For example, if the simulation of punching process didn't predict the tool damage occurred in the real world experiment, The 3D topography of the damaged tool and the real load curve should be informative data that could be used as input file to modify the boundary condition previously defined. Such a function of simulation software should be favorable to industry since people always don't only rely on the accuracy of simulation. A software that could take the experiment results into consideration should be much better than those tell you to just try again and again.
The kind of problem you propose would be an inverse type of problem solving. There are codes who are able to do that kind of studies for specific cases. In some cases, the inverse problem can be resolved directly, in other cases it is solved iteratively as an optimization problem.
The latter way can be done with most codes, if they can be solved in a batch mode. In that case an optimizer program can run the model with varying boundary conditions or other model parameters in order to obtain a specific result, minimize a problem, or optimze a solution. Of course, it is to the user to determine which parameters to change, and which results (the object function) to monitor.
Thank you! @Dirk and @Hojjat
I don't know how to respond separately here, so I will do it together.
You answers are useful in a way, but I realized that the problem may be not that easy as I or we had thought. The thing is the basic method for any simulation software is one cause relate to one effect, I mean it is in a linear way. But if we inverse the process, it isn't necessarily in a linear way. One effect may be induced by several possible cause.
For example, in ideal condition, apply tensile force to a steel rod without surface defects along the center axis, the angle between the fracture surface and the axis will be 45 degree. But from the effect-45 degree, we can't make the conclusion that the rod has no surface defects and the force is applied along the axis. From this point of view, maybe the idea I proposed is unrealistic theoretically.
To narrow the idea, if the cause is a1, a2, a3, a4. the effect is b. Maybe with the solution you proposed, it's possible to deduce a1 on condition that we know all left (a2,a3,a4 and b).
Yes, Hojjat. I guess so. Python and Abaqus are just tools that base on a kind of "real world cause-simulated effect" model, there is not input port for real world effect. Normally we do this manually, there is no algorithm about it embedded in the software.
Thank you anything for your participation.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Mechanics
- Classical Mechanics
- Continuum Mechanics