <strong>and</strong> paper, reformulating the model into an equivalent index 1<br />

problem.<br />

• Build the capacity <strong>of</strong> recognizing <strong>and</strong> resolving high index problems<br />

into the modeling tool.<br />

Because Modelica is designed as a multi-domain modeling tool, it has to<br />

support automatic index h<strong>and</strong>ling. This does not mean that automatic<br />

index reduction is the silver bullet that solves all high index problems in<br />

the best possible way. There are a few exceptions when manual derivation<br />

<strong>of</strong> an index 1 problem is preferable to the automatic procedure. This<br />

is the case when the automatic procedure gives results with numerical<br />

disadvantages, at least with the current version <strong>of</strong> the Modelica language<br />

<strong>and</strong> the index reduction algorithms in the Dymola tool. Examples <strong>of</strong> these<br />

rare cases are presented in Chapter 4. But independent <strong>of</strong> an automatic<br />

h<strong>and</strong>ling, model users have to underst<strong>and</strong> the implications <strong>of</strong> high index<br />

problems in order to provide the right initial conditions.<br />

Partial Differential Equations<br />

Partial differential equations (PDE) are the mathematical formulations<br />

that permits the highest level <strong>of</strong> detail for system level models. They<br />

provide the tool corresponding to the magnifying glass with the highest<br />

magnification power. PDEs form an essential part <strong>of</strong> the mathematical<br />

description <strong>of</strong> physical systems depending on space <strong>and</strong> time. They arise<br />

in a large number <strong>of</strong> modeling applications, but are not that common in<br />

systems level modeling. The reason for this is that PDE can have widely<br />

differing properties <strong>and</strong> that they are much more difficult to deal with<br />

numerically than ODEs. Using PDE for dynamical systems means that<br />

we have at least one more independent variable apart from time. Usually<br />

these are spatial coordinates, but variables can also be distributed in other<br />

dimensions like particle sizes in crystallization processes.<br />

Numerical solutions to PDE for design calculations are used in almost<br />

all engineering disciplines <strong>and</strong> are established as independent research<br />

areas, e. g., Computational Fluid Dynamics (CFD) <strong>and</strong> Finite Element<br />

Methods (FEM).<br />

Consider a set <strong>of</strong> PDEs expressed over the solution domain Ω which is<br />

a compact region in IR m+1 <strong>and</strong> its boundary δ Ω ∈ IR m , compare Figure 2.3<br />

for a two-dimensional example. If the independent variables are separated<br />

into time t <strong>and</strong> a vector <strong>of</strong> spatial variables x, then the dependent variables<br />

u can be written as<br />

u = u(x, t)<br />

where u ∈ IR n , x ∈ IR m <strong>and</strong> t ∈ IR ≥ 0. In mathematical texts the independent<br />

space <strong>and</strong> time variables are usually treated alike, here Ω is used<br />

30

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!