Perform a parameter optimization
What is a parameter optimization?
Parameter optimization is used to identify optimal settings for the inputs that you can control. Workspace searches a range of values for each input to find settings that meet the defined objective and lead to better performance of the system. After a simulation analysis, you can perform a parameter optimization or a sensitivity analysis, However, for engineering applications, you typically perform a parameter optimization before a sensitivity analysis because changing the system settings is often easier than changing the variability of the inputs. For example, adjusting the temperature setting is easier than reducing the variability of the temperature.
Consider the construction project scenario introduced in Monte Carlo simulation. The initial simulation answers the business question "What percent of projects will take more than 30 business days?". To explore ways to reduce the percent of projects that extend past the upper limit of 30 business days, you can use parameter optimization.
Perform a parameter optimization
When you perform a parameter optimization, Workspace searches for alternative input settings that optimize an output based on the objective and the search range you define. For the construction project scenario, you want to find optimal settings for the inputs that reduce the percent of projects that extend past 30 days (% out of spec).
-
Choose
.
- Define your objective by choosing the goal and the metric that you would like to affect for one of the outputs in your model. In the construction project example, you want to minimize the % out of spec (the percent of projects that extend past 30 business days) for Project Time.
-
To define the search range, enter the low and high settings of the
parameter values for the inputs that you can control. For inputs that you
cannot control, check
Noise.
Workspace
reruns the simulation and searches within these ranges to find settings that
improve the results based on your objective.
Tip
Consider using search ranges that are as wide as possible to broaden the search area and increase your chances of meeting your objective. Do not exceed levels that are unfeasible or unsafe for your system. You can repeat the parameter optimization and see how changing the search range affects the estimates of performance.
- Click Optimize Parameters.
Examine the results
Workspace displays the results of the parameter optimization, assumptions, and guidance for next steps.
-
View the results in the workspace.
The initial results for the construction projects example
indicated that around 15% of the projects will extend past 30 days. Following
the parameter optimization, you can now expect around 8% of the projects to
extend past 30 days.
Note
Each time you repeat the simulation, the results will vary because the simulation is based on randomly selected values for the inputs.
- Under Assumptions, compare the new settings to the previous settings and confirm that the new settings are feasible for your application.
-
You can do any of the following tasks:
- To view DPMO (Defects Per Million Opportunities), observed performance, summary statistics, percentiles, and model assumptions, click More Results.
- To view
information about next steps, hold the mouse pointer on
.
- (Optional) Perform another parameter optimization with wider ranges.
- Continue to optimize parameters until you are satisfied with the results. Next, you can perform a sensitivity analysis to examine the effects of changes to the input variation. For more information, go to Perform a sensitivity analysis.
Switch the view
Edit the model
After you analyze the results, you may want to return to the model and change inputs or outputs, and then rerun it. This allows you to test a number of "what if" scenarios allowing you to gain insight into the behavior of your system and make better decisions.
-
To change the model, choose
.
By default,
Workspace
overwrites the existing values in the current model. To preserve the current
settings and work in a copy instead, choose
.
- Choose .