June 1, 2021
Tutorial - Running an Optimization
Using the optimization dialog
Numerical optimization is an important tool for automatically finding a desired parameter set for a
model. Hopsan contains an optimization dialog from where optimizations can be controlled. This guide
shows how to optimize control parameters for a simple hydraulic position servo using the Complex-RF
algorithm. Optimization in Hopsan can either be performed from a dialog or by loading previously
saved (or manually created) script files. This tutorial demonstrates both methods.
1. Open the model
The Welcome tab has a box with example models. Open the Position Servo example model.
This model describes of a hydraulic position servo, where a PI-controller is used to control the
position. This consists of a proportional gain (Kp ) and an integrator gain (Ki ). The values of
these gains will decide the speed and stability of the servo. It is thus desirable to find as good
controller parameters as possible.
2. Simulate
Simulate the model by clicking the simulate icon.
Simulate current project (Ctrl-Shift-S)
Now double click on the “Scope” component on top of the model. Two curves will be plotted, the
reference signal and the actual position. The reference signal is a step, and the position shows
the corresponding step response. As you can see, it is quite oscillative. We want to adjust the
control parameters to reduce its rise time, overshoot and settling time by using the optimization
feature.
3. Open the dialog
Click the optimization icon to open the dialog.
Open optimization dialog (Ctrl-Shift-Z)
1
June 1, 2021
The first page should look like above. You can either write an optimization script by hand, load
a script file or start the script wizard. Let us begin with the latter. Click on the wizard icon.
Script Wizard
4. Optimization settings
The first wizard page will open. From here you can choose the general optimization settings.
• Set Algorithm to Complex-RF
• Set Number of iterations to 1000
• Set Number of points to 4
Leave all other settings unchanged for now.
2
June 1, 2021
5. Choose optimization parameters
Next page controls the parameters in the model that shall be optimized.
Check the following parameters in the list:
• GainI → k#Value
• GainP → k#Value
Then give them the following min and max values:
Parameter Min Max
GainI, k#Value 0 0.001
GainP, k#Value 0 0.04
6. Choose objective functions
The optimization algorithms use objective functions to calculate how good a certain point in
the parameter space is. From the third page this can be chosen. In this model we want the
actual position to be as close as possible to the reference signal. To do this, check the following
variables in the list:
• Position_Sensor → out → Value
• Step → out → Value
Then choose “Minimize” and “Absolute Average Difference” in the boxes. Click “Add Function”.
3
June 1, 2021
Apart from minimizing the average difference, we also want to minimize the overshoot of the
step response.
• Uncheck Step → out → Value
Now choose “Minimize” and “Overshoot”, and then click “Add Function”. In the box called
“arg1”, change the value to 0.7.
Now we have added a second objective function. The total objective function will the sum of
these. As they may not be equally important or equally scaled, it is possible to weight the
individual functions. Change “Weight” for the first function to 10. Now the page should look
like this:
4
June 1, 2021
7. Start the optimization
Next page shows the generated script code. Modifications can be made if necessary, but leave it
as it is for now.
Now change to the "run" tab. From here the execution of the optimization can be monitored:
Click on “Start Optimization”.
Start Optimization
During the optimization the objective function values and the values of the optimization param-
eters are shown for the different points. The rows are sorted from best to worst.
5
June 1, 2021
8. Evaluate the results
Once the optimization has finished, we can evaluate the results. Parameters from a point can be
written back to the original model by using the “Apply” buttons.
Click the “Apply” button for the point with green text (the best point). Then go back to the
original model and click the Simulate button again.
Simulate current project (Ctrl-Shift-S)
Now plot the step response by double clicking the scope component on top of the model. If
a desirable result has not been found, it can be necessary to change objective functions or
parameters and re-run the optimization.
6
June 1, 2021
Optimizing from script files
It is often necessary to re-use the settings from the optimization more than once. This can be done by
saving the optimization scripts to a file, and loading them again later.
1. Follow steps 1-7 above
2. Saving a script file
Click on Save script file in the dialog.
3. Loading a script file
To load a previously saved script file, click on Load script file in the dialog. You will be taken
directly to the code tab. Go to the "run" tab and start the optimization as before. Note that
the settings on the first three pages are not updated by the script. Running the wizard again will
re-generate the script and overwrite the loaded code.
4. Loading a script file from command terminal
It is also possible to load a script file from the command terminal. Just write “exec” followed
by the path to the file. Hopsan will automatically understand that this is an optimization script
and start the dialog. You can use “pwd” to show current directory and “cd” to change directory
if necessary. Example: >> exec ..\folder\[Link]