Challenge

Finite Element Analysis is a mass business at automotive OEMs with huge turnovers due to automated execution with help of Simulation Data Management Systems (SDMS). To assure simulation quality standards for (purely) simulation-based product approval requests quick expertise of what parts/events are in causal relations to other parts/events.

The causal in-depth analysis of how and why components deform and move as they do, still needs high-level experts which is not available at the necessary scale.

In case of a acceleration sensor the main question is, if a specific control task can be resolved by a given sensor set? If no, why are the sensor signal pattern like they are and which parts form the signal pattern?

Solution

Up to now in-depth analysis is still a high level expert task and has to be done manually by analyzing time varying interactions in animated post-processing. Nevertheless, it is time consuming, subjective and also depends on the daily constitution of the engineer.

The Andata approach gives an objective, physically based automation of in-depth analysis by automated modelling of causal relations.

Applications

• Causality Studies: Which components have an influence on another component (e.g. sensor) or on an event (e.g. head injury) in a certain load case?
• Design Target Evaluations: Is the occurring chain of effects in the load case intended in this way? The designer can, for example, check whether the deformation flow is taking place as “planned” or investigate deviations
• Estimation of Uncertainties: Unsafe events occur (e.g. predetermined fractures, frictional contact, etc.). Do they have an influence on a matter under consideration? Shouldn’t these partial results be used as a basis for decision-making?
• Sensitivity Analysis: What effects do small changes in the simulation have on the relevant components?
• Performance Comparision: Which model deals better with a certain state of affairs? Evaluation of causal chains with regard to defined criteria and goals. Development status, vehicle variant, vehicle model or cross-load case comparison
• Metamodel: The interaction structure can be interpreted as a meta model and can be used for inter- or extrapolation. E.g., how strengthened or weakened some metal sheet has to be to reach a significant change in the causal chains.

References

N. Stalanich, T. Hinterdorfer, D. Lämmerhofer, A. Kuhn, R. Garcia Gomez, D. Böhmländer: Identification and Analyzing of Causal Chains in CAE, NAFEMS Artificial Intelligence und Machine Learning in der CAE-basierten Simulation, Munich, Oktober 2023