Optimization - Configuration File

Configuration File

Finally, all the specified information can be viewed in the configuration file. Other special settings are also possible here, which are of particular interest to simulation experts.

Together with the stl files, the configuration file is the only input parameter for optimization.

The following settings can be made under User Setting:

onlyFEM

An analysis of the design space is carried out.

Further settings can be made in the FEM area:

solver= Extern CG	Conection to solver: Externer solver from AMendate z.B. AMendateCudaSolver. Uses an integrated conjugated gradient solver.
solverIP=localhost	IP of the external solver, localhost for the same workstation. IP for cloud, for what the matrix is built locally and sent to the calculation unit. Large amounts of data can be moved with a corresponding amount of time.
solverPort=1234	Port which is used to access the CudaSolver. This can be selected arbitrarily, according to the specified value when starting the solver.
eigenThreads=2	Number of threads that can be used to build the matrix. At least one core should always remain free.
solverMaxMemory=10	Definition of the maximum available memory for the stiffness matrix. Corresponds to the resolution ot the model and thus determines the calculation time. With GPU Solving, the GPU RAM normally should not be exceeded (1 GB to 14 GB if necessary with CPU also over 100 GB). We are mainly using Nvidia Quadro P5000 with 16GB Memory. When starting an optimization, an initial mesh is created with a resolution as fine as possible without exceeding solverMaxMemory value. This is driven by the overall part volume including Design and Non-Design spaces. Then the optimization starts with a way lower resolution for the first iterations to drive a fast optimization and reduce the most unnecessary material. With ongoing optimization, the resolution is increased back again up to the finest resolution for the final iterations. As the material volume is already decreased, there is no need for the total amount of elements and the maximum amount of memory will not be used. This is the reason why experts can try to set the value for solverMaxMemory even higher than the available memory.

The following output files can be selected:

export_stl_name_MC_Smooth	Result geometry: stl with MarchingCube smoothing algorithms
export_stl_name_MC_Smooth_intersection	The optimization result is intersected with the design space in each iteration on a voxel basis
export_all_allCases	Each defined export will be exported per loadcase
Stresses
export_ply_name_Stress_RGB	Stresses: ply in color
export_ply_name_Stress	Stresses: ply values for nodes
export_mrc_name_Stress	Stresses: mrc values for nodes
export_ply_name_Stress_Prop	Stresses: ply values for facets
Displacements
export_ply_name_Dis_RGB	Displacements: ply in colour
export_ply_name_Dis	Displacements: ply values for nodes
export_ply_name_Dis_Prop	Displacements: ply values for facets
Result Geometry
export_stl_name_MC_Smooth_intersection intersectionDetail =	The current optimization result is intersected with the design space on a voxel basis in each iteration. 0...3 Setting the resolution on the basis of which the intersection is performed. A good value here is 2.

Summary

Definition of design type, calculation type, optimization goal and symmetry
Definition of output files