I am doing MOGA optimization using ANSYS workbench. I need to set an "equality constraint" to the optimization problem. Unfortunately all the offered operators are "equality constraints" only (i.e "greater than/equal to" and "less than/equal to")
Did you face such problem ? would you tell me how to manage such constraint in ANSYS ?
In ANSYS Workbench, you can apply equality constraints in the optimizer using the DesignXplorer tool. DesignXplorer is the module within ANSYS Workbench that handles design exploration and optimization studies. To apply an equality constraint, you would typically use Response Surface Optimization (RSO) or Direct Optimization methods. Here's a general step-by-step guide on how to apply an equality constraint using RSO:
1. **Setting Up Your Model:**
- Import or create your simulation model in ANSYS Mechanical, Fluent, or any other relevant module.
- Define the parameters (design variables) that you want to optimize and the responses (objectives and constraints).
2. **Create an Optimization Study:**
- In ANSYS Workbench, go to the "Engineering Data" tab and create a new Optimization Study or Add a Design Point if you already have one.
3. **Set Up Design Variables and Responses:**
- Within the Optimization Study, define the design variables and responses you want to optimize.
- Specify the range or initial values for the design variables.
4. **Define Constraints:**
- Add a new constraint by right-clicking on "Constraints" and selecting "Add Constraint."
- Choose the type of constraint as "Equality."
5. **Input Constraint Equation:**
- In the constraint setup window, input the mathematical equation that represents your equality constraint.
- The equation should be of the form "Constraint = 0".
6. **Specify Optimization Method:**
- Choose the optimization method you want to use. For example, you can use Response Surface Optimization (RSO) or another suitable method.
- Set optimization parameters such as convergence criteria and maximum iterations.
7. **Run the Optimization:**
- Click on "Start Optimization" to run the optimization study.
8. **Review Results:**
- After the optimization completes, you can review the results, including the optimized values of the design variables, the constraint values, and the achieved objective values.
Remember that the exact steps might vary based on the version of ANSYS you are using and the specific details of your simulation model. Make sure to consult ANSYS documentation or tutorials specific to your version for more accurate guidance.
Additionally, ANSYS Workbench provides various tutorials and documentation on optimization workflows, including applying constraints, which can be beneficial to understanding the process better.
Hello Mostak Ahamed ,
I am appreciated for your concern, but I think there is misunderstanding about " constraints" in my discussion . The ANSYS set the constraints for the objective ( equality and inequality ones), I meant in my discussion the constraints on the parameters or in other word (as ANSYS nomination ) "parameters relationships" they have only inequality relationships which I need to set equality relationships "constraints" for them.
Regards
In general optimization algorithms dont deal well with equality constraints. Hence, nobody offers them. Equality does not exist in numerical schemes a=0.0 is more likely a = 1.e-16 etc.
A couple of work arounds are (1) resolving the equality constraints for design variables. Like in a = b+c, just remove a as a design variable and only keep b and c. then insert a as an equation.
(2) resolve equality constraints into two inequality constraints. a = 0: a > -epsilon; a < epsilon; epsilon = 1.e-3. Don't make epsilon too small as it may make it harder to converge. You may also want to normalize your constraints to bring a between 0 and 1.
- Badges
- Science topic