Analyzing Numeric Integration

You can analyze the numeric integration to inspect state variables that might slow down the simulation of large models.

Note: You cannot use analyzing numeric integration when you use any of the following integrators:
  • Euler
  • Any Rkfix integrator
  • Cvode
  • Any integrator used for inline integration

Before you begin: To have a model to work with, do the following:
  • Create a behavior experience from the Dymola model example "Coupled Clutches". See Creating a Behavior Experience from a Model.
    Note: To select the model example Coupled Clutches in the first step, replace step 1 in the reference with the following: In the Dymola Behavior Modeling app, from the Behavior Tools section of the action bar, display the model example flyout by clicking Model Examples , then select Coupled Clutches .
  • Display the Simulation Setup dialog box, see Setting up a Simulation. In the Debug tab, activate the option Which states that dominate error.
  • Simulate the model. See Simulating and Generating Results.
    Note: The simulation must be finished before analyzing numeric integration.
  1. From the Scenario section of the action bar, click Simulation Analysis .

    Tip: You can also, from the Variable Browser, do any of:
    • Under the header, to the right, click Simulation Analysis .
    • Right-click any signal and select Simulation Analysis .
    For information on how to display the Variable Browser, see Simulating and Generating Results.

  2. Click the Numeric Integration tab.
    The Numeric Integration tab contains the following columns:
    • The state variable name.
    • The number of times the variable limits the step size of the numerical integrator. Counts greater than zero are highlighted.
    • The number of times the variable dominates the error, that is, has an integration error larger than any other state variable. Counts greater than zero are highlighted.
    • The number of times the variable contributes to more than 10 % of the integrator error.
    Notes:
    • You can click any column header to sort the lines in descending or ascending order.
    • The window can be minimized. It is then displayed with the icon on the right side of the screen.
    • If no information can be displayed, the Numeric Integration tab is not displayed.
  3. To display the result for a variable in a plot, right-click the variable in the window and select Plot Error.
    The following is the result:
    • In the plot that appears a marker is plotted at each time point where such an event occurs, and the value on the vertical axis is the contribution to the numeric error in percent. This plot gives a better understanding of the variable´s effect on the numeric integrator at various stages of the simulation.

      To see an example, right-click J1.w in the window and select Plot Error. The result is:

      You can see that the variable J1.w dominates the error around t=0.8, but not at t=0.9-1.0 although the error is still larger than 60 %. The variable J1.w limits the step size near t=0.65 and T=1.15.

    • The variable that you selected is also highlighted in the Variable Browser.
  4. To study the behavior of the numeric integrator, click Plot Integrator Step Size.
    A plot appears, displaying the internal integrator step size, the scaled local error estimate from the integrator, and the integration order.
    Notes:
    • Rejected steps (and subsequent restart with smaller step) are indicated with red cross.
    • You can use the scaled local error estimate to, for example, see if the integration step size is being limited by the error estimate or something else, like frequent restarts because of events.
    .