 ## General Mechanical

#### ANSYS DOE module to reduce high number of loads to equivalent reduced number of loads

• ahm.eltaweel
Subscriber

I have a loaded part that is exposed to more than 30 forces.

I would like to reduce the 30 loads to only equivalent 5 loads while maintaining the state of stress at some selected points within ±10% of the stress values obtained at the same point in the case of 30 forces.

The bounding limit for each equivalent force will be defined.

The orientation of these equivalent forces is predefined as well.

Is it possible to utilise the DOE module to define the magnitude of these equivalent forces?

If yes, what is the protocol to follow?

As I understand, the current ANSYS version 2022 supports only up to 20 input parameters in the DOE module, correct?

• Ashish Khemka
Ansys Employee

Are you looking to perform a variability analysis?

Guidelines for Selecting SSA Variables (ansys.com)

Regards Ashish Khemka
• ahm.eltaweel
Subscriber

I am looking to do conduct a computational experiment design.
In the simulation (FE) level I have up to 30 direct forces.
However, this high number of loads can't be realised in the physical experiment set up.

I am looking to conduct computational DOE in order to find out an equivalent state of loading of only 5 loads which give a similar state of stress compared to applying the full 30 loads.
I am not looking only for the significance of each input parameter on the output (i.e. sensitivity), but I am seeking computational assistance in finding out an equivalent reduced number of loads.

Best regards Ahmed
• peteroznewman
Subscriber
How many selected points will be used to decide on how closely the 5 forces are creating a state of stress that is similar to the 30 forces?
How many stress components will be used at these selected points? All six components: three normal and three shear components?
• ahm.eltaweel
Subscriber

3 control points.
At each point two normal stresses are compared:
-Max. Principal stress
-Min. Principal stress

Deviation of up to ┬▒10% is allowed.
• peteroznewman
Subscriber
Is this a linear system?
A linear system has only linear contact, such as Bonded or No Separation, or no contact at all.
A linear system has only linear material models.
A linear system has only small deflection.
If you have a linear system then the 30 forces can be applied separately and are additive to the six components of stress at a point.
The sensitivity of each force is immediately known at each of the 3 control points by simply applying a single force.
The same can be said about the 5 forces you want to use to replace the 30 forces.
When you say 30 or 5 forces, are the directions of the forces fixed, so you are only talking about the magnitude of the force, or are there three components for each of the 5 forces?
Max Principal and Min Principal Stress is a nonlinear combination of the six components of stress, so there is an advantage of capturing the six components of stress at the 3 control points, not just the two Principal Stresses to take advantage of the linearity of the system between input forces and output stress components.
The 30 forces will create a target value of the six components of stress at the 3 control points. Then a Least Squares Fit will find magnitude of the 5 forces that minimizes the error of the stress components from the 18 target values.
• ahm.eltaweel
Subscriber
Yes, the system is linear.
Yes, the 30 forces are of known orientation.

The orientation of the 5 sources was assumed (in terms of angulations w.r.t the coordinates system) to simplify the analysis (experience-based assumption) however, how complex the problem could be if the orientation (in terms of three directional components) of the 5 sources isn't defined?
i.e. the magnitude of each force of the 5 forces to be defined in terms of its three directional components.

• peteroznewman
Subscriber
Instead of 5 variables, there would be 15 variables, so it is 3 times the work to get 15 sensitivities.
The Least Squares Fit can do either 5 or 15 variables. I recommend figuring out how to do it with 5 variables first. 