Launch of the DFG-funded “PlauSim” project in April 2026

The DFG-funded research project “PlauSim – Plausibility Checking of Simulations with Increased Transferability and Reduced Application Limitations” has started at the Chair of Engineering Design (KTmfk) at FAU. The project focuses on the automated assessment of structural-mechanical FE simulations.
Simulations are an important part of virtual product development. At the same time, FE software is increasingly being used by engineers who do not have in-depth expertise in finite element analysis. PlauSim addresses this challenge by using data-driven methods and deep learning to automatically determine whether an FE simulation is plausible – with the long-term goal of extending this capability to previously unknown simulations. This is intended to prevent incorrect decisions in design and simulation setup and to reduce unnecessary iterations.
Three research questions are at the heart of the three-year project: How can the transferability of plausibility checking to unknown simulations be improved? How can additional simulation knowledge be used to provide comprehensible explanations of the results and recommendations for correcting errors? And how can the amount of labelled training data required be significantly reduced?
To address these questions, a similarity metric for FE simulations will first be developed and validated using expert knowledge. Based on this, machine-learning models and data-preparation methods will be investigated with regard to their transferability. An ontology will link the classification results with formalised knowledge, enabling detailed explanations of plausibility assessments. In parallel, active and semi-supervised learning methods will be investigated to reduce the need for labelled data.
The results will ultimately be integrated into an assistance system and evaluated with simulation experts. The goal is to provide generalised, comprehensible and resource-efficient support for assessing structural-mechanical FE simulations – and thus to take another step towards more broadly applicable virtual validation in product development.