Finite Element Analysis - Science topic

Martin Veidt Thankyou for the help. I just found that I was applying temperature as boundary condition instead of load.

Kindly check Ansys Composite PrepPost (ACP).

Link to check - https://www.ansys.com/en-in/academic/students

Please select all the nodes.

Martin Veidt : Thank you for your comment, but there is no density dimension in the VUSDFLD subroutine nor Strain.

Based on what dimension density can be changed or assumed?

Can you please explain a bit specifically?

I presume, you are using concrete with GFRP on the bottom face to avoid tensile failures. Which material model is used for concrete ? There is a large stiffness contrast at the interface of concrete and loading plate. Given the choice, I would for coarse mesh at the loading place, and surrounding concrete. Another way is to provide graded stiffness at this interface. This is by increasing the E value of concrete layers, at top most face, and reducing by 10% to 25% based on mesh size. This shold give smooth stress flow from loading plate to concrete, and analysis may not terminate. Try and give feedback.

Thanks Martin Veidt and Christian Willberg for immediate responses.

I tried it 2 to 5 and then increased up to 50 MPa too. Still failed the simulation with excessive distortion in the selected region. When I am increasing the value up to 5000MPa, there was the same scenario, but number of distorted regions tends to decrease. It is important that, this model has no applicable load. Thermal stress, Viscoplastic stress, and oxide growth stress (only for the oxide layer) are available.

I didn't work on this model, sorry.

Hi Murana Awad

The reduction in stiffness will be high if the defect/crack is present in regions of high stress. In a cantilever beam, the maximum stress occurs at the fixed end. Therefore, if the defect/crack is near the free end you can expect a greater reduction in the natural frequency from that of an undamaged beam as compared to when the defect/crack is present elsewhere. You can find a discussion on the same in my research article. I have attached it below for your perusal.

Hope it helps.

Regards,

Jatin

Hello Martin,

Thank you for your input!

I checked my material definition and interaction. I have applied a general contact algorithm (explicit) between the particle and substrate. The contact properties are a friction coefficient of 0.5, pressure overclosure as hard contact, and specified a heat generation of 0.9.

Regards,

Arif

Hi Yunus Rezvanifar,

The best way to do this is to use a subroutine. If the ABAQUS has your material equations, it’s easier to use the USDFLD; otherwise, use the UMAT subroutine. Check the links below. They are articles that can help you write the UMAT subroutine.

Best wishes.

Dear Yunus Rezvanifar ,

Field variables in abaqus are quantities that are used to perform calculations when running user subroutines. They do not describe a specific

physical quantity but are often used to describe solution dependent material properties.

Integration points are referred as material points to distinguish between the

Langragian and Eulerian approach.

it all depends on the objective of the analysis: if the bolt has only to be modeled to include the effect of its pretension, the axisymmetric option seems reasonable. If the goal is to analyse the screw, a 3D model is obviously necessary.

For that thing you may try the following (but again in ANSYS, I hope similar options will be available in COMSOL):

1. Plot the radial deformation of nodes at the interface near the cavity in time history postprocessor it will show at what time point (or load step) diameter of cavity starts increasing.

2. If you are interested in understanding any leak through that cavity, contact pressure zero or contact status open may be used as the indication of the same.

Regards

Thanks, Jigneshkumar for sharing these two links,I will go through it.

Actually what main problem I am facing in my model is that it is a two-step study, and the problem is that in step 2 (time-dependent) the initial value is not taken by step1 (stationary). I tried many things but in this linking of two steps, it is showing errors.

Thanks

I could only provide you some general hints since I am not a specialist in this domain.

1) Could you check literature on the Voronoi model. I am aware that a professor at Ohio State University by name Dr. Siddharth Ghosh works in this domain.

2) Alternately, you could do limited physical uniaxial component level compression and shear tests to understand the stress-strain / load-deflection properties of the test coupon. You could assume the tensile strength as a small number or zero. You could do this for various ratios of asphalt to binder, and various sizes of asphalt, so that you have a reasonable level of characterization. Alternately, there could be literature out there with these results. Temperature would play a significant role. The structure would be brittle under cold conditions and visco-elastic as the temperature increases. You could suitably create a rheological model. Your results depend on the operating temperature. If the material characterization is available (either through aforementioned experiments or literature), you could model the structure as a non-homogenous and anisotropic model with temperature dependent material properties. If there are steel rebars, that could also be modeled, or alternately I believe there is a method called as "the smear technique" for modeling concrete with rebars. This technique should be quite established.

You could use any FE package of choice and do the analysis. Initially you could work on simple compression and shear FE models to ensure that your results match with literature or experiments at coupon level.

3) You could also check commercial packages (or write to them) if they have any readymade constitutive relationship for asphalt. You could try Abaqus, ANSYS, MSC Marc, Altair, etc. I just googled "Finite element analysis asphalt" and it throws up some interesting literature.

My answer is quite generic, but at this moment this is the best I could think of.

For aerospace grade materials (Aluminum and Ti alloys) lot of data is available in open from NASA. I am giving some links here. For your specific case you may try with google "NASA material data handbook + your material".

Nils Wagner Here You are!

It is not true that stress is not transmitting across the body. The results show a max von Mises stress of 1694 MPa, which is a sign of very localised stress singularity, perhaps at a sharp notch or edge. Such a high max value in the color band makes your whole body appear in blue color. Simply change the maximum scale value in your color band (e.g. 500 MPa).

Can you share your Abaqus model (.inp)?

Hello,

First, you need to know what your damage initiation criterion is? After completing your job, select the damage initiation criterion from the Field Output dialog box. Check the frames and legend. Find out when the first point value is greater than one. You can read the time from the step time. To find the point location, select the Contour from the options menu, then go to the Limits tab, and toggle on the Show location to observe the location of the point.

Also, you can refer to this link: https://caeassistant.com/product/abaqus-for-beginners-structural-engineers-in-the-field-of-civil-engineering/

Hi Filippo,

There is different software you could use to obtain a 3D model from a stack of CT images. If you have a license you could use Avizo for instance or if you look for a free alternative there is 3D Slicer. Some software can directly create a tetrahedral mesh (the last version of Avizo for instance) but other will export a surface mesh (made of triangles) that you will have to transform into a volumetric mesh afterward. You can indeed segment separately cortical and trabecular bone to obtain more precise FE models, here is an exemple on sabertooths: Figueirido & al 2018 Distinct Predatory Behaviors in Scimitar- and Dirk-Toothed Sabertooth Cats but some recent papers don't make the difference between those two and as it sounds you don't have any experience with segmentation or FEA maybe it would be better to start with simpler models. If you really want to pursue in that direction there is indeed some automatic tools in different segmentation software allowing you to segment a few slices of your stack and apply the threshold to your whole stack. As you have microCT images the cortical and trabecular bone should appear quite different on the images and this should be feasible. This is a bit complicated to answer in more details as I don't know which licenses you have or what your CT data look like but I hope that helped!

Best,

Narimane

In this case, I highly recommend you use ABAQUS scripting. Because it's more convenient than linking ABAQUS with Matlab, and your run time decreases significantly. In this regard, I suggest you use Micro Manger in ABAQUS to write an ABAQUS script.

Hi ,

You could face this issue due to many reasons, if you have any third party security software or when the components responsible for the installation of Window Update are corrupt. I would suggest you to follow the steps given below.

Method 1: Run the Windows Update troubleshooter.

Follow these steps to run Windows Update troubleshooter.

1.Press Windows key + W and type troubleshooting.

2.Select Troubleshooting.

3.Select View all on the top left corner.

4.Run Windows update troubleshooter.

Move to next method if issues persists.

Method 2: Temporarily disable third party security software.

If you have installed any third party security software like antivirus program/firewall, they can interfere with the download. I would suggest you to temporarily disable and check if you are able to download the update.

Method 3: Reset Windows Update Components.

I would suggest you to Reset Windows Update Components and try installing the update by placing the computer in Clean Boot.

Reset Windows Update Components

Note: This section, method, or task contains steps that tell you how to modify the registry. However, serious problems might occur if you modify the registry incorrectly. Therefore, make sure that you follow these steps carefully. For added protection, back up the registry before you modify it. Then, you can restore the registry if a problem occurs. For more information about how to back up and restore the registry, click the following article number to view the article in the Microsoft Knowledge Base. How to back up and restore the registry in Windows:

You can follow the steps given in the link below for reference.

Troubleshooting problems with installing updates

(All articles are applicable for Windows 10 also)

Hope it helps

Hi Rameshbabu C ,

NASTRAN is a solver and it is used as processor so you cannot compare Ansys and Nastran. Every FEA package has its own pros and cons. Ansys being a very user friendly software it is widely used in academic institutions and in few companies. Aerospace industry mainly use Hyper mesh for meshing work and Patran to preprocess the stress analysis work. Hope this helps

Regards

Paul Vizhalil

Have a look at FEniCS, which is a very powerful library running in all OSes via docker. In Ubuntu you can install the packages directly. Writing the code is in Python, but actually the package wraps it into a C++ and solves very efficiently. All codes are ready for parallel computation. Basically, you write the weak form, what you want to solve, hence, you can use linear or nonlinear material equations without any problem. I use it for multiphysics computations and can only recommend to begin with it.

The community is very helpful as well, if you cannot use a feature or not working as expected, have a look in this blog

I may recommend their tutorial to learn the basics

and also this book in more advanced use of FEniCS in continuum mechanics

This book has an older version of the code, please have a look for the changes in the Errata to this book in my web site

Have fun!

Emek

Good afternoon! If the model solution stops, it is desirable to check the boundary conditions, it is also possible to increase the size of the computational grid. The pile is usually monolithic and the moment of inertia is evenly distributed.

Rossana Fernandes Can you share the two Nastran models (.bdf)?

The answer to your question could be pages. I will suggest few things assuming the definitions of the material models and boundary conditions replicate the given specimens in the test (in case you are trying to validate your model with experimental results).

0- Apply displacement rather than forces.

1- Turn off the force convergence criteria and only keep the displacement.

2- Exclude the extra displacement shapes for SOLID65 (Keyopt 1).

3- Use finer mesh as SOLID65 is mesh dependent.

4- Use bond-slip relation by employing COMBIN39 element.

NOTE: check the value of uniaxial tensile strength as it seems too small for the given concrete strength. You may use ACI or any other code to find that value.

Let me know.

Good luck.

I do not think you will find a specific value for the damping as it depends on the type of skin, thickness, and age of the human. Furthermore, the skin itself will have a varied damping value for the same person from one spot to another (due to the boundary conditions).

What you mean by " I have applied displacement on a surface"

why you aren't applied load and record the displacement ?

I propose my explanation: when you impose displacements in Step 1, the upper sphere is fully constrained, so the solver is able to converge. When, instead, you remove displacements, the sphere is not constrained anymore – in fact, you have no friction at the contact, so the sphere is not prevented to rotate (this happens also you have a vertical load because the program does not know that the sphere is in equilibrium).

If you apply displacements and, then, also the load, such a load has no effect on the lower region because the applied displacements “dominate” the sphere response; in other words, you impose a displacement that behaves like a constraint that cannot be changed by the further load in Step 2, which perhaps will only modify the stress in the upper portion of the sphere.

A final remark: your geometry can be modelled by a symmetric 3D model, or even by an axisymmetric model with a significant reduction of computational time.

I have created 44 Video Lectures for ABAQUS Python Scripting:

suitable for beginners and experts.

The corresponding Code can be found at github:

I am giving ABAQUS Subroutine/UMAT Lectures here:

future lectures will come later this year

I am giving ABAQUS Subroutine/UMAT Lectures here:

future lectures will come later this year

I am giving ABAQUS Subroutine/UMAT Lectures here:

future lectures will come later this year

The cycle load module can completely solve this problem

here is a link:

Manar Naser Can you share your Abaqus model (.inp format)?

Dear Edward,

Abaqus2Matlab is a program for the extraction of ABAQUS analysis results to MATLAB. It provides a link between ABAQUS and MATLAB, and enable you to perform postprocessing of ABAQUS results in MATLAB.

As an alternative, you can extract the displacement result and save them in an excel file. You should have a file containing coordinates and the displacement magnitude. Then, import the data into MATLAB using xlsread function.

To get the length and height, starting from the coordinate that the indenter strikes the pipe and moving in two directions, write a while loop in MATLAB, that saves the coordinate and the corresponding displacement magnitudes in a matrix, while: abs(displacement_magnitude) >= 0

Now you have the coordinates of the displaced nodes and the magnitudes of indentation at each node. the maximum value of displacement gives you the depth, and the maximum - minimum of the displaced node coordinates gives you the length of indentation.

There is no real difference. It is merely the choice of shape functions that differs. In a hierarchical approach the shape functions are set up such that we only add a shape function when increasing the order of the approximation (see p-FEM), while in other versions of FEM all shape functions need to be updated (IGA, SEM, etc.). Each approach has obviously some advantages and drawbacks of their own. But in principle the range of possible applications is the same.

Cheers,

Sascha

Dear Jatin,

Not always does a finer mesh result in a more exact solution. A mesh convergence study should always be performed to guarantee the descending trend of the error as the mesh size gets smaller.

Having this verified, yes, a finer mesh reduces the stiffness of the model. Because FE approximates the the PDE solution by forcing the element into specific modes of displacement which yields a stiffer element. But as the element size decreases, the FE solution converges to the analytical solution of PDE.

Eigenvalue can be physically interpreted as how stiff the structure is in the eigenvector direction. So it follows the same pattern as stiffness.

Useful content to understand tolerances in the link below.

In my openion - [1] Revise boundary conditions [2] Revise input parameters & default co-ordinate axes.

Please, see the following helpful link:

Dear Ana,

Check that if your model fully supports these outputs. Make sure that the analysis type, material properties, and element type are defined in a way to produce these outputs.

Dear Yunus,

After putting some more thoughts on the question we've discussed about in regard to the osseointegration process, and based on what we know from mechanobiology of bone healing process, I speculate that we may have either endochindral or intramembraneous ossification at the existing gap between the implant's thread and its neighboring bone. I think what does matter there is the extent of micromotion, and the relative motion between the implant and bone can determine the outcome of the differentiation process.

On another note, I think there should be some difference between bone healing and osseointegration process, as the former is between the two dynamic structure, but the latter is between a dead, man-made material and a dynamic structure.

Last point: another important consideration is that to differentiate between bone remodeling and bone healing (also osseointegration) process, which unfortunately in many papers this point was not taken into account.

Let's see what other fellow researchers' thoughts are in regard to your interesting question.

Best of luck

Hi Jose Gaxiola,

Suppose that there is a file of nodal relative motions composed of lines with format X, Y, Z, value. I suggest:

1. Create nodeset from nodes of contact surface.

2. In Initial step create temperature predefined field with type Mapped field using created nodeset.

3. Import XYZ Point data from the mentioned file.

4. Request nodal temperature output.

5. Visualize initial nodal temperatures.

Yes, when you install Ansys there is an option that asks you to install a file its name is examples.

This file has a basic USERMAT file.

Hi Mayank Vikhona,

You can use springs with connectivity type: Connect two points. The first point is a node at the bottom of the slab, the second is a ground point (should be fixed).

In case anyone wonders how to do it, I managed it using excel (as I am not really good at writing scripts in Python).

I looked for the closest nodes (between the two surfaces) in the undeformed state by looking at the coordinates of all the nodes of interest. After identified, I obtained the row position where the data of this node was. Next, I obtained the X,Y,Z coordinates of that closest node by indexing.

Once I had those coordinates for the closest node, I also obtained the final coordinates of this node in the deformed state.

Finally, I used the formula for the relative position which I wrote on my question, and that's how I got the relative movement of one node from another one (closest in undeformed state). I repeated the process for all nodes within the surface and that allowed me to track the relative movement of one surface to the other.

Hi Mohammad Ahmed Basri

How did you solve the problem? I am also getting the same nonlinear solver failed to find equilibrium error. My analysis ran for 6 hours and then I got a termination error.

Sarra Bouriachi ,

For what kind of materials?

Linear elastic: Its 3D plot has no special meaning

It is significant for materials with anisotropic and nonlinear properties.

yes, you can do.

Skewness is one of the primary quality measures for a mesh. Skewness determines how close to ideal (i.e., equilateral or equiangular) a face or cell is.

The following table lists the range of skewness values and the corresponding cell quality.

According to the definition of skewness, a value of 0 indicates an equilateral cell (best) and a value of 1 indicates a completely degenerate cell (worst). Degenerate cells (slivers) are characterized by nodes that are nearly coplanar (colinear in 2D). Highly skewed faces and cells are unacceptable because the equations being solved assume that the cells are relatively equilateral/equiangular.

Hi Abolfazl,

If I understand your problem right, I don't think that the number of DOF is the cause of this error. In order to define a ZeroLength element between two nodes, the location of those must be identical. So the location of green and black nodes should be the same. Then you could define p-y, t-z, and q-z materials separately. There is a term of direction in predefined ZeroLength element (-dir), in which you can define a direction for these elements. For example, if you referred to the y-direction as the number of 2, you would write 2 for the elements in the y-direction (-dir 2).

I hope this information can help you in solving the problem.

Dear all,

Probably it is too late to respond to this question. However, I wanted to put my two cents here.

Selecting interior nodes is always a challenge and simply selecting a surface and the nodes attached to a surface is not always going to work. The best way of selecting nodes is selection based on a coordinate system and a geometry referred to that coordinate system (as also suggested by ). In Mechanical (by default) you can select nodes based on a cube geometry which is related to a Cartesian coordinate system (in the background). You can developed algorithms that uses different shapes e.g. cylinder, sphere and etc. based on either Cartesian or Cylindrical coordinate systems.

I developed an ACT sometimes back that enables selecting nodes in ANSYS Mechanical based on different shapes and coordinate systems. This ACT is available in ANSYS app store right now. I put a pdf document here hat shows the ACT capabilities.

Hi all

The examples I provide do increase stiffness. The first example from internal maths, the second example from autoconstraining connecting dependent RBE2 dofs to ground and the last, connecting rotational dofs to ground.

If you look at hour glass modes where element formulation provides internal resonances, there I'd say that you lose stiffness .

Connecting beams to membranes, you will likely deal with all of the above.

To elaborate on the RBE2 - a rigid beam element - using it to connect a beam (6dof) to a solid (3dofs) - you get problems with the 3 rotational dofs. If the dependent dofs are on the solid, these get constrained to ground.

The practical workaround is to add a thin shell onto the solid element and to define it using a material with zero density to avoid spurious resonances in the thin shell.

Just my 2 cents

C

Dear Hajar Rhylane ,

The torsional stiffness for circular cross section is indeed equal to the polar moment of inertia as the cross section is axisymmetric. But as the non-circular cross sections are not axisymmetric their cross section would bulge or warp when the shaft is twisted. If no value is entered for Ixx then ANSYS will compute the torsional stiffness (Ixx) as 'Iyy+Izz' which is correct for circular cross section but not for non-circular cross sections.

The expressions for allowable twist angle, allowable shear stress and torsional stiffness are available in literature. You can input this value in ANSYS as Ixx.

Best Regards,

Shlok

In the ABAQUS program, you need to select one hysteresis models which have similarities to your experimental curve.

Bob Senyange Sir and Victor Krasnoshchekov Sir, thank you for your comments.

The simplest solution you can use in an Abaqus simulation is to place viscous dampers at the boundaries of your model as described by Lysmer and Kuhlemeyer (1969) and make sure your boundaries are not too close to the structure and foundation. This works well in the presence of a homogeneous soil profile and perpendicular/tangential waves with respect to the sides of your model. The next step would be to look at more advanced approaches such as: infinite elements, absorbing layers (e.g., CALM = Calm Absorbing Layer Method), PML (Perfectly Matched Layers) or even hybrid approaches mixing FEM and BEM formulations. However, I do not know if these methods are available in Abaqus software (perhaps using user-defined elements?) nor the particular problems that may arise when calibrating their parameters, numerical stability, etc.

In APDL main menu, you should use Select Enitities, It s a usefull tools for selecting and unselecting everything such as node, kepoints, element and etc.

First use Select>>Entities

then use "Element", "By Num/Pick" , after that select the element you want.

After that you should use again Select>>Entities, then used "Node", "Attached to"

after that select the element. then use Plot>>Nodes.

I made an example and put it in the attached photo, I hope it is useful.

In CAD, there are ways to simplify and reduce your design time. if you use creo, there is a feature called Skelton concept, with Top-Down and Bottom-Up Design. we are using this concept to reduce the design time. every new design we make few master models, linked with family table and reference. this method reducing more than 75% of our design time. even we can have parametric drawings, no need to make separate drawing for the new design.</