Altair's Student Guides - Instructor's Manual - CAE and Design Optimization - Basics

Student Project Summaries

CAE and Design Optimization - Basics Contents

Introduction.............................................................................................2 Installation Instructions: .......................................................................2 Redesign of Aluminum Housing.................................................................3 Description of the Problem ....................................................................3 Design Objectives .................................................................................4 The Analysis Model ...............................................................................5 The Optimization Model.........................................................................6 Results.................................................................................................6 Further Work ........................................................................................6 Summary .............................................................................................8 Frequency Optimization of Vehicle Floor ....................................................9 Description of the Problem ....................................................................9 Design Objectives ............................................................................... 10 The Analysis Model ............................................................................. 10 The Optimization Model....................................................................... 11 Results............................................................................................... 12 Further Work ...................................................................................... 12 Summary ........................................................................................... 13 Shape Optimization of Engine Block ........................................................ 14 Description of the Problem .................................................................. 14 Design Objectives ............................................................................... 14 The Analysis Model ............................................................................. 15 The Optimization Model....................................................................... 16 Results............................................................................................... 17 Further Work ...................................................................................... 17 Summary ........................................................................................... 17 Battery Tray Design ............................................................................... 19 Description of the Problem .................................................................. 19 Design Objectives ............................................................................... 19 The Analysis Model ............................................................................. 20 The Optimization Model....................................................................... 22 Results............................................................................................... 22 Further Work ...................................................................................... 22 Summary ........................................................................................... 22

1


CAE and Design Optimization - Basics

Introduction This material is best used after reading the book CAE And Design Optimization – Basics. Access to HyperWorks software is not essential for you, the instructor. Of course, if you choose to solve the problem yourself before working with your students, you will need HyperMesh and OptiStruct. This book describes 4 assignment problems that highlight different applications of HyperWorks. Each problem is independent, and is complete in itself. Students may choose to do more than one, depending on their interest. To make best use of this material you will need a computer with a soundcard and speakers. Your computer should have a media-player programme (such as Windows Media Player) and an Internet Browser that supports JavaScript. The material can be copied to a server and accessed by clients. You can customize the HTML files to suit your requirements. After opening the file, doubleclick on any text to edit it. Use the save changes link on the left of your Browser window when you are finished.

Installation Instructions: 1. Copy the folders to your computer or to your server. If you are working on a server, it is a good idea to set the folders to “read only” to prevent inadvertent modifications. 2. The videos are best played in full-screen at a resolution of 1024 x 768. You may need to install the CamStudio Codec to view video on your computer. To do this, right-click on the file camcodec.inf and choose Install from the popup menu. You may need administrator privileges to do this. 3. Ensure that JavaScript is enabled on your browser. 4. Each folder contains one HTML file. Double-click on it to open the instructions. 5. Data files are provided as relevant – IGES files, HM files, etc. 6. In case you need support, contact your distributor or email [email protected]

2



Redesign of Aluminum Housing Areas covered:

Software used:

• • • • •

Geometry Abstraction for FE modeling Automatic Mesh Generation Constrained optimization with Stress constraints Design interpretations of FE results Transfer of an FE model to a CAD modeler for postoptimization detailed-design

• • •

HyperMesh OptiStruct HyperView

Description of the Problem A military tank carries projectiles in tubes. To protect the projectiles, a flap closes the exit of the tube. When ready to fire the flap opens, leaving the exit clear for the projectile. Each flap is controlled by an actuator, which is an electromechanical device consisting of a motor, a cam, a gear train, and a few other components. All components are mounted inside a housing, which is then sealed. The housing itself is bolted to the tank body. In the current design the housing is machined from an Aluminum casting.

3



The housing-manufacturer has been informed that the units have developed cracks near the mounts. No data on conditions causing the failure, pattern of failure, etc are available. No changes can be made to the mounting points on the tank. The internal components (i.e. the actuator) also cannot be modified or relocated. The design problem is: How would you redesign the housing? The supplier wants to change the housing material from Aluminum to Steel. One option is to maintain the same dimensions and switch the material to Steel. The student should be asked to consider that since the forces that cause stresses are due to the deadweight, increasing the stiffness will not help if the mass also increases. Steel is stiffer than Aluminum, but is also heavier.

Design Objectives Since we have no information on the variation of the actual loads with time, we will use an equivalent static analysis. In this approach, we use a factor of safety to allow the static design to mimic the dynamic conditions. Interested students can look up the MIL-Standard for shock loads to be rigorous. Since the unit has worked for a substantial amount of time before failing, failure is likely to be either because of a special-case overload, or due to fatigue. The latter is probable, since Aluminum does not have an infinite fatigue life. Steel is a better choice from this point of view. Our approach, for optimization, will be to reduce the stresses, regardless of the permissible stress. This is a fair approach to take since we do not know the original factors of safety. To achieve this, we first perform a baseline analysis - an FEA of the initial design to establish stresses for the given load. Since we are performing linear analysis, the actual value of the load does not matter.

4



We then run the optimization with the goal of reducing the stress to less than that in the baseline analysis, under the same loads, using steel. To sum up, build an Analysis Model, calculate the stress in the component, and build an Optimization Model using the existing design as the design space. We constrain the stress to be 0.75*baseline stress, and our objective is to minimize the mass.

The Analysis Model Since OptiStruct uses an Optimization Model, the Analysis Model need not be perfect. It should, of course, be adequate. Your students should perform the baseline analyses with at least two different element-sizes to confirm that the solution is fine enough. The housing poses a couple of challenges from the point of view of meshing. The first, of course, is which elements to use. The second is in the details to keep. We will choose to use solid elements since we believe shear dominates (the walls are thin, but the length:thickness ratio is fairly small). We will neglect the lip that's grooved into the top face. We will also omit the fastening holes. Hex elements are better than tetrahedral elements but are harder to create. It's a trade-off between spending more time meshing and more time solving. For a student working on optimization, tetrahedra are adequate, provided the student has verified that the results are adequately captured. Solid elements have only 3 degrees-of-freedom per node (the x, y and z components of deformation). Since we will define the density of the material, the FE solver will calculate the mass of each element. We will also define special elements called mass elements to represent the masses of the internal components - since they are not being designed, we will not model the components themselves.

5



Once these properties are assigned, we will define the acceleration due to gravity (direction and magnitude) so that the inertial forces can be calculated. After the analysis, we will review the von Mises stress and record this for use as the design constraint for the optimization model.

The Optimization Model Stress as a response requires special attention. Since sharp corners can cause stress to be singular, OptiStruct uses an average measure of the stress. This is why stress as a constraint, unlike displacement, is specified for the complete model rather than at specific points.

Results Topology optimization decides an equivalent element density for each element in the design space. A density of 1 means material must fill the element, while a density of 0 means the element needs no material. For most elements, this value will vary between 1 and 0. As a designer, you will need to exercise your judgment. For example, you may decide that you will omit material from all elements whose density is less than 0.3 (or 30%). Having made this decision, you will need to take the geometry back to your CAD modeler, smooth it out (that is, use geometrically regular edges or surfaces, etc.) and re-evaluate the design for a satisfactory stress.

Further Work The assignment brings home the advantages of the Optimization-as-anintegral-part-of-CAE, as well as the advantages of using HyperWorks: • • • • •

designs are more likely to pass subsequent verification designs are not unnecessarily over-designed easy setup of the FE model easy creation of the Optimization model convenient reporting options with easy viewing

Depending on their proficiency, students may want to

6

Student Project Summaries • • •


estimate the change in center-of-gravity, since this affects the forces at the mounting screws. The center-of-gravity can be a constraint for optimization investigate the effect of mesh size on the optimization results use shape optimization to choose a fillet radius for the various fillets

7



Summary By the end of this assignment, the student will know how to • import IGES files into HyperMesh • use the Model Browser • zoom, pan and rotate • change colors of entities • control visibility of geometry • create collectors for material data, elements, forces and restraints (SPCs) • measure distances and the diameter of circles • find the centers of circles • use consistent units • check for different types of element-edges - free, shared, etc. • fill and stitch surfaces • create fixed points • use temp nodes • create mass elements • create a load case • ask for reactions to be output by the solver • plot stress contours • view deformed shapes • use unaveraged stresses to check if the model is adequate • organize elements in collectors to define the design space • specify responses that the optimizer needs • specify stress constraints and deflection constraints • set the objective • control the optimizer • check for convergence and violation of constraints using the text output files • check for convergence and violation of constraints using HyperView • view density plots in HyperView • run OS-Smooth to export data back to a CAD modeler

8



Frequency Optimization of Vehicle Floor Areas covered:

Software used:

• • • • • •

Geometry Abstraction for FE modeling Automatic Mesh Generation Constrained optimization with frequency constraints Use of manufacturing-constraints Design interpretations of FE results Transfer of an FE model to a CAD modeler for postoptimization detailed-design

• • •


Description of the Problem Typically made of sheet metal, the floor separates the seating / cargo area from the underbody. As with most other components in a vehicle, avoidance of resonance is important. Vehicle designers provide ranges of frequencies to component designers, who must ensure that these are complied with. Plane sheet panels exhibit poor stiffness and NVH performance due to their flexibility. A common and cost-effective approach in the automotive industry to improve the stiffness and NVH performance of sheet panels is the addition of beads. The design challenge is to make changes to the proposed shape, without altering mounting points or other assembly aspects, to ensure that the base frequency is above 40 Hz.

9



Design Objectives The design problem involves the specifications, the methods of addressing these, and the potential for success. The last point is the most important: in design, a negative result is also extremely useful. It is very possible that the floor, with the given thickness, will not be able to attain the required characteristics. Beads (also referred to as swages) are subject to manufacturing constraints. Design requirements of punch-and-die sets often conflict with requirements of product-design. Within the floor itself, a bead may interfere with another component, so we will want to keep the bead-depth down to a reasonable level. In an application like this, topology optimization is a definite possibility. However, since this is a floor, we do not want to use cutouts. Size optimization can also help us suggest a better thickness than the given 3.25 mm. We will use topography optimization since beads are an accepted stiffening-approach, and are most likely to be accepted by both designers and manufacturing-engineers. Our design approach is fairly simple: build an Analysis Model, then build the Optimization Model with the base frequency and bead characteristics as constraints. The Objective is to maximize a weighted sum of the first 5 frequencies.

The Analysis Model In these instructions, only one mesh will be created. Your students should perform the analyses with at least two different element-sizes to confirm that the solution is fine enough. Shell elements lend themselves readily to the problem. 10



How do we decide the element size? Numerical models sometimes introduce spurious modes of vibration, and it is hard to filter these out without access to test data. In this problem, we will use values that keep the solution-time within limits. Since shell elements have 6 degrees-of-freedom per node, a 20,000 node model has 1,20,000 dofs. This can be quite time consuming! Your student should review the basics of modal analysis: the Rayleigh-Ritz method is essential reading, while a quick review of Simultaneous Vector Iteration (SVI) and Lanczos is recommended. We will ask the software to calculate the first few mode shapes. Since changes in the geometry can change mode shapes, we will ask OptiStruct to track the modes. This is a very useful feature, whereby OptiStruct uses the eigenvalues to ensure that it is indeed constraining the correct modes. We will also use two of HyperMesh's strongest features: autocleanup and QI Meshing. We specify the range of element properties we want to accept and allow the software to determine which features to retain. Used with care, it can substantially speed up the model-building process.

The Optimization Model Why did we identify that the base frequency as a constraint and not as the objective? Because at this stage we will be satisfied if we can exceed the given value: there is no need to take it even higher. Your student should realize that there are other ways to alter resonant frequencies. For instance, adding a point of support can make a substantial difference. We study how to organize elements in collectors to define the design space, and how to specify manufacturing constraints for topography optimization. 11



We also use the Mode-tracking option since changes in geometry can result in a change in eigenvalues: for instance, what was the 3rd mode may become the 4th, after the change.

Results Since the solution time for the floor can be quite substantial, the instructions for this section use a sample geometry. We study how to check for convergence and violation of constraints using the text output files, how to check for convergence and violation of constraints using HyperView. We also view bead plots in HyperView and use OsSmooth to export data back to a CAD modeler

Further Work The assignment underscores the advantages of treating Optimization-as-an-integral-part-of-CAE, and the advantages of using HyperWorks: • • • • •

designs that are more likely to pass subsequent verification designs that are not unnecessarily over-designed easy setup of the FE model easy creation of the Optimization model convenient reporting options with easy viewing

You may choose to assign further investigations to your students based on their level of proficiency on CAD, the time available, etc. Some of the areas that could include • •

12

the use of size optimization to suggest a better sheet thickness the inclusion of a Static load case to include loadbearing capacity in the design requirements



Summary By the end of this assignment, the student will know how to • import IGES files into HyperMesh • use the Model Browser • zoom, pan and rotate • change colors of entities • control visibility of geometry • create material collectors • create component collectors • create load collectors • create loadcases • measure distances • use consistent units • work with different types of edges - free, shared, etc. • fill and stitch surfaces • measure the diameter of circles • find the centers of circles • use temp nodes • use autocleanup to do all of the above • create restraints or SPCs • specify options for the eigenvalue-solver • organize elements in collectors to define the design space • specify responses for the optimizer • specify topography constraints - i.e., the bead parameters • use manufacturing constraints • set the objective • control the optimizer • check for convergence and violation of constraints using the text output files • check for convergence and violation of constraints using HyperView • view bead plots in HyperView • export data back to a CAD modeler

13



Shape Optimization of Engine Block Areas covered:

Software used:

• • • • •

Geometry Abstraction for FE modeling Automatic Mesh Generation Free-shape optimization with Stress constraints Design interpretations of FE results Transfer of an FE model to a CAD modeler for postoptimization detailed-design

• • •


Description of the Problem An assembly model for the engine of a model aircraft has been proposed. Any design changes must only be incremental - structural changes are ruled out by the project-deadlines. Instead, the designers have asked that attention be paid to the outer profile of the engine block. The radii chosen at the CAD stage were not based on any calculations. The goal now is to start with the given shape and check if changes to the shape can reduce stresses.

Design Objectives Engine design is multi-disciplinary, since heat transfer and fluid flow (for combustion) are at least as important as structural design. There are also forces that the engine experiences because of the movement of the piston. It may seem that a topology optimization on the engine block to suggest material layout is a good idea, but there are thermal considerations that are important, and these have not been provided to us. Remember that thermal heat capacity of the engine, its heat-dissipation characteristics, etc. are more important than weight-reduction. 14



There is no functional basis for the outer-profile of the engine block. We can only work with the exterior, and with areas away from the mounting and assembly points. Ideally we should carry out an analysis of the entire assembly. Unfortunately that is not within the scope of the design. In actual usage, a range of allowable values for subsystems (stresses, natural frequencies, etc.) should be provided to the component designer by the system-integrator. In the absence of these, we will concentrate on the behavior under static loading. Since we do not have any data on the loading conditions, we will use a compare-with-original approach. Even here, we are not given any idea of the loading. Therefore we will assume unit acceleration along each of 3 orthogonal directions. We will want to monitor the stress in several elements, so we will ask the optimizer to minimize the maximum of these stresses - this is called a minmax optimization. Our design approach is build an Analysis Model, build the Optimization Model with stress as the response. The Objective is to minimize the maximum stress in the area of interest.

The Analysis Model Since OptiStruct uses an Optimization Model, the Analysis Model need not be perfect. It should, of course, be adequate. Your students should perform the analyses with at least two different elementsizes to confirm that the solution is fine enough. Deciding which elements to use is not easy. Since the engine block is a complicated shape, solid elements are the only option, but should we use tetrahedra or hexahedra? The latter are superior in performance but the effort of creating a mesh is quite high. Since our focus is on design, we will choose to work with 15



tetrahedra - this allows us to focus on the design aspects without getting tied down in meshing complications. Deciding what size of elements to use is not easy either. We will start with a problem-size that is relatively small so that we can first ensure that the problem is well-posed. Since solid elements have 3 degrees-of-freedom per node, a 10,000 node model has 30,000 dofs. This is reasonable for static analyses. Our model will have 1 material (Steel), 1 element type (tetrahedra), 3 loads (accelerations along three axes), 1 set of restraints, and only 1 type of solution (static). To take advantage of symmetry, we will build a solid from the IGES surfaces and split it along the plane of symmetry. Our problem involves Body Loads, which act at all points in the body. We will create a load case that is a combination of other loads: after just one solve, we will want to see the stresses separately for the xdirection acceleration, for the y-direction accelerations, for the z-direction acceleration and for all three acting together.

The Optimization Model It is a good idea to ensure that the students understand the full power of free-shape optimization. The software can actually suggest the external shape - this is remarkable. Encourage your students to experiment with setting up different parts of the component as design variables. But do not use a brute-force approach. Choosing all external nodes and identifying them as design variables can prove to be expensive, since the optimizer will set up variables for each node! Free size optimization hides some maths from the user. Interested students may want to read the on16



line documentation on Shape Optimization for an indepth understanding.

Results Our principal interest is in the shape-variables. We will want to apply the recommended values to the FE model. This causes the FE model to be modified to match the recommendations of the optimizer. We will then build surfaces from the FE mesh and export IGES geometry for CAD usage.

Further Work The advantages of the Optimization-as-an-integralpart-of-CAE are clear, as are the advantages of using HyperWorks: • • • • •

designs that are more likely to pass subsequent verification designs that are not unnecessarily over-designed easy setup of the FE model easy creation of the Optimization model convenient reporting options with easy viewing

You may choose to assign further investigations to your students based on their level of proficiency on CAD, the time available, etc. Some of the areas that could include • • • •

application of free-shape to other areas in the model - for instance, wall-thicknesses the inclusion of natural frequencies as a response a study of the effect of design changes on the center-of-gravity of the engine block inclusion of a displacement constraint limiting the maximum movement of the nodes

Summary By the end of this assignment, the student will know how to • import IGES files of assemblies into HyperMesh 17

Student Project Summaries • • • • • • • • • • • • • • • • • • • • • • • • • •

18


use the Model Browser zoom, pan and rotate change colors of entities delete unwanted imported data control visibility of geometry create material collectors create component collectors edit a component collectors to ensure it is using the correct material create load collectors create body loads create load combinations use multiple load cases create restraints or SPCs apply symmetry BCs create solid geometry cut a solid with a plane measure distances use consistent units perform an analysis and obtain baseline results choose nodes that constitute the design space for Free Shape optimization specify a minmax objective control the optimizer check for convergence and violation of constraints using the text output files to check for convergence and violation of constraints using HyperView to view and apply shape changes to export data back to a CAD modeler



Battery Tray Design Areas covered:

Software used:

• • • • •

Geometry Abstraction for FE modeling Automatic Mesh Generation Constrained optimization with FRF excitation Design interpretations of FE results Transfer of an FE model to a CAD modeler for postoptimization detailed-design

• • •


Description of the Problem A design for the tray to hold the battery of an automobile has been proposed. The basis for the model is a study of earlier designs. The designer wants to know how the tray will perform under the range of excitations it is likely to experience when mounted on the vehicle. Also, the CAD designers have asked whether ribs can be added to minimize the deformation. The goal is to start with the given shape and check if the FRF displacement can indeed be reduced, and to suggest to the designer the likely increase in weight. The starting point of the problem is the IGES file of the CAD assembly. A frequency-domain load is applied to simulate a frequency-sweep: from 0 to 1000 Hz.

Design Objectives The proposed design is made from 2 mm thick sheet metal, and is designed to assemble with predesigned components. It has to carry 2 batteries that together weight 15 kg. 19



Design for static loads is quite easy, but the problem here is that when the engine is running, the tray is expected to receive excitations at a wide range of frequencies. How can we design it for these dynamic loads? On a test-rig, we would load the tray and excite it at frequencies within the specified range. In testingterms, we would "sweep" through all frequencies. FRF (short for Frequency Response Function) excitation is the way to do this using a numerical model. We will first evaluate the natural frequencies of the proposed design. We will use these naturalfrequency-and-mode-shape pairs to perform a "modal response" analysis for frequencies within the specified range. Next, we will ask OptiStruct to tell us how to stiffen the material to minimize the displacement at a point of interest. Rather than using topography optimization, we will use topology optimization. We will use the results of the optimization to decide where to place ribs under the tray. Topography optimization and size optimization could also have been used in this case.

The Analysis Model The model is quite simple, geometrically. The mesh is likely to be quite small, so the challenge does not lie in the FE mesh preparation. It lies in specifying and interpreting data that simulates a frequencysweep. The element choice is easy - shell elements. It does not matter much whether we mesh the outer surface, the inner surface or the mid-surface. A quick fix on geometry errors and automeshing can give us an adequate mesh. In this project, however, we will see that it is sometimes easier to fix the mesh than to fix the geometry.

20



Since the model is small, we will not need to worry much about the element size, though the student must, of course, ensure that the results are adequately accurate. We will need to be careful in our choice of units. Analysis results and data are specified in cyclesper-unit-time. If if we use SI units we can work with Hz (cycles / second). A baseline analysis establishes the mode shapes of the proposed design. The specifications call for an excitation between 0 and 1000 Hz. We therefore use the Lanczos method to obtain mode shapes upto 3000 Hz. We will use the modal-analysis method, where we approximate the transient behaviour as a weighted sum of the mode shapes. To reduce error, we use mode shapes upto 3000 Hz thrice the excitation range. The restraints are easy to setup - clamps at the specified locations are best represented by SPCs. The loads are a little more complicated. Specifying FRF loading requires care. We specify the frequency range through which we want to sweep using the TABLED1 card. The DAREA card specifies the amplitude of the excitation, and the RLOAD2 card puts the frequency and the amplitude together. We then specify the frequencies at which we want to evaluate the responses. We use the FREQ1 option to choose intervals of evaluation. Finally, we use the ~EIGRL option to specify which modes are to be evaluated - all evaluated modes are used in the modal-evaluation of the FRF response.

21



The Optimization Model The bottom of the tray is our design space. We want to minimize the FRF response at a selected node. Our constraint is the mass. We will be liberal with the mass constraint for the first setup. If the results look promising, we can investigate the effect of tightening the mass constraint. You should encourage your students to experiment with the constraints.

Results Our principal interest is in the element thickness. We will see what OptiStruct has recommended, and use these to suggest where ribs should be used under the tray.

Further Work The advantages of the Optimization-as-anintegral-part-of-CAE are clear, as are the advantages of using HyperWorks: *designs that are more likely to pass subsequent verification *designs that are not unnecessarily over-designed *easy setup of the FE model *easy creation of the Optimization model *convenient reporting options with easy viewing

You may choose to assign further investigations to your students based on their level of proficiency on CAD, the time available, etc. Some of the areas that could include *the use of topography optimization (remember that mass should not be used as a response!) *the use of minmax optimization to minimize the frequency at more nodes than one *a study of the effect of design changes on the center-of-gravity of the battery tray

Summary By the end of this assignment, the student will know how to 22


• • • • • • • • • • • • • • • • • • • •


import IGES files of assemblies into HyperMesh use the Model Browser zoom, pan and rotate change colors of entities control visibility of geometry create material collectors create component collectors create load collectors create FRF loads using the RLOAD2, DAREA and TABLED1 cards create FRF evaluation frequencies using the FREQ1 card use the Modal-method to evaluate FRF response create load combinations create restraints or SPCs measure distances use consistent units perform an analysis and obtain baseline results view and animate mode shapes control the optimizer check for convergence and violation of constraints using the text output files to check for convergence and violation of constraints using HyperView

23

Altair's Student Guides - Instructor's Manual - CAE and Design Optimization - Basics

Recommend Documents