Aluminum profiles play a vital role in industries such as automotive and aerospace, where lightweight, high-strength, and corrosion-resistant materials are crucial for safety and efficiency. Their manufacturing, however, relies on multistage process chains including extrusion, quenching, straightening, bending, and aging each of which introduces uncertainties that affect the final component performance. Traditional optimization approaches address individual steps in isolation, overlooking the interactions across the full chain. This limits predictive accuracy, increases scrap, and constrains sustainable production.

The proposed project introduces a holistic modeling framework that integrates all key forming steps while explicitly considering both reducible uncertainties (e.g., process settings, measurement errors) and irreducible uncertainties (e.g., material variations). By linking detailed finite element simulations with surrogate models and data-driven methods, the project aims to capture uncertainty propagation across the entire chain. This makes it possible to predict critical-to-quality (CTQ) parameters such as springback, wall thickness, energy consumption, and crash energy absorption with greater accuracy and computational efficiency.

Preliminary work has shown that coupling individual process simulations can significantly improve prediction reliability, but computational demands remain a barrier. The project therefore incorporates active-learning-based surrogate modeling to reduce the need for thousands of full-scale simulations, enabling tractable global sensitivity analysis and robust optimization.

The overarching aim is to generalize this approach to realistic engineering practice by:

1. Developing coupled models that bridge white-box physical simulations with grey- and black-box surrogate models.
2. Quantifying the sensitivity of CTQs to interrelated uncertainties across process steps.
3. Integrating sensor data to update and validate predictions dynamically.

All developments will be benchmarked against experimental data obtained from fully realized process chains at TU Dortmund University. The outcome will be a validated modeling framework that ensures robust, sustainable, and energy-efficient production of aluminum profiles, thereby supporting lightweight engineering with quantified reliability under uncertainty.