COMPLAS 2025

Ensuring Model Consistency in Inverse Problems: Application to Industrial Aluminium Quenching

  • Pachnek, Florian (AMAG rolling GmbH)
  • Simon, Peter (AMAG rolling GmbH)
  • Nemetz, Andreas Walter (Johannes Kepler University Linz)
  • Zeman, Klaus (Johannes Kepler University Linz)

Please login to view abstract download link

Modelling and simulation play a crucial role in modern manufacturing processes, thus enabling the optimization of production strategies and product properties while reducing the consumption of material and energy, experimental efforts and costs. In the industrial production of aluminium plates, sophisticated cooling strategies and distortion control during quenching are essential for ensuring highest product quality. Rapid cooling processes are typically modelled using convective boundary conditions, the parameters of which can be identified through inverse techniques [1]. Heat transfer coefficients (HTCs) can be determined from experimentally measured temperature profiles using parameter identification algorithms along with suitable forward models. While inverse methods have been extensively studied in the literature [2], their integration into advanced forward models remains limited. This study presents an inverse quenching model to identify time-dependent HTCs that subsequently are applied as boundary conditions in a thermomechanical finite-element spray quenching model to predict flatness of aluminium plates. The key interface between these numerical models are the HTCs that are obtained from the inverse approach and afterwards used in the forward model. Due to numerical and computational constraints, the inverse model and the forward model often differ from one another in mathematical formulation, discretization methods, parameters and their meanings, solvers, or even in software environments. To ensure consistency, rigorous verification processes are required to guarantee that the HTCs maintain the same physical meaning across models, thereby leading to accurate thermal field predictions. To achieve precise results, the inverse and forward model must be considered together as an integrated framework, often requiring iterative refinement of both models to ensure consistency and reliability in the entire simulation process. The presented approach enhances numerical modelling for the analysis and optimization of industrial rolling and quenching processes.