algorithmic modeling for Rhino
I'm working now on Millipede, and i want to record or store data from FE output in each optimization step, to use it later for animation showing optimization process. But Record component doesn't give me any result.. It is any other option to store FE data for each step ?
What type of results would you like to store? The component generates results only when needed [in order to conserve cpu resources and memory, for example each optimization step requires only the displacements at nodes so no need to compute stresses etc...]. You need to use the results visualization component and for example link it to some custom C# component that simply clones results on each update or streams them to files.
The direct output of the solver is an FESystem object that contains the data structure of the complete model + results [if computed] and it cannot be cloned as it is reused at each step.
the simplest solution would be to set optimization to 1 step at a time then use the appropriate result or visualization component [depending on what you want to store] and link its output to some C# component that adds a new branch to a tree structure at each step. Then create a component that produces your custom visualization and by selecting different branches from the recorded tree you can have your animation.
Hello, Thank You for respond.
I know i can store geometry after I visualized them.
But I wanted to store FE data to easy use later visualization component to get some particular stress line in points, or to play with Iso Mesh on different Steps of optimization.
If I store each step, only fixed geometry, i loose possibility to change Iso value. And storing Mesh od stress line for 100 Iso steps for 30 steps of optimization or can occupy lot of ram :) But I don't have another option...
You can always generate a mesh that stores the stiffness factors in the color component, and the principal stress vector [need only one of the two vectors since they are orthogonal] in the face normals list [one vector per face], you can use the texture coordinates to store the actual stresses [2 float numbers] etc.. That would be quite compact storage scheme, i don't think there would be a problem even with 100+ meshes which in any case is more than the optimization steps you need on average.