Machine learning-based constraints for manufacturing-oriented topology optimization of rollbonded cooling plates
Machine learning-based constraints for manufacturing-oriented topology optimization of rollbonded cooling plates
The thermal management of batteries is one of the key challenges of electric mobility in the automotive sector. Lithium-Ion batteries as well as other common battery types operate best in a certain temperature range, as the energy capacity drops outside of the range. Furthermore, aging effects can occur, reducing the battery lifetime. Therefore, a thermal management system is part of the battery pack. Battery cooling plates with integrated channel structure that are passed by a fluid for indirect liquid cooling by forced convection are a common realization of such systems. One of the most innovative and promising technologies for the efficient production of such plates is Rollbonding. The technology is characterized by competitive cost, great design freedom, and flexibility, as it enables the production of channel patterns without the need for a dedicated forming tool for every design. Designing battery cooling plates is challenging because conflicting thermal and hydraulic targets must be considered while ensuring manufacturability. Topology optimization is a well-known technology that can help designing optimal structures. One of the major challenges of topology optimization is the generation of designs that are feasible for the dedicated manufacturing technology. This paper addresses this issue by presenting an optimization strategy that considers the relevant manufacturing constraints based on machine learning models. These models are trained based on simulation data such that parameters like maximum material thinning and minimum channel height can be predicted and consequently considered within the optimization. Furthermore, a 3D parametrization strategy is presented that enables the representation of realistic 3D channel shapes as they result from the manufacturing process. This way, the hydraulic and thermal performance can be predicted with better accuracy compared to classical 2D methods.
Battery cooling, Flow, Forced convection, Manufacturing, Rollbonding, Thermal, Topology optimization
Schewe, Frederik
779acaaa-87af-4b5a-b88f-a9c6f2d2e0f8
Elham, Ali
676043c6-547a-4081-8521-1567885ad41a
Schewe, Frederik
779acaaa-87af-4b5a-b88f-a9c6f2d2e0f8
Elham, Ali
676043c6-547a-4081-8521-1567885ad41a
Schewe, Frederik and Elham, Ali
(2025)
Machine learning-based constraints for manufacturing-oriented topology optimization of rollbonded cooling plates.
Structural and Multidisciplinary Optimization, 69 (1), [5].
(doi:10.1007/s00158-025-04192-8).
Abstract
The thermal management of batteries is one of the key challenges of electric mobility in the automotive sector. Lithium-Ion batteries as well as other common battery types operate best in a certain temperature range, as the energy capacity drops outside of the range. Furthermore, aging effects can occur, reducing the battery lifetime. Therefore, a thermal management system is part of the battery pack. Battery cooling plates with integrated channel structure that are passed by a fluid for indirect liquid cooling by forced convection are a common realization of such systems. One of the most innovative and promising technologies for the efficient production of such plates is Rollbonding. The technology is characterized by competitive cost, great design freedom, and flexibility, as it enables the production of channel patterns without the need for a dedicated forming tool for every design. Designing battery cooling plates is challenging because conflicting thermal and hydraulic targets must be considered while ensuring manufacturability. Topology optimization is a well-known technology that can help designing optimal structures. One of the major challenges of topology optimization is the generation of designs that are feasible for the dedicated manufacturing technology. This paper addresses this issue by presenting an optimization strategy that considers the relevant manufacturing constraints based on machine learning models. These models are trained based on simulation data such that parameters like maximum material thinning and minimum channel height can be predicted and consequently considered within the optimization. Furthermore, a 3D parametrization strategy is presented that enables the representation of realistic 3D channel shapes as they result from the manufacturing process. This way, the hydraulic and thermal performance can be predicted with better accuracy compared to classical 2D methods.
Text
PrePrint_ML-Constraints_Topology_Optimization
- Accepted Manuscript
Restricted to Repository staff only until 9 December 2026.
Request a copy
More information
e-pub ahead of print date: 9 December 2025
Keywords:
Battery cooling, Flow, Forced convection, Manufacturing, Rollbonding, Thermal, Topology optimization
Identifiers
Local EPrints ID: 510602
URI: http://eprints.soton.ac.uk/id/eprint/510602
ISSN: 1615-147X
PURE UUID: 74cd1c72-bf73-493f-9ca1-16e72f89adc0
Catalogue record
Date deposited: 14 Apr 2026 16:32
Last modified: 15 Apr 2026 02:05
Export record
Altmetrics
Contributors
Author:
Frederik Schewe
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics
