Computational Structural Mechanics - Science topic

Dear all,

If we have a job with 2 steps, the first step is static, general and the second one is static, Riks. so how can we apply the imperfection for only the second job? wish your helping!

Victor J. Аdlucky thanks for your reply. Also, does this explain the very high buckling factor fron linear buckling analysis?

Hi all,

I am new to Ansys workbench. I have designed a spiral-like spring looking to simulate for compression, tensile and shear. Any idea of how I could get the set-up(settings) right. From meshing to solving. Any suggestion will be very much appreciated.@

Create entire sketch in modeling software and import it to Ansys Workbench.

I also want to delete particles in DPM model . Load this UDF , I want fluent16.1 to show the particle trajectories. the following error occurs : Error: received a fatal signal (Segmentation fault). Error Object: #f.

Divya Brahmavrathan, as I can see in the image there is you have defined the function, so for that first of all, you need to define the response spectrum (RS) from the define menu in SAP2000. Now, there is two IS 1893 codes available in the drop-down list. I suggest you just go with IS 1893:2016 and now again run the Pushover analysis and go to ATC-40 and modified the function. where select your predefined RS and then there is SF, which is known as a Scale factor, in other words, the multiplier for your RS. you should multiply only g=9.81 value in it and also assure that what spectra are you defined before is it elastic or inelastic. If you are willing to do the fragility analysis in that case you should go with elastic RS. it means the R-factor should be taken like 1.

To ensure better perfomance on your computer the following simulation condition simflags="-P all" could be specidied in the go line. This means that all processors available will be used. If you want to use a smaller number of processors you can substitute "all" with a desired number, e.g. simflags="-P 4".

Dear Aamir:

There are many reasons why various kinds of elements are used. Shall give a simple example. Let us say we would want to mesh a simple 3D structural problem with general purpose solid elements. FE offers us at least the following choices:

(1) Tetrahedral element with four nodes. This is the simplest element possible. It is a linear element. Good if you have no other choice. We use this element when the geometry is so complex that becomes otherwise difficult to mesh. It is a robust element that does not give you problems while solving. It could be beaten to shapes as sharp as a needle and the solver does not complain much except to throw-up a warning or two and continue solving. The disadvantages are it is too stiff in bending and your mesh size should be large. This means, you would consume more memory and solution time. This element does not capture stress gradients adequately. There is also another problem. When the mesh density is high, a single node is shared by too many elements. Now nodal stresses are averages of the nodal contribution of each element. Each nodal contribution is extrapolation from gauss points. We end-up getting unwieldy nodal stresses. But the element centroidal stresses could be still acceptable. So, what do we do? We go to the next better element.

(2) Tetrahedral element with 10 nodes. This element is the same as the above but has mid-side nodes. Good when you import solid models from CAD and create a quick FE mesh. This element could also be used to mesh complex geometries but is a bit finicky. FE uses serendipity (commonly iso-parametric) elements where every element is mapped to a master element. This makes the numerical integration through Gaussian quadrature easier. When we do this, we run into what are called as Jacobians (involves a simple matrix and some basic differentiation). When any element matches well with the master element, the transformation functions are robust. If the master element and the finite element differ greatly, the mapping becomes quite fragile, for example when you have folded elements, kinked edges, etc. FEA does not like bent edges in elements even with mid-side nodes simply because the master element does not have kinked edges. While such problems are almost nonexistent in a linear tetrahedral element, it starts to show-up when you have mid-side nodes. However, since these elements have mid side nodes, they are a lot flexible than item #1. But still, the stress gradients at sharp corners, crack tips, etc. are not adequately captured. The number of nodes and elements and hence the computational time is even larger than item #1. So, what do we do? We go to the next better element.

(3) Tetrahedral element with 10 nodes but mid nodes of a vertex moved close to an apex. These are called as Barsoum’s quarter point (or one-third) element. By moving the mid nodes towards the apex, we bend the stresses sharper this approaching stress singularity in sharp edges better, as in a crack tip. However, this is only a desperate attempt to capture stress singularity. You could either use this element to capture steep stress gradients or use many elements to capture similar effect. Barsoum’s elements just offer us some extra convenience instead of huge mesh density in crack tips. Nothing more and nothing less!

(4) Brick element with 8 nodes called as a hex element due to its six faces. A simple, but excellent element. Creating a good-looking mesh with this element is an art. Quite a flexible and well-behaved element and if you have a hex mesh, enjoy!

(5) Brick element with 20 nodes, called as hex element with mid nodes. Same as above, but with mid nodes. Naturally flexes like a snake, but hogs on computer time. You don’t use it unless you would have to.

(6) Then you have special elements, the same as above, but takes care of shear and volumetric locking. Shear and volumetric locking are easy to understand but have been irritants.

Only the very bests to you!

-Sathya Prasad

A boundary condition or a constraint is missing.

Probably it is the constraint that must be applied between the layers and the matrix.

we use a beam under bending load to proof the euler classic buckling fomula for a slender culomn

If you consider the boundary conditions &

column slimming limit

ofcourse you can do that for other problems.

For brain MRI Images you may consider this dataset.

http://www.med.harvard.edu/AANLIB/

Hello Mr. Walallawita,

there is no need to convert it manually. In the "details" method of "bolt pretension" in mechanical you can choose "adjustment" instead of "load" in the "define by" section.

After that just measure the force reaction on your bolt.

Either way I doubt you need to do a simulation. Calculating the adjustment is quite easy if you are familiar with applied mechanics.

Regards

Gino

Artificial intelligence, Genetic Algorithms (and its variants like Genetic Programming and Gene Expression Programming) are very good tools to use in your research for optimization purpose. I would suggest you to understand the basic concepts first (you can find alot of material regarding this on internet). Also you must have some data in hand before taking a start (data can be yours or secondary data from already published literature). ANN and GEP are the most common techniques used in Civil Engineering and you can goto www.sciencedirect.com and give these keywords.

I have also attached some articles on GEP and ANN which are mostly related to structures.

Hope it helps.

Hello Sandeep,

Could you explain what do you mean by your first statement "To avoid Membrane Stresses, try to constrain 3D Cohesive elements at the edges in the directions perpendicular to symmetry directions"?

I am facing similar problem in one of my simulation.

Thanks in advance.

The irregular objects can be meshed with the help of refinement option.

This can be very helpful:

Dear Researcher,

The span thickness ratio of the beam under consideration should help the researcher determine which theoretical framework is adequate for modelling/describing the cantilever beam. The researcher should also find out the beam theory being implemented in the software in use, and whether the beam theory appropriately defines the cantilever beam problem considering the span thickness ratio.

IMHO; the best option would be to read the Getting-Started-Manual which will teach you this and lots of other things you need to successfully model with FEM.

This seems like a multi-step algorithm. First you need to select the area of interest, which I think you have done using the first *GET command. There is no straight path to selecting the keypoints associated with an area, so you will need to select the lines associated with the area first using LSLA command, and then selecting the keypoints associated with the selected subset of lines using KSLL. So far, seems like it is straightforward.

The tricky part is to sorting or ordering the keypoints in an ascending or descending order. That can be done in a *DO loop and using the *GET command.

First, count the number of selected KP:

*GET,KP_CNT,KP,,COUNT

Next, you can find the the smallest KP number:

*GET,KP_NUM,KP,,NUM,MIN

You can store the KP_NUM in an array for stepping in a loop. Next step is implementing a loop like this:

*DO,CNT,2,KP_CNT

*GET,KP_NUM,KP,KP_NUM,NXTH

!---HERE YOU CAN STORE THE NEW KP_NUM IN AN ARRAY

*ENDDO

Note that:

1- You can directly store the KP numbers in the array through *GET command. You only need to give the correct array location to *GET parameter input.

2- This loop has to be done inside the loop that selects areas individually.

A code like the following may work. I haven't tested that, though.

*GET,A_NUM,AREA,0,COUNT

*DIM,AREA_LIST,ARRAY,A_NUM,5

*DIM,AREA_LIST(1,1),AREA,,NUM,MIN

*DO,CNT,2,A_NUM

*GET,AREA_LIST(CNT,1),AREA,AREA_LIST(CNT-1,1),NXTH

*ENDDO

*DO,CNT,1,A_NUM

ASEL,S,AREA,,AREA_LIST(CNT,1)

LSLA,S

KSLL,S

*GET,KP_NUM,KP,,COUNT !---IT SEEMS THERE CAN ONLY BE FOUR KP PER AREA, SO THIS LINE MAY BE REDUNDANT

*GET,AREA_LIST(CNT,2),KP,,NUM,MIN

*DO,I,2,KP_NUM

*GET,AREA_LIST(CNT,I+1),KP,AREA_LIST(CNT,I),NXTH

*ENDDO

*ENDDO

Hi Hinesh,

Do you know how to specify the hourglass keyword and its parameter in Abaqus explicit?

Thanks,

Chirag Sachdeva, I tried other time steps and I ran into the same problem.

Abhimanyu Kumar, I did make that mistake but I also put in my load in Pa so the units are still consistent. The deformation doesn't make sense. Small deformations would make sense if the units are consistent and the load is small--which I think is this case here.

Farshad Nazari, why wouldn't 3D solid elements not work?

Hello

SnS is tin sulfide. 

The interaction between soil and wall can be modeled in  different ways : 1) representing the soil as lateral springs connected to the wall element which you have shown in your present figure 2) Zero thickness interface element can be provided at the soil-wall interface which will capture both sliding and gaping at the interface. These interface elements could be node to node or surface to node. 

Stiffness of spring may vary with depth depending upon whether soil is clay or sand. In case of clayey soil stiffness may be constant through out the depth. However, in case of sandy soil/granular soil stiffness may increase with depth. 

To begin with, you can take Young's modulus (E) of the soil at that depth as the spring stiffness k. 

In case of interface elements you will have to perform convergence study to check that the sliding at the interface is converging to a constant value and inter-node penetration is within limit. 

To bring thickness of wall into picture you will have to define wall as 2-D (i.e.,area) element. In present case, you have defined wall as 1-D beam element ( i.e., a line element). So you can not show thickness in it.  For further details on interface modelling you can refer the linked article. 

Dear Prashant: 

It is conveniently possible to model the reinforcement (non-prestressed as well as prestressed) in concrete by using one-dimensional (1-D), i.e. axial force transmission truss-type finite element. It could be modeled either through a discrete approach, or embedded approach; however, convenient approach appears to be the smeared reinforcement model in the concrete. 

Because the reinforcing/ prestressing bars are relatively thinner, the 1-D elements can be embedded (smeared) within the concrete, such that only the mechanical properties of such embedded elements correspond to the re-bars'/ tendons' material properties; whereas, rest of the finite elements possess the mechanical properties of the concrete. Thus, the strain compatibility is ensured especially at the interface between the concrete and the rebar contained inside. Specifically, you may model perfect bond of the prestressing tendon with the surrounding concrete, and some calculated prestress may be applied as a predefined field, which will help you model the prestressing tendon in concrete. 

Now, the prestressing tendons could be made of any material that has been chosen for a problem in hand, such as the fiber reinforced polymer (FRP) in your case. The material properties assigned to these embedded elements should pertain to the FRP in your case. I have also seen your definition of material property module (in the snapshot) and feel that it requires a re-look.

Please note that the approach I have explained above is one of the successfully applied approaches, and other approaches are also available and possible to adopt for such modeling of the prestressing FRP tendons in prestressed concrete member. 

Hope this information is helpful in your research work. 

Good luck,

Vasant Matsagar.

make a remote point

The computational homogenization techniques will solve your problem. You can carry out an RVE (representative volume element) analysis to obtain the 6x6 stiffness matrix, then figure out your bulk modulus and shear modulus. If your network is random enough, you might be able to get an equivalent isotropic material. If you want to avoid the difficulty of applying right boundary conditions, you can try SwiftComp, a general-purpose multiscale constitutive modeling code based on the recently discovered mechanics of structure genome. The code can be directly launched in the cloud at https://cdmhub.org/resources/scstandard. Plug ins for various commercial finite element packages are available.

please follow the help file

is your model 3d using solid element or 2d-(using plain stress element or using plain starin element)?

Yield lateral displacement is found from experiment.Usualy cyclic protocol is in term of yield lateral displacement.

i will try to send you abaqus file for one concrete beam column joint but this week i am so busy with final examinations

 Best regards

Hi Deepak

For a Constrained Layer Damping (CLD) to work, one must allow shear sliding between the layers. 

Could a problem be that your BCs somehow constrain such movement between the CLD layers?

You write that you are using a plate element so the cross section should be uniform in rotation, but one never knows. Just a thought. Take a look at the maths in the theory manual and see if it can shed some light and read out also ply motion from your analysis if this is possible.

As a side note. Sandwich plate elements have a tenedency to overestimate the damping effect as the cross section does not rotate uniformly for all CLD configurations.

/C

Don't do this, unless you know exactly what you are doing and how the increased tolerance affects results.

If your analysis converges with higher residual, this usually means that you need more Newton iterations. You can use the solver controls. Try Analysis=discontinuous first; if this does not work, try to change other solver controls. This way is better because it only affects how you get a solution, not what is accepted as solution.

Hadi, it depends upon the procedure you're employing. Bolt pretension load would work only with abaqus standard procedure, while thermal load would work in both standard as well as explicit. It all depends on what exactly you're trying to simulate, and how the modeling has been done.

Probably misunderstanding something: Why not just add a second step? The initial state of the second step will be the final state of the first step.

Thank you. The discussion submitted previously in eng-tips is about Nastran and very general. However I will try to use the connectors.. and then report later here the results. 

Consider it as a beam on elastic foundation and analyse by various methods.Compare the results and publish as conclusion

in riks method LPF will never become constant as the response moves from pre- to at- to post buckling regimes

Mr. Willberg and Mr. Pedrazzi thank you both for your reply.

Hello Steven,

There is a possibility that the elements are penetrating the neighboring surfaces (due to mesh discretization).

There is also a possibility that the adjust=yes definition on the tie constraint definition is adjusting the slave nodes which in turn is distorting the mesh.

Switch off the tie constraint definition, run a datacheck and see if it runs to completion. Its either the tie constraint definition or poor mesh that's causing the problem.

Aren, put a point mass between the two springs, instead of the rigid plate, and check the dynamics

Dear Haider,

Thank you for answering me

I have already read this file. However, they just shown me how to perform eigenvalue buckling analysis. They did not show me how to determine critical buckling load from analysis. In my opinion, In order to get the critical buckling stress or load, we have to multiply the value of the first eigenvalue (obtained from eigenvalue buckling analysis ) to the applied load. However, in imperfection-sensitive structure, the first eigenmode may not characterize the deformation that leads to the lowest buckling load. We need to find other eigenmode. Therefore, we can not obtain the critical buckling load by this way.

Do you have any solution in this case?

Dear Mr. Ankit,

The answer depends on the purpose of your analysis.  If the purpose is to obtain the overall  stress resultant distributions, the thin shll element is enough for that purpose.  However, if the purpose is to get information regarding the stress concentration in the vicinity of the geometric discontinuity (i.e. the sudden change in cross section), the solid element should be used.

actually, it is dynamic stiffness. Kd=F(w)/U(w)，w  denotes frequency

@nawras h. mostafa 

yeah i have been changing time and yes it is indeed a critical factor, yes i really need to apply it gradually to verify what's in the paper and then do my own work on the model. 

the problem that they didn't mention the rate at which load was applied! all they said is that the load increase was using 2 KN increments and then the maximum loading was left for 30 minutes then it was unloaded again. 

do you have any suggestion on loading rate, noting that the load is nothing but a static load that was increased to get the deflection on each level.

The value of f_tol  needs to be corrected, and the default value is 0.05 in ABAQUS,

more reasonable initial crack will be obtained when a smaller f_tol was set in ABAQUS.

I hope this answer can help you.

Eigen frequencies and modes are dependent on the stiffness and mass interaction between nodes. As you know, for a given mesh and no. of eigenvalues requested during extraction, you've only certain no of eigenvalues accurate (this is due to eigensolving technique).

Mass distribution is easy, as many codes utilize lumped mass system.

Stiffness distribution will be tricky as depending upon type of elements, each node will have certain no of stiffness components. So theoretically it's possible to retain in rigid body system with springs, but practically it'll be quite difficult. Just look at stiffness matrix of a simple spring. Each node interacts with the other node (off diagonal terms in stiffness matrix). Similarly in a beam element, each dof will interact with other dof of the same node + the node at the other end. Simple springs generate uncoupled stiffness matrix. If you know a way to account of inter dof interaction, then yes you can do what you want, otherwise no :)

This is really no different than a nonlinear rotor dynamics problem, and Guyan reduction is extremely effective in removing higher order frequencies that push the time step to extremely small values.

The C3D8R have DOF=1,2,3 and the DC3D8 have DOF=11.  Could you just use the C3D8RT elements to have DOF=1,2,3,11?  Then you can specify all the material properties at once and even set DOF=11 to zero for step 1 and then DOF=1,2,3 to zero for step 2 if you wanted.  Is this a possibility?

Hello Zohre Fakharshakeri

pleasecheck the pdf related to your topic

As part of a recent study of steel-plate composite walls (which have a concrete core), I completed a sensitivity analysis of the effect of hourglass controls in LS-DYNA.

We used the Winfrith concrete model and found that hourglass types 4 and 5 reduced hourglass energy by the largest amount (to ~10%) when QM was between 0.01 and 0.20. We were comparing the influence of these parameters to experimentally measured behavior and found that QM = 0.10 was best. Hourglass type 6 with QM = 0.10 gave reasonable results with hourglass energy at about 20%. It should be noted that as displacement increased (especially beyond yield of the section, the hourglass energy increased in our study. If you want more details, see Ch 3 of the linked document.

DR Amjad, i check this link and i found it benefit.

You can use this site , I think will be useful 

Thank you. Can any one please elaborate, how to use higher order governing equation to implement the shear force condition at the free end of a cantilever beam using Differential Quadrature Method (DQM).

Hello Samuel Ayinde

   Your question is kind of surprising. What you are investigating is a standard problem in finite element analysis where you have an irregular mesh with variable material properties.  What you need is a mesh generator to generate the required mesh. By virtue of the rectangular element shapes mesh generation is an elementary problem. Next your FE software should be able to assign different material properties to the elements. Then the assembly of the global stiffness matrix will proceed as usual with each element stiffness matrix being computed from

     K_e =     ∫B^TD_iB d(vol)

where D_i is the D-matrix for the i^th element.   The presence of different material properties does not affect the procedure.  But beware!  The type of mesh you used with narrow rectangular elements will not give accurate results. For accurate solution the elements should be approximately square near the points of load application.  In that case the mesh would have been finer than that the one you used.  THe FE analysis of solids with narrow material inclusions can be quite expensive because of this requirement of square elements. 

dear sir,

Diaphragms are adopted in concrete box girder bridges to transfer loads from bridge decks to bearings. Since the depth of diaphragms normally exceeds the width by two times, they are usually designed as deep beams. However, diaphragms may not be necessary in case bridge bearings are placed directly under the webs because loads in bridge decks can be directly transferred to the bearings based on Jorg Schlaich & Hartmut Scheef (1982). This arrangement suffers from the drawback that changing of bearings during future maintenance operation is more difficult.

1.Initially you have to create a node set (example: "output nodes") 

2.Now, create a display group(DG)

Tools -> Display group -> Nodes -> Node Set -> output nodes and save the DG.

3. To extract the data click on Tools -> query -> Probe values-> Field output (spatial displacement U1/U2/U3) -> ok. Click on Display group and select the group that you have created in the previous step.

4. Click on 'Write to File'. Now the displacement values for the required nodes will be extracted as a report file (.rpt).

Hope this helps.

thank you Ahmad :)

I think you should look up strain rate sensitivity. I wrote down some papers I've cited which might help.  The basic equation is 𝜎=𝑘(d𝜀/dt)^𝑚 where sigma is the flow stress, epsilon dot the strain rate, and m the strain rate sensitivity. Sorry if the eq is ugly.

The flow stress of a material is a function of strain rate. The higher the strain rate, the higher the yield stress. If you do not know the exact relationship of your material, it should be sufficient to know that it will probably yield above it's normal yield point due to the blast. So, you can do what the above review suggested to determine if it is in the elastic range; just know this is now a conservative estimate.

If you want to go further into this, the sensitivity parameter is not simply related to the strain rate or the strain.  I believe Picu talks about this.

Y. Chen et al

Stress–strain behaviour of aluminium alloys at a wide range of strain rates

International Journal of Solids and Structures 46 (2009) 3825–3835

R.C. Picu et al

Strain rate sensitivity of the commercial aluminum alloy AA5182-O

Materials Science and Engineering A 390 (2005) 334–343

Dear Nabeel,

Various methods can be used to solve the problem. Whichever method is chosen, however,  the result should be same. The finite element and finite difference methods are OK, but you need a computer and appropriate software at hand. For quick analysis, one can apply a simpler approach based on the Ritz or Galerkin procedures.

Regards,

SL

Mousafa,

I think any critical state models at the supercritical (wet) zone lead to softening

In Abaqus Modified Cam-Clay is called "Clay plasticity"

Do not forget to specify  elastic properties, and then yield surface

Here is an example

I agree with Djamel Boutagouga

yes my dear Carlos Mendez

Thank you, Abderrachid.

Hi Bimal

Just tabulate the time-displacement data in amplitude option and then assign that in boundary condition as a displacement control.

The msg-file tells you that the solution does not converge because the force is zero, but the displacement increment is not. The most probable cause for this is some unconstrained rigid-body mode (i.e., incorrectly defined boundary conditions).

I wonder what class of material models you are looking for. Damage would be a typical class for which the stiffness decreases with load cycles. Kinematic hardening is not meant to model this effect.

If you are really looking for damage mechanics, here are 2 classical books:

Jean Lemaitre: A Course on Damage Mechanics. Springer (1996)

Jean Lemaitre: Engineering Damage Mechanics. Springer (2005)

For the question 'can prestress be defined in a truss element' - Yes it is possible.

You could apply a 'Predefined field' in the initial step. The field is of the type 'Mechanical', and 'Stress'- Choose S11 for axial stress in Truss element.

Thank you very much

this is not possible in GUI

! first: select all nodes with reaction forces in y-direction

! (e.g. something like nsel,s,loc,y,0 or in GUI); then read APDL-script:

/post26

*get,numbnod,node,,count ! no. of nodes with reaction forces in y-direction

node=0 ! initialize

node=ndnext(node)

rforce,2,node,f,y

*do,i,1,numbnod-1

node=ndnext(node)

rforce,3,node,f,y

add,2,2,3

*enddo

plvar,2

hello dear friend

you should or you can write some codes in matlab and write the exchanging matix but it needs a good ability in coding and finite elemnt

i think its better to resolve your model with new condition and method you want

in this way you can reach to what you want easier and with good result.

So Flac 3D gives good results.

Very easy.Draw the MOHRs circle and calculate  principle stresses graphically or analytically

@Maxim Ilyushkin

Thank you very much. I will have a try as your advice.