Defining local stress constraints through DRESP1 in the design region for Topology Optimization is allowed. The Stress-Norm method is used to aggregate stress responses.

It is not recommended to use the global stress constraint along with a mass/volume constraint. The constrained mass/volume may not allow the stress constraint to be satisfied.

The stress constraint definition in a topology optimization is a global constraint and does not target local stress concentrations. These areas can be addressed subsequently through size, shape, and free shape optimization or a combination thereof. Artificial stress concentrations are filtered out during topology optimization with stress constraints. These include regions around rigid connections, concentrations due to hard geometric features such as corners, etc.

Stress constraints may not work well in a model where there is a large differential in response values between design and non-design spaces. In these cases, it is recommended to modify the problem formulation to say, compliance based for example.

If you are trying to optimize a design for its first mode frequency and have constrained the first mode frequency to be greater than your target value, you may encounter the following:

While targeting first mode frequency, it is recommended that you constrain the first few modes (1-5 for example), as opposed to just the first mode. This results in better stability of the problem and improved convergence. The same applies for buckling responses as well.

The reason is that the particular non design part (mounting location) is not needed for the optimal load transfer. If there is a connection to adjacent parts, there should be some loads. Loads will ensure that the region will not be disconnected from the design space.

It is recommended to optimize for cutouts first through topology optimization, and subsequently for thickness, through free sizing.

While running free-size optimization with buckling responses, it is recommended to use a base thickness value. The optimization process has the tendency to drive the design to simultaneous failure of a large number of buckling modes, potentially as many as the number of design variables. This can only be regulated with a properly chosen base thickness that would prevent formation of an increasing number of local buckling modes.

It is not unusual to encounter elements with negative jacobian while running a free shape optimization. Some ways to minimize the possibility of this failure are:

In an acoustic analysis, an interface matrix is calculated between the fluid and structural grids. This interface between the fluid and structural grids is termed the ‘wetted’ surface. Accuracy in calculating this interface is important for determining the accuracy of the solution. While running an optimization, the original interface calculated at the start of the optimization is used throughout the optimization and is not updated during the course of the optimization. As long as the interface is not affected during the optimization, for example, optimizing a component away from the fluid domain, the optimization can be run. However, if during the optimization, the interface is expected to change sufficiently enough through topology, topography or shape for example, the optimization solution may not be accurate. This is based on the implementation as of version 10.0.

The following are some commonly used/suggested optimization problem formulations:

The density method used for topology optimization is mesh-dependent. It also causes checkerboarding (density value changing rapidly between elements). Models with second order elements do not show this behavior and use of minimum member size constraint treats the checkerboarding issue.

Volumefrac response is the material fraction of the designable material volume. The volume response is the total volume, which includes design and non-design volume.

Both formulations are identical as long as the same density rr  is used for all materials in the model. When minimizing the volume, the density value in the material card does not have to be given.

If different densities rr exist in the design and the non-design space, then the optimization result may be different for minimizing mass and minimizing volume.

In a case where the design mass and the non-design mass are almost identical, but the design volume is much smaller than the non-design volume (Vdesign << Vnon-design), minimizing mass will give better results.

On the other hand, when the design volume and the non-design volume are almost identical, but the design mass is much smaller than the non-design mass (rrdesign << rrnon-design), minimizing volume will yield better results.

You can change the move limits (the allowable change of the design variable in each iteration step) for the first iteration. The move limits are defined for each optimization type differently. The parameters are DELTOP, DELSHP, DELSIZ on the DOPTPRM statement. They can also be defined for size and shape variables on the DESVAR card.

In HyperMesh, use the opti control subpanel in the optimization panel on the Analysis page.

To restart an OptiStruct job, two files are needed: filename.fem (containing input data) and filename.sh (containing design variable information from the end of last complete iteration).

Refer to Restarting OptiStruct in the Running OptiStruct section of the documentation.

Yes, but in order to do this, you must define the topology optimization problem differently.

Instead of defining the complete thickness of the shell element as designable, a core non-designable thickness, T0, must be provided in addition to a maximum thickness, T, which includes the core thickness and a designable region (where the ribs can be grown). The topology optimization can then be performed as usual.

The rib patterns can be interpreted by visualization of the density contour plots. Wherever the density values reach 1.0, a rib with the height of the -core plus the designable thickness is needed. Where the density values go to 0.0, no rib is required and it just retains the core thickness.

Example:

Component minimum core thickness: 1.5 mm
max thickness with ribs: 3.5 mm

The thickness on the PSHELL card is set to 3.5 mm and T0 on the DTPL card is set to 1.5.

This can lead to 2mm high ribs (max).

Since shell elements are being used, the aspect ratio has to be taken into account. Do not let the expected thickness be greater than the in-plane dimensions of the elements.

Modeweights are assigned from within a model analysis subcase using the MODEWEIGHT subcase information entry.

From within the HyperMesh panels, access the optimization panel and the responses subpanel. If the wfreq or comb is chosen as the "response type," you can assign up to six modeweights to six different modes.

If the value of the subcase information entry NORM is left as 0.0 (default), then the normalization factor is estimated during the zero-th (analysis only) iteration of the optimization run (such that the static compliance and weighted frequencies are equally weighted). If you provide a value other than 0.0, that value will be used in the optimization. In HyperMesh, the value of NORM may be defined in the response panel when the "response type" is comb.

The normalization factor (NORM) is used for normalizing the contributions of static load case compliances and the inverse of eigenvalues when using the combined compliance index as the objective function. A typical structural compliance value is on the order of 1.e4 to 1.e6. However, a typical inverse eigenvalue, as in the case of using the combined compliance index as the objective function, is on the order of 1.e-5. If NF is not used, the linear static compliance requirements dominate the solution. Refer to Responses in the User's Guide for more information on the "Combined Compliance Index" and the use of the NORM value.

In the first method, you can tell OptiStruct to use only a certain fraction of the designable volume (i.e. constraining the volume).  OptiStruct will redistribute and reorient the amount of the material within the design domain while optimizing the objective function and satisfying any other constraints specified.

In the second method, you can choose their objective to minimize the volume. Here, OptiStruct will minimize the volume to arrive at a final topology that satisfies other constraints.

In order to increase stiffness, minimize compliance should be used with forces and maximize compliance with prescribed displacements.

The compliance is defined as:

Compliance ~ Force · Displacement

When prescribed displacements are used, the reaction force must be increased to increase the stiffness. This means that the compliance has to be maximized.

In case the forces are given, a stiffer structure means having lower displacements. To achieve this goal, the compliance needs to be minimized.

If the objective of the design problem is to minimize volume response or mass response, the initial material fraction will be set to 0.9 by default. If a mass or volume constraint is used in the design problem, the initial material fraction will be the value corresponding to the value defined on the constraint. When mass or volume response is not being used to define the objective or the constraint, the material fraction will default to 0.6 at iteration 0.

The initial volume fraction can be assigned to 1.0 (or any other value between 0.0 and 1.0) by defining the MATINIT parameter on the DOPTPRM bulk data entry. This can be done from the HyperMesh interface by changing the value of the MATINIT parameter and checking the corresponding check box in the opti control subpanel, under the optimization panel (located on the Analysis page).

Yes, any type of combined optimization can be performed with OptiStruct. It is recommended, however, that different types of optimization be performed separately first. This will help you understand the performance of the structure with different kinds of optimization before attempting a combined optimization.

To set up an optimization problem using more than one optimization type, access the optimization panel and choose the subpanels for the optimization types to be performed (topology, topography, size and/or shape). If more than one subpanel is chosen in the optimization panel, (both topology and topography, for example) OptiStruct will automatically perform the combined optimization.

For details on setting up different types of optimization, refer to the OptiStruct tutorials of the documentation.

Nothing is wrong.  You have defined a design space for your topology optimization. In an analysis-only run, the density of the design space is set to 1.0. In the first iteration of the topology optimization, the density of the design space is less than 1.0, unless you have expressly set the MATINIT value, defined on the DOPTPRM bulk data entry, to 1.0. Therefore, in the first iteration, the structure appears to be not as stiff as in an analysis-only run.

For topology optimization runs with mass (volume) as the objective, the default for MATINIT is 0.9. For runs with constrained mass (volume), the default is reset to the constraint value. If mass (volume) is not the objective function and is not constrained, the default is 0.6.

To perform an analysis only run with the modal parameters set to the values that will be used for the first optimization iteration, do not use the ANALYSIS command, and set MAXITER = 0.

You can open the iteration history file .hgdata using Altair HyperGraph (or HyperView) while OptiStruct is still running and create plots of the Objective, Constraints, and Design Variables against Iteration. Clicking Apply on the Edit Curves panel, the view can be updated periodically.

The von Mises stress constraints may be defined for topology and free-size optimization through the STRESS optional continuation line on the DTPL or the DSIZE card. There are a number of restrictions with this constraint:

There are several barriers for buckling constraints in topology optimization:

Because of the above reasons, for the time being, reasonable success can only be expected for one class of design problems -- shell structures with non-zero base thickness.

In order to find the rigid body modes, you can add a diagnostic command OSDIAG,46,1 to the I/O options or subcase information sections of the OptiStruct deck. The diagnostic command will output the dofs that are constrained and provide the rigid body mode results to the user.

The sensitivities of volume are so low for topography optimization that they do not make a significant contribution toward the optimization. It is recommended that using both volume and mass as responses be avoided.

This can be achieved through the use of DEQATN, DRESP2, and DRESP1L cards. A DEQATN is defined to sum a number or responses.  This DEQATN is referenced on a DRESP2 definition, which uses the DRESP1L continuation to select the same DRESP1 response for a number of different subcases.

Within HyperMesh, this can be achieved by first defining a dequation. Then, in the response panel, select function as the response type, select the already-defined dequation, click edit, select Responses_by_loadstep from the available options, and select the same response for a number of different loadsteps.

For example, consider a control arm with three loadsteps. The control arm consists of two components; one of which is designable. It is necessary to add the regional compliance of the designable component for three loadsteps together, and then minimize the resulting compliance. The constraint on the volume fraction is not to be more than 0.1. A section of the input file is shown below:

$

$  OBJECTIVES Data

$

$

$HMNAME OBJECTIVES       2objective

$

DESOBJ(MIN)=5

$

$

DESGLB        9

$

$

$HMNAME LOADSTEPS       1brake

$

SUBCASE       1

 SPC  =       1

 LOAD =       2

$

$HMNAME LOADSTEPS       2corner

$

SUBCASE       2

 SPC  =       1

 LOAD =       3

$

$HMNAME LOADSTEPS       3pothole

$

SUBCASE       3

 SPC  =       1

 LOAD =       4

$

BEGIN BULK

$HMNAME DESVARS        1topo_arm

$TPL    1       0.0     0.0     1       1       14

$

$  OPTIRESPONSES Data

$

DRESP1  3       comp    COMP    PSOLID                                14

DRESP2  5       wcomp          1

+       DRESP1L        3       1       3       2       3       3

DRESP1  9       vol     VOLFRAC

$

$HMNAME DEQUATIONS       1eq1

$

DEQATN  1       f(a,b,c)=a+b+c

$

$

$  OPTICONSTRAINTS Data

$

$

$HMNAME OPTICONSTRAINTS       8vol

$

DCONSTR        8       9        0.1

DCONADD        9       8

$