Cantilever with Concrete properties

[laurenboag1]

Hello,

I have typed the following command lines for mofem, however i have encountered an error.

../tools/mofem_part -my_file cantilever_conc.cub -my_nparts 1

mpirun -np 1  ./bone_adaptation -my_file out.h5m  -my_order 2 -my_output_prt 1 -ksp_type fgmres -pc_type lu -pc_factor_mat_solver_package mumps -ksp_atol 1e-12 -ksp_rtol 1e-12 -snes_monitor -snes_type newtonls -snes_linesearch_type basic -snes_max_it 1000 -snes_atol 1e-6 -snes_rtol 1e-8 -ts_type beuler -ts_max_snes_failures 1 -ts_monitor  -ts_dt 0.5 -ts_final_time 40 -my_load_history Cant_load_history.txt -mass_postproc -ts_monitor -ksp_monitor -young_modulus 33000 -poisson_ratio 0.15 -c 0.4 -rho_ref 0.0058 -psi_ref 1.33e-4 -m 1.0095 -n 0.9987 | tee log

I have attached my error message, load file and mesh journal for reference.

I am unsure why this error has occurred.

Kind regards, 

Lauren

[Lukasz Kaczmraczyk]

Lauren,

We can not tell from the error more than something is wrong with convergence. You need to show us more what happens before the error, how the iterations look like. The source could be wrong input parameters, or error with boundary conditions, in particular, that you have not restrained rigid body motion.

Kind regards,

Lukasz

[laurenboag1]

The only restraint I have applied is fully restraining the cantilever at one end as I had previously done for the model with bone properties.

However I have applied a constant load rather than a stepwise function.

Kind regards,

Lauren

[Lukasz Kaczmraczyk]

Hello, 

You can make mode running if you set:

-ts_dt 1e-4 -ts_final_time 1e-3

However, you need to check units, all have to be consistent. What is the dimension of your model, it is in meters, centimeters or millimeters, what is a unit of young modulus, energy, etc.

Regards,

Lukasz

[laurenboag1]

Young modulus and energy are in MPa, model is in mm, and force is in Newtons. I believe this should be okay?

Will the small time step effect my comparison between the models or will it simply be an additional factor to comment on?

Thanks,

Lauren

[Karol Lewandowski]

How did you calculate m and n coefficients? I would suggest to start safely with commonly used values 3 and 2, respectively.

Also, can you send your load history? It should look similar to this: 

[Lauren Boag]

I am now receiving this error when previously the code ran with these parameters.

I do not understand why this will no longer work.

[Karol Lewandowski]

You had to change something, because it's clearly diverging. 
Also, when you want your density to increase without changing the reference energy density, you would have to increase the loading.

Cheers,

[Lauren Boag]

When viewing the results in paraview, I am setting the minimum bound of rho to the ref density of 0.0058 however this is showing almost no optimisation.

From experimenting with the minimum bound I have managed to obtain results which look more accurate when the minimum bound density is 0.035 which is an arbitrary number in relation to my data.

Is it possible I need a smaller mesh size to optimise the density more correctly?

The output of my concrete analysis shows the correct trends but, not the correct optimisation in relation to the reference value.

Would it be easier if I sent you my out.vtk files as it may be easier to see than explain.

Kind regards,

Lauren

[Karol Lewandowski]

What do you mean by 'more accurate' or 'more correctly'? 

By setting the bounds in paraview you are just showing densities that are within the specified threshold.

If you are getting mainly values lower than a reference density, that's ok, it means that for a given loading lower density would be good enough.

If your goal is to increase the density above the reference, then (like I said before) you would need to increase the loading. 

[Lauren Boag]

I mean more accurately matches the pattern in the waffenschmidt paper.

Although the paper uses bone, I thought I was aiming for the concrete optimisation to match the evolution trend?

My current model almost all values are above the ref density. There is very little change occurring between time steps.

[Lauren Boag]

I have changed loading and results are looking much better.

Thank you so much for your help.

Hope you have a good Christmas.

Best wishes,

Lauren