optiSLang

from Wikipedia, the free encyclopedia
optiSLang

Ball Dynardo OSL P RGB.jpg
Program screenshot
Basic data

developer Dynardo GmbH
Publishing year 2003
Current  version 7.4.0
(May 2019)
operating system Windows , Linux
programming language C ++ , Python
category CAE program
License proprietary
German speaking No
optiSLang product page

optiSLang is a CAE - Software for editing multidisciplinary tasks of parametric sensitivity analysis , multidisciplinary optimization, reliability and robustness analysis and robust design optimization (RDO) using the stochastic analysis . Existing simulation processes from any CAE programs as well as pre- and post-processors can be connected using a graphic editor via ASCII files and made accessible to a parametric sensitivity study, optimization or stochastic analysis. Optimizations can be done using gradient methods ,Evolutionary algorithms or adaptive response surface methods are carried out.

Methodology

Sensitivity analysis :
Thevariance-based sensitivity analysis usedin optiSLang examines the influence of the input parameters on the system responses and thus classifies sensitive or sensitive parameters of the formulated system. The optimization variables are varied evenly distributed and evaluated independently of one another. In contrast to the local, derivation-based sensitivity analysis, the variance-based sensitivity analysis covers the entire parameter space.

Coefficient of Prognosis (CoP)
The CoP is a model-independent measure for evaluating the model quality and is determined by the equation . is the sum of the squares from the errors of the predictive model, which are obtained by a cross-validation method. Here, the original support points are divided into groups and an approximate solution is then built up from the set of support points reduced by the amount. The quality measure of the model is thus only built up at those points that are not part of the approximate model. Since the prediction error is used instead of an approximate solution, it is possible to apply this method to both regression models and interpolation models.

Metamodel of Optimal Prognosis (MOP):
If unimportant variables from a model are removed, this can improve the prediction quality of an approximate solution. This idea was based on the Metamodel of Optimal Prognosis (MOP) . Here, the optimal input variables and the most suitable substitute model (polynomial approach or moving least squares with a linear or square basis) are determined. Due to its independence and objectivity, the CoP used here is an ideal measure for quantification and thus for comparing different models and different parameter sets.

Multi-target optimization :
The optimal substitute model determined by using CoP and MOP, taking into account the optimal selection of the input variables, can serve as a preconditioner for global optimization strategies, such as evolutionary optimizers, adaptive response surface methods, gradient-based optimizers or biological optimizers, or as a direct single-target optimizer. Several mutually conflicting target functionals can also be considered in order to find meaningful weighting factors for a possible subsequent target optimization in order to develop an optimal design.

Robustness assessment:
In the variance-based robustness assessment, critical model responses are examined. In optiSLang , input variables generated by random principles are generated and the system responses are evaluated using a probability density function. Here the statistical properties of the model response to mean value, standard deviation, quantile values ​​and higher stochastic moments are examined.

Reliability rating:
The reliability or reliability is a measure of the reliability of a result. In the probabilistic safety analysis (PSA) , scattering influences are expressed as random variables and defined using a distribution function with stochastic moments and mutual correlations. The result of the reliability assessment is the complementary reliability ( ) and the failure probability, which can be represented on a logarithmic scale.

Interfaces

optiSLang was developed for a variety of problems, for example mechanical, technical, mathematical, investigative problems and a variety of other questions can be investigated. For this purpose, optiSLang has various interfaces to external computer programs:

History

Since the 1980s, research teams at the Universities of Innsbruck and the Bauhaus University Weimar have been developing algorithms for optimization and reliability analysis in connection with finite element calculations. The result was the Structural Language (SLang) software . In the year 2000, CAE engineers carried out optimizations and robustness assessments in the automotive industry for the first time. Dynardo GmbH was founded in 2001 and in 2003 the SLang- based software optiSlang was launched as an industrial software solution for CAE-based sensitivity analyzes , optimizations , robustness assessments and reliability analyzes . From version optiSLang 4, the software was fundamentally equipped with a new graphical user interface in 2013 and the interfaces to external CAE programs were revised.

Related Links

Individual evidence

  1. ↑ ChangeLog: ANSYS optislang. Dynardo, May 2019, accessed June 24, 2019 .
  2. ↑ a b Product website
  3. ↑ a b Thomas Most, Johannes Will: Sensitivity analysis using the Metamodel of Optimal Prognosis (MOP) . In: Proceedings of WOST . 8, 2011.