Menu
Search Contact

Sensitivity Analysis

Multiple Simulations

causal correlation analysis using bayesian neural networks causal correlation analysis using bayesian neural networks
graph based algorithms can be used for causal assessment and deep analysis graph based algorithms can be used for causal assessment and deep analysis

CAUSALITY

  • Transition to graph based meta models enables considerations of interactions
  • Using conditional probability models
  • Perform conditional correlation analysis

Clustering

  • find similar deformation patterns
  • get quick overview of your data
  • multiple methods available like hierarchical clustering, distance based clustering (e.g. k-means), density based spatial clustering or spectral clustering

clustering of full vehicle deformation pattern clustering of full vehicle deformation pattern
variance analysis on clustered geometry of a dodge ram variance analysis on clustered geometry of a dodge ram

statistics

  • multiple descriptive statistical methods available
  • evaluation over time, spatial or across simulations
  • arbitrary entities from nodes to components or selections

projection

  • Project high dimensional deformation pattern on a 2D map
  • Fast assessment of deviations or bifurcations also over time
  • Use multiple methods (SOFM, t-sne, pca, ica, umap, ...)
  • Map deviations back to the geometry
use dimensionality reduction to find similar events and back deviations back to geometry for berrer understanding (isomap, umap, pca, tsne) use dimensionality reduction to find similar events and back deviations back to geometry for berrer understanding (isomap, umap, pca, tsne)
Sensitivity analysis: results of anova (based on the car body deformation pattern) Sensitivity analysis: results of anova (based on the car body deformation pattern)

Sensitivity

using dimension reduction enables sevaral methods for sensitivity analysis

  • ANOVA
  • Sobol Indices
  • Surrogate Modelling
  • ...

2 Simulations

subtraction of 2 models to find sensitive areas subtraction of 2 models to find sensitive areas

subtraction

  • get a quick overview from your model deviations
  • subtract arbitrary results on arbitrary entities (nodes, elements, pids, clusters, ...)
  • perform model mapping to compare different models

causal chains

  • Use graph based models to perform causal analysis
  • Assessment of causalities by evaluating error propagation chains
  • Event detection to simplify assessment across time steps
Causal analysis using causal deviation chaines; graph models; graph analysis; ml, ai Causal analysis using causal deviation chaines; graph models; graph analysis; ml, ai