And if the system is stiff or highly non-linear, is there a general (or nearly general) rule-of-thumb to verify the results?
A proposition for evaluating the accuracy of your simulation :
1/ make a high resolution simulation with high accuracy and keep the trajectory as the 'true' trajectory
2/ change the resolution and evaluate the error compared to the 'true' one.
3/ make the same as in 2/ with different methods
4/ Plot the accuracy versus speed to select the best method.
http://www-rohan.sdsu.edu/doc/matlab/toolbox/simulink/ug/simulation5.html
the first the assumptions for your system minimum
second the incremented of the step in simulation
you can converter the data of nonlinear to look-up table to near the data
The accuracy is determined by the nature of inputs and the assumptions used.It also depends the use of custom made models
You can also check the Refine Factor on the preferences of your simulink model. Default is 1, it's better to make it 4. Check this out if you would like
http://www.mathworks.com/help/simulink/gui/data-import-export-pane.html?refresh=true#bq9_f96-1
and this
http://blogs.mathworks.com/seth/2009/07/14/refining-the-output-of-a-simulation/
there are no factors specified but it depends on your work what the no. of input hidden and output components
cannot be said very clearly..but various factors which contribute may include
1) The solver which you are using.
2) The step size of simulation.
3) Refine factor.
The factors asked are a method used and simulation parameters such as initial step size and tolerance.
At use the method (solver) and/or simulation parameters could not provide convergence of solution (i.e. stability) which cases may take place especially for stiff differential equations simulation accuracy is lost. It may be got at right option of simulation method and parameters.
- Badges
- Science topic