Hello !
I am using Concrete Damage Plasticity model to model an RC structure. I am doing displacement controlled analysis.
The reaction force output is alright as long as the concrete is in the elastic range but as soon as damage occurs, the reaction force output shows noise.
Have anyone else encountered the same response and how to solve it ?
Hi Ghaffar,
In an explicit analysis including damage or in this case plasticity, oscillations are to be expected. There are two sources for these oscillations. The first is the "mismatch" between the frequency at which you request the output and the stable time increment in the analysis. The stable time increment is related to stress wave propagation speed in an element. You can imagine that the output at a certain integration point is different when the stress wave is at the integration point compared to the situation when the stress wave is not at that point. This will introduce oscillations in the output and there is not a whole lot you can do about that. The standard practice is to filter out the high frequency response, where the frequencies much higher than the eigen frequencies of interest are filtered out. ABAQUS has a couple of filtering options like "Butterworth" that are very suitable for these problems.
Another source of oscillation is the local degradation of the material. Depending on the stress-strain curve you use for the plasticity model, if the degradation is sudden, then dynamic stress distribution will cause high freq. oscillations, which will not damp out if you have no damping in the system. Adding damping is quite a painful process since its not easy to determine the type of damping (alpha/beta..etc.) and the damping factors that will give the correct behavior without dissipating too much energy.
The response that is shown in the figure indicates a rather quasi-static behavior. If that is the case and you have test data to validate the model, I would advice to start filtering the output and comparing it to the test results to determine the filter settings. If you don't have tests to validate the analysis, you can perform an aigen frequency analysis to determine the eigen frequencies of the system. In case of a quasi-static analysis you can look at the first couple of eigen frequencies and use that as a cut-off frequency in the filtering process.
Dear Ghaffar Khan,
There can be many different reasons but the most important ones are: the material constitutive law followed by its parameters and the solver (one can get different results for Newton approach or Riks). Nevertheless your problem description is kinda short so maybe you can share some files if you want to get the proper help.
Hi Ghaffar,
Are you running an explicit or an implicit analysis? In an explicit analysis noise is expected, but in an implicit (standard) analysis not. Maybe a bit more background information on the analysis and on the noise that you are witnessing. Are these high frequency oscillation?
Hello, thank you for your answers.
By noise I mean, the reaction force oscillates. you can find the reaction force graph attached.
I am analyzing a shear wall in ABAQUS/Explicit. I am displacing the top beam and measuring reaction forces along the constrained bottom face.
I am using Concrete Damage Plasticity model for modeling concrete and embedding reinforcing bars in it.
Hi Ghaffar,
In an explicit analysis including damage or in this case plasticity, oscillations are to be expected. There are two sources for these oscillations. The first is the "mismatch" between the frequency at which you request the output and the stable time increment in the analysis. The stable time increment is related to stress wave propagation speed in an element. You can imagine that the output at a certain integration point is different when the stress wave is at the integration point compared to the situation when the stress wave is not at that point. This will introduce oscillations in the output and there is not a whole lot you can do about that. The standard practice is to filter out the high frequency response, where the frequencies much higher than the eigen frequencies of interest are filtered out. ABAQUS has a couple of filtering options like "Butterworth" that are very suitable for these problems.
Another source of oscillation is the local degradation of the material. Depending on the stress-strain curve you use for the plasticity model, if the degradation is sudden, then dynamic stress distribution will cause high freq. oscillations, which will not damp out if you have no damping in the system. Adding damping is quite a painful process since its not easy to determine the type of damping (alpha/beta..etc.) and the damping factors that will give the correct behavior without dissipating too much energy.
The response that is shown in the figure indicates a rather quasi-static behavior. If that is the case and you have test data to validate the model, I would advice to start filtering the output and comparing it to the test results to determine the filter settings. If you don't have tests to validate the analysis, you can perform an aigen frequency analysis to determine the eigen frequencies of the system. In case of a quasi-static analysis you can look at the first couple of eigen frequencies and use that as a cut-off frequency in the filtering process.
Hi Farid,
Thank you for your kind advice.
I shall use a filter to reduce the noise.
- Badges
- Science topic