OS-E: 0825 Multi-Model Optimization (MMO) – Design Scaling with Pattern Repetition in 2 Different C-Clips

Multi-Model Optimization can be used in applications that require optimizing parts of different sizes. This is accomplished by using the SCALE continuation line on linked DTPL and DSIZE entries in the models on which the scaled design is to be applied.

In this example, Topology optimization design variables that are linked between the two models are used (Figure 1). The entire model is the design space made up of PSHELL elements. The Boundary Conditions are different for each model.

Model Files

Refer to Access the Model Files to download the required model file(s).

The model file used in this example includes:

c_clip_mmo.zip

Model Description

Model 1 and Model 2 are subject to differing boundary conditions. The opposing forces applied to Model 1 try to move the clip arms outward, whereas, in Model 2 the applied forces try to move the arms inward. Additionally, Model 2 is also scaled by 2.0 in the X direction and 3.0 in the Y direction.

In this example, focus on generating a linked common design for the two models. However, since the two models are of different sizes, the SCALE continuation line to scale the design from Model 1 to Model 2 are used.

FE Model
Element Types: CQUAD4; CTRIA3

The linear material properties are:

MAT1
Young’s Modulus: 2.1E5
Poisson's Ratio: 0.3
Initial Density: 7.9E-9

Multi-Model Optimization Setup

The optimization setup for each model is:

Optimization Type: Topology Optimization
Responses: Mass Static Displacement
Objective: Minimize Mass
Constraints: Static Displacement (2 Grids)

First, set up the individual topology optimization models and test them individually to determine if there are any errors or important warnings during runtime.

A main input deck is created that utilizes the ASSIGN,MMO entry to identify the models to be included in the Multi-Model Optimization. The Model Name can be specified on the third field of the ASSIGN entry (this name can be used to identify and qualify the responses used on the DRESPM continuation line, if applicable). In the current run, the main file is named main.fem and the multiple models are named c_clip.fem and c_clip_scaled.fem.

Note: In the main file, the individual models should be located in the same directory (working directory) as the main.fem file.

ASSIGN,MMO,bc1,c_clip.fem
ASSIGN,MMO,bc2,c_clip.fem
$
BEGIN BULK
$
ENDDATA

A minimum of one set of Design Variables or Design Domains should be linked between the multiple models identified for Multi-Model Optimization. Linking is accomplished by using the same Identification Number (ID) for the design variables or domains. This is used to internally map design domains to match the optimized results into a single model.

Note: The presence of other unlinked design variables and domains can lead to different final optimized designs for the common design space in the multiple models.

Linked design variables and domains should be identical in all respects (except for the COORD and SCALE continuation lines). The applied design parameters and manufacturing constraints should match in all models. This restriction does not apply to unlinked domains. Additionally, to facilitate internal pattern repetition, a COORD continuation line is required on the corresponding design variable or domain entry (DTPL, DSIZE, and so on). This allows identification of the location and direction of the repeated pattern across linked design domains.

The design domains are linked as:

For Model 1, the DTPL entry with ID 1 and for Model 2, the DTPL entry with ID 1 are linked. OptiStruct internally selects one of them as the main and applies internal design mapping across models.

The SECOND continuation line identifies the scaling factors that are applied to the design.

The COORD continuation line allows the identification of the anchor point and mapping of the design among multiple models. This continuation line is mandatory.

Results

The scaled design has a similar pattern to the original design (Figure 3).

In the final result, the design is scaled in Model 2.