optiSLang (Adv. Optimization)

SoS 3


Robustness evaluation in virtual product development

Manufacturing tolerances, material scatter, random load or other stochastic effects cause scattering properties of components or structures that are usually of spatial distribution. To ensure your product quality, to avoid recalls or to fulfill safety requirements, you can consider these random effects by using appropriate statistical models and methods already in the design phase. 

SoS supports your design process by identifying random fields and the visualization of scatter shapes. Thus, you can locate areas with a high probability of failure or variation („hot spots“) and identify the causes of the scatter.





Visualization of statistical quantities on FEM meshes may simplify the evaluation of robustness significantly. Visualization helps in particular if the critical regions are a priori not known or if their positions vary randomly (e.g. in crash simulation, metal forming simulation, etc.). Further, visualization helps to increase the acceptance of statistical results in CAE and production departments.

Statistics on Structures supports a robustness analysis in these ways: 

    • Analyze random  properties of structures
    • Identify consequences of manufacturing tolerances and random loads
    • Inspect statistics directly on the structure
    • Easily detect hot spots and potential failure locations
    • Improve robustness and product quality
    • Understand  the cause of scatter
    • Decompose response scatter using random field parametric
    • Generate  random designs by using the random field parametric
    • Eliminate random noise in consecutive analysis

Application examples


Post processing window

New features in SoS 3 (since SoS 2.x)

    • Flexible and “Easy to use” work flows
    • Improved numerical efficiency
    • Limitations on FEM mesh size for random field identification dramatically relaxed
    • Simulation of random fields

General features

    • Import and export of FEM field data (element data or node data) of multiple samplings
    • Visualization of single design and statistical result values
    • Identification and visualization of the scatter shapes and their amplitudes
    • Script-based re-evaluation with changed input data
    • Batch processing on Linux
    • Treatment of failed designs, statistical outliers and eroded elements (e.g. cracks)

File formats

    • Native formats: SoS binary database, SoS script
    • Scalar parameters (e.g. varied random inputs): CSV, optiSLang
    • FEM meshes and data (e.g. from FEM simulations of parameter studies): LS-Dyna, NASTRAN, STL (AutoForm through LS-Dyna)
    • Various image file formats (e.g. photos of measurements)

ChangeLog: SoS (Statistics on Structures)

Version 3.1.0, Dec 2014
    • FEM kernel now supports linear and quadratic shell and 3D continuum elements. It further optimises memory layout for 1D, 2D and 3D grids (e.g. signals, images, voxels).
    • Visualisation introduces transparent rendering, iso surfaces, cutting planes, improvements to palette, a ruler
    • Random field models:
      – Introduce cross-correlated fields, local error measures
      – New: Field Metamodel of Optimal Prognosis (F-MOP) for sensitivity analysis of field data. Visualises field CoP directly on the FEM mesh
    •  Importer introduces algorithms for treatment of incompatible meshes
      – Project data between the reference mesh and an independently meshed design structure, including small displacements
      – Identify geometric deviations
      – Automatically translate and rotate design meshes into reference coordinate system
    • Improved integration with optiSLang:
      – Generation of field samples based on random field models
      – Automatic evaluation of F-MOP
Version 3.0.2, Sep 2014
    • New: MS Windows: Associate file extensions (script and data base files)
    • New: Allow extraction of minima and maxima from selection in visualisation (“cluster extraction”)
    • New: Visualize relative values in percent (%).
    • Fix: Enforce optiSLang compatible identifiers during export. Allow more characters during import.
    • Fix: Now use the default system file dialogs, improving user experience and performance
    • Fix: “Export scalar data to CSV” dialog returned with undesired error messages in certain situations.
    • Fix: Allow greater resolutions when importing image files. Note: Old database files containing image data will be incompatible.
    • Fix: Stability and performance improvements in visualisation
    • Changes in binary file format.
Version 3.0.1, Feb 2014
    • New: Automate simulation of random fields with optiSLang.
    • New: Export deformed geometries to STL format.
    • New: Extend supported platforms to Linux CentOS 5.4 and above.
    • Fix: MS Windows: Improve performance of graphics engine while interactively changing the camera.
    • Fix: Export of selected eroded elements to optiSLang will not take the interpolated value, but create a missing response value. Hence the eroded element will be ignored in optiSLang’s MOP solver.
    • Fix: Export of scalars to CSV and optiSLang ignored inactive/active designs. Now, only active designs will be exported.
Version 3.0.0, Dec 2013
    • New user interface, new 3D rendering engine
    • Flexible and “Easy to use” work flows
    • Embedded scripting
    • Greatly improved numerical efficiency, including multi-threading
    • Limitations on FEM mesh size for random field identification dramatically relaxed
    • No mesh coarsening any longer required
    • Algorithms for simulation of random fields
    • Detection of geometric deviations between non-matching meshes
    • Import and export of various random field data (as input or as generated data)
    • Image processing