ANSYS Workbench 12 introduces parametric features to find a satisfying design for a hook. The hook should withstand a load of 6000N - the maximum stress should not exceed the yield stress. Some possible candidates for best design will be determined.
ANSYS Workbench 12 introduces parametric features to find a satisfying design for a hook. The hook should withstand a load of 6000N - the maximum stress should not exceed the yield stress. Some possible candidates for best design will be determined.
ANSYS Workbench 12 introduces parametric features to find a satisfying design for a hook. The hook should withstand a load of 6000N - the maximum stress should not exceed the yield stress. Some possible candidates for best design will be determined.
ANSYS Workbench 12 introduces parametric features to find a satisfying design for a hook. The hook should withstand a load of 6000N - the maximum stress should not exceed the yield stress. Some possible candidates for best design will be determined.
Copyright:
Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online from Scribd
Download as pdf or txt
You are on page 1/ 38
Design Exploration with ANSYS V12: The response surface method
Step by step Structural Mechanics example
An ANSYS Tutorial
2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Analysis model
2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Finding a good design
In this introduction to DesignXplorer, we will use the parametric features of ANSYS Workbench to find a satisfying design for a hook. The hook (see following slide for geometry) should : withstand a load of 6000N the maximum stress should not exceed the yield stress; be robust enough so as to resist scattering of the load and unavoidable variations of its dimensions (manufacturing variations); And of course, its weight must be as low as possible. To reach our goal, we will use a DOE analysis: parameters are varied continuously over given ranges and the performance of the hook is examined over these ranges. Some possible candidates for best design will be determined.
2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Hook geometry The model is provided as a DesignModeler file. Parametrization of the model is shown below.
DS_Angle: initial value is 130 DS_Thickness: initial value is 15mm (thickness in XY plane)
DS_Depth: initial value is 20mm (thickness in Z direction) DS_LowerRadius: initial value is 50mm 2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Simulation environment
Hook is made of steel. Cylindrical support applied to restrict radial and axial motion. A 6000N force is applied downwards.
2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Initial status about the model
The mininum safety factor value is 0.54, indicating a failing design (yield stress is exceeded). Initial weight of the hook is 752 grams. 2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Defining the parametric simulation model with ANSYS Workbench
2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Getting started Start ANSYS Workbench 12.0 Insert a Static Structural (ANSYS) system in the schematics Right click on Geometry, select Import Geometry then Browse and select the hook2.agdb file. Select and double-click the hook2 item. Select Model, RMB then Edit (or double click) Note that the project page now contains a Parameter Set bar that holds the DM parameters 2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
A note on geometric parameter
In this examples, the geometric parameters are defined from ANSYS DesignModeler and are automatically collected in the parameter set regardless of their names. If some of them are not required for a parametric analysis, simply deactivate them. Geometric parameters can also be defined directly from your CAD system using a prefix to flag the ones that are relevant for the simulation. You can also mix parameter sources: some could be imported from the CAD model and additional ones defined in ANSYS DesignModeler.
2009 ANSYS, Inc. All rights reserved.
ANSYS, Inc. Proprietary
Setting up the model
(1) Make sure the hook material is Structural Steel (default steel material for WB). (2) Under Mesh: insert a sizing, pick the hook body and set the size to 5mm (you may have to switch the unit system to mm,N so the proper units are displayed); Insert a mapped face meshing and select all faces (the mesher will only map the faces that can effectively be mapped).
(2)
(1)
2009 ANSYS, Inc. All rights reserved.
10
ANSYS, Inc. Proprietary
Creating the simulation environment
Under Static Structural insert a cylindrical support and free the tangential degree of freedom (radial and axial should be fixed). The support is applied to the upper cylinder (1). Insert a force of -6000N in the Y direction (set the force definition to components). The force is applied on the small surface inside the hook (2). Under solution, insert Total deformation, Equivalent Stress and a Stress tool (make sure this one is set on Yield stress per material) (3). Solve the model (optional). 2009 ANSYS, Inc. All rights reserved.
(2)
(1)
(3)
(1)
(2) (3)
11
ANSYS, Inc. Proprietary
Setting the output parameters
Output parameters are the quantities we want to investigate: Mass of the hook; Safety factor; Optional: maximum displacement and maximum equivalent stresses. To set these as parameters, go to: Geometry: under the properties of the bodies, check mass; Solution: check the minimum for the Safety factor (optionally also maximum for displacement and equivalent stresses).
2009 ANSYS, Inc. All rights reserved.
12
ANSYS, Inc. Proprietary
A note on parameters ANSYS Workbench allows to create parameters from most of the items in the simulation tree: Material properties Contacts Loads Boundary conditions Results If you are using APDL code in the Mechanical Simulation Environment: the input arguments (arg0 to arg9) can be parameterized The output (for post-processing operations) can be exposed as output parameters
2009 ANSYS, Inc. All rights reserved.
13
ANSYS, Inc. Proprietary
Save your project
In the file menu, select Save Project.
2009 ANSYS, Inc. All rights reserved.
14
ANSYS, Inc. Proprietary
Response surface modeling and optimization
2009 ANSYS, Inc. All rights reserved.
15
ANSYS, Inc. Proprietary
Principles What is the Response Surface method: Based on the number of input parameters (ANSYS Workbench Simulation, CAD, etc.), a given number of solutions (design points) are required to build a response surface; A Design of Experiments, or DOE, method determines how many and which design points should be solved; Once the required solutions are complete a response surface is fitted through the results allowing designs to be queried where no hard solution exists. Rather than providing values for discrete sets of parameters, the Response Surface provides a continuous variation of the output quantities with respect to the input parameters.
2009 ANSYS, Inc. All rights reserved.
16
ANSYS, Inc. Proprietary
Parametric variations for the hook
We are going to perform the deterministic analysis of the hook for the following parameter ranges: DS_Thickness: 15mm to 25mm DS_Depth: 15mm to 25mm DS_LowerRadius: 45mm to 55mm DS_Angle: 120 to 150 We do not need to specify how many points are to be taken for each parameter: the DOE method will give us the necessary points. Also, after performing the analysis, we will be able to give an estimate of the output parameters for ANY value of the 4 parameters (between their min and max values).
2009 ANSYS, Inc. All rights reserved.
17
ANSYS, Inc. Proprietary
Setup of the response surface
Go back to the schematics page and insert a Response Surface cell from the Design Exploration Toolbox. Select Design of Experiments then set the upper and lower bounds of each input parameters RMB on Design of Experiments then select Update WB will start the simulation for the 25 points required in this example. 2009 ANSYS, Inc. All rights reserved.
18
ANSYS, Inc. Proprietary
Evaluation of the response surface
Once the simulations have been performed, select the Response Surface then Update
2009 ANSYS, Inc. All rights reserved.
19
ANSYS, Inc. Proprietary
Understanding the results
Once the response surfaces have been built, several tools are available to investigate the model: Min/max search; Surface plots; XY plots; Sensitivities; Optimization and trade-off plots.
2009 ANSYS, Inc. All rights reserved.
20
ANSYS, Inc. Proprietary
Min/Max search In the schematics, insert a min/max search system( you may have to activate the beta options to see it). Drag it over the response surface cell. RMB on Min-Max Search then Update The min/max values give a first indication of the performance we can expect from our design.
From this table, we can read that the safety factor ranges from 0.36 to 2.08 and the mass from 517grams to 1469 grams. This indicates a pretty wide variation for both parameters. We need to further investigate to find interesting design points.
2009 ANSYS, Inc. All rights reserved.
21
ANSYS, Inc. Proprietary
Responses
In the Outline of the Response surface, you will see a Response Points folder and a default Response point under it (usually the center of your design space) Each response point can be affected different charts: Spider chart to examine the value and variation of all output parameters on a single graph Local sensitivity to examine the weight of each parameter around the response point Response: 2D or 3D graph to represent the variation of one output with respect to one or two input parameters New response points can be added using the table. Charts are then introduced from the project pages Toolbox 2009 ANSYS, Inc. All rights reserved.
22
ANSYS, Inc. Proprietary
Responses Charts 2D Curves
Select the Response point, then the Response chart. The variation of the first output parameter with the first input parameter is displayed (here, mass vs Thickness) Set the X axis or Y axis values to the variables to be represented. The cursors at the bottom of the Properties window allow to change the values of the input parameters that are not used in the chart 2009 ANSYS, Inc. All rights reserved.
23
ANSYS, Inc. Proprietary
Responses Charts Surface plots
Toggle the Mode option to 3D to get a response surface and set X, Y and Z axis. Use the mouse buttons to rotate or zoom on the surface Moving the mouse over the surface provide the actual value of X, Y and Z
2009 ANSYS, Inc. All rights reserved.
24
ANSYS, Inc. Proprietary
Responses Charts Sensitivities
Select the Response point, then the Sensitivity chart. The chart can be displayed as a bar or pie chart. It shows the local sensitivity of each output parameter to the input parameters Here, we see that the safety factor is mostly influenced by the thickness and depth of the hook, then by the lower radius angle has no effect on it. Thickness or depth increase will increase the safety factor, while the radius will lower it The mass is influenced by all parameters any increase of the parameters increases the mass.
2009 ANSYS, Inc. All rights reserved.
25
ANSYS, Inc. Proprietary
Some conclusions so far
From what we have seen, and especially from the sensitivity plots, we can say that: To reduce the mass, we will need to lower the angle (it has no effect on the safety factor); A decrease of the radius value both lowers the mass and increases the safety factor, so the radius should be kept low; The thickness and depth do increase both mass and safety factor and will require a closer examination to find the right trade-off between minimal mass and reasonable safety factor.
2009 ANSYS, Inc. All rights reserved.
26
ANSYS, Inc. Proprietary
Optimization: finding good designs
Drag and drop a Goal Driven Optimization system from the toolbox over the response surface. A Goal driven optimization can also be inserted on the parameter set. However, since we already have computed a response surface, we will reuse it to perform the optimization task
2009 ANSYS, Inc. All rights reserved.
27
ANSYS, Inc. Proprietary
Setting up the goals
RMB on Optimization, then select Edit In the table, set the optimization goals: Safety factor close to 1.1 set its importance to Higher Mass should be minimized In the properties, select Screening as optimization method with 1000 samples 2009 ANSYS, Inc. All rights reserved.
28
ANSYS, Inc. Proprietary
Candidate designs
Select Optimization in the schematics then RMB and Update or Update Optimization from the project page main menu Three candidate designs are now shown, with ratings (the more stars, the better!) . Candidate A is in this case the best choice, among the 1000 sampled points. Candidate B is slightly below 1.1, so we may not want to consider it. 2009 ANSYS, Inc. All rights reserved.
29
ANSYS, Inc. Proprietary
TradeOff plots A tradeOff plot is a representation of the sample set we used for the goal driven optimization. It can be a 3D or 2D view of input or output parameters versus other parameters. Usually, a 2D plot is more readable than a 3D plot. In our particular case, since we are looking at 2 output parameters at the same time, the 2D plot is better suited. 2009 ANSYS, Inc. All rights reserved.
30
ANSYS, Inc. Proprietary
TradeOff plots In the outline, select Trade-off plot, then set X-Axis to Safety Factor Minimum (Yaxis should be mass) Under response and derived parameters, check only the mass and the safety factor. This graph provides a lot of information about the performance that can be achieved (mass is in kg) The colors indicate how good (blue)/how bad (red) the configurations are with respect to the design objectives (note the vertical blue line for a safety factor of 1.1). 2009 ANSYS, Inc. All rights reserved.
Feasible The red-dotted contour shows the performance you can expect from your design. For example, a mass of 1 kg and a safety factor of 1.1 can be achieved, while a mass of 0.55 kg for the same safety level can not. You can not reach any point far outside of this contour.
31
ANSYS, Inc. Proprietary
TradeOff plots Extracting some samples
You can reduce the number of samples displayed on the graph by moving the Pareto Front cursor in the properties of Outline The points that are removed first are the less adequate for our goals. Since we aim for a safety factor of 1.1, several points are located on the vertical line corresponding to this value.
Candidate A from the optimization
2009 ANSYS, Inc. All rights reserved.
32
ANSYS, Inc. Proprietary
Finding a feasible design
The process that lead to candidate A is the following: Create the response surface; Sample 1000 random points on this surface; Sort the samples to find the one that best fulfills our goal. The values given for candidate A may not represent a design that can be manufactured. You may want to round the values of the dimensions to something that can be manufactured. To do so, go back to the optimization table, right click on Candidate A then insert as response point Select the response surface item in the schematics 33 ANSYS, Inc. Proprietary
2009 ANSYS, Inc. All rights reserved.
Finding a feasible design
Now change the values of the parameters to more acceptable ones. To do so, you may want to look at sensitivities or curves to know how much each of the changes will affect the mass or safety factor. Whenever you change a value, the mass and safety factor are automatically updated. In our case, and based on the knowledge previously acquired, we can select values as shown above (it even provides a slightly higher safety factor for a lower mass!)
2009 ANSYS, Inc. All rights reserved.
34
ANSYS, Inc. Proprietary
Checking the accuracy of the new candidate
It is recommended to have a final check with a full simulation of the model with the selected values so as to check the accuracy of the prediction. Select your modified response point, RMB then Insert as Design Point: the point will now appear in the projects Parameter set Select this point, then RMB and Update Selected design point: ANSYS Workbench will evaluate the response for the point, based on a full simulation
2009 ANSYS, Inc. All rights reserved.
35
ANSYS, Inc. Proprietary
Final model The table of design points now shows the updated values, which are in good agreement with the values predicted by the response surface: we can validate our new design
If we compare the new design and the one we started with, we have raised the safety factor from .54 to 1.1 and the mass has been reduced from 752 grams to 690 grams (-9%).
Initial design Final Design
2009 ANSYS, Inc. All rights reserved.
36
ANSYS, Inc. Proprietary
Good practice for the response surface analysis
The response surface techniques used in DesignXplorer allow to work on a set of 5-10 input parameters at a time. Using more parameter will require much more simulations to provide a reasonable accuracy of the response surfaces. Always check your design candidates with a full simulation so as to remove the approximations due to the response surfaces.
2009 ANSYS, Inc. All rights reserved.
37
ANSYS, Inc. Proprietary
Benefits of the deterministic analysis
Benefits: Using the deterministic analysis, we have been able to investigate the performance of the design over continuous ranges of the input parameters, using a limited number of simulations. From the response surface, we are able to identify the key parameters really influencing the design. Also, we have been able to identify several candidates for the final design. Using engineering practice helps then to choose the final one. The deterministic approach helps identify the performance that can be reached with the actual design (TradeOff plots) 2009 ANSYS, Inc. All rights reserved.
