Periodic Reporting for period 1 - BatCAT (Battery Cell Assembly Twin)
Okres sprawozdawczy: 2024-01-01 do 2025-06-30
BatCAT is the project that realizes the manufacturability programme from the BATTERY 2030+ Roadmap, creating a digital twin for battery manufacturing that integrates data-driven and physics-based methods. It develops a cross-chemistry data space for two technologies, (1) Li-ion and Na-ion coin cells and (2) redox flow batteries, addressing a triple challenge in digital manufacturing:
(i) Design,
(ii) Operation
(iii) Trust.
(i) By improved product and process design and optimization, product quality and process efficiency increase. This requires decision support that makes complex decision problems accessible to human decision makers. The digital twin technology from BatCAT provides an interpretable industrial decision support system (IIDSS) based on multicriteria optimization. Surrogate modelling connects the high-level analysis firmly to ground-truth data.
(ii) Process operation and control is improved by acquiring and analysing sensory and operando data at real time, facilitating live interventions within an Industry 5.0 real-time environment. BatCAT follows a rigorous approach to actionable modelling, combining data-driven methods with deductive reasoning based on ontologies and formal methods (answer set programming and BPMN-based model checking) to guarantee a reliable behaviour.
(iii) The approach from BatCAT produces trustworthy models: Machine learning always retains a clearly characterized connection to the ground truth, and any decision support or decision making from inductive reasoning is safeguarded by constraints through formal deductive reasoning. All our models and methods are explainable, and all our data are FAIR and explainable-AI-ready (XAIR). The digital twin is validated in pilot production lines for (I) coin cells and (2) redox flow batteries, proving its transferability across chemistries.
The project is closely connected to the Advanced Materials 2030 Initiative, BIG-MAP and BATTERY 2030+, BEPA, DigiPass CSA, EOSC, EMMC, and the Knowledge Graph Alliance, ensuring a community and industry uptake of the results.
(i) Design,
(ii) Operation
(iii) Trust.
(i) By improved product and process design and optimization, product quality and process efficiency increase. This requires decision support that makes complex decision problems accessible to human decision makers. The digital twin technology from BatCAT provides an interpretable industrial decision support system (IIDSS) based on multicriteria optimization. Surrogate modelling connects the high-level analysis firmly to ground-truth data.
(ii) Process operation and control is improved by acquiring and analysing sensory and operando data at real time, facilitating live interventions within an Industry 5.0 real-time environment. BatCAT follows a rigorous approach to actionable modelling, combining data-driven methods with deductive reasoning based on ontologies and formal methods (answer set programming and BPMN-based model checking) to guarantee a reliable behaviour.
(iii) The approach from BatCAT produces trustworthy models: Machine learning always retains a clearly characterized connection to the ground truth, and any decision support or decision making from inductive reasoning is safeguarded by constraints through formal deductive reasoning. All our models and methods are explainable, and all our data are FAIR and explainable-AI-ready (XAIR). The digital twin is validated in pilot production lines for (I) coin cells and (2) redox flow batteries, proving its transferability across chemistries.
The project is closely connected to the Advanced Materials 2030 Initiative, BIG-MAP and BATTERY 2030+, BEPA, DigiPass CSA, EOSC, EMMC, and the Knowledge Graph Alliance, ensuring a community and industry uptake of the results.
Please confer part B. It is not possible to detail this here, since we are limited to 7 480 characters for the part A summary. This can all be read in part B, and in our deliverables.
During M1–M18, BatCAT pushed battery-manufacturing digital twins beyond the state of the art by joining formal, interoperable knowledge with practical modelling and early pilot-line data. We started a modular DT platform that uses BPMN/DMN workflows and brings uncertainty handling into the process, so evidence and assumptions stay visible when decisions are made. We created an ontology-to-ASP mapping (from the Minimal BatCAT Core Ontologies) that translates ontology terms into logic rules, enabling explainable, constraint-aware reasoning and reuse across the project. DTU set up a Materials Cloud archive space under BATTERY 2030+ and demonstrated dynamic workflow orchestration for materials/AI pipelines, improving data flow and automation.
On modelling, we advanced mesoscale calendaring simulations that link microstructure changes to transport-relevant metrics, and we delivered an electrode-scale Multiphysics model for Li-ion operation both designed to couple into line-level optimization. We also bridged scales by connecting thermophysical property models to CFD-class process tools.
On data and pilots, partners started the DKMS/knowledge base (LinkAhead/CaosDB) and agreed a consortium-wide PID strategy for making data, models and workflows reliably findable and pilot preparation progressed for the two use cases (coin cells and redox-flow), with first in-line sensing at lab scale and equipment acquisition.
Indicative impacts: faster model-to-pilot iteration, traceable decisions, and FAIR/XAIR-ready data others can reuse. Next needs: integrate components on both pilot lines (towards TRL5), stabilize data pipelines and PID/metadata governance, capture KER/IP for exploitation, and keep aligning with community standards
On modelling, we advanced mesoscale calendaring simulations that link microstructure changes to transport-relevant metrics, and we delivered an electrode-scale Multiphysics model for Li-ion operation both designed to couple into line-level optimization. We also bridged scales by connecting thermophysical property models to CFD-class process tools.
On data and pilots, partners started the DKMS/knowledge base (LinkAhead/CaosDB) and agreed a consortium-wide PID strategy for making data, models and workflows reliably findable and pilot preparation progressed for the two use cases (coin cells and redox-flow), with first in-line sensing at lab scale and equipment acquisition.
Indicative impacts: faster model-to-pilot iteration, traceable decisions, and FAIR/XAIR-ready data others can reuse. Next needs: integrate components on both pilot lines (towards TRL5), stabilize data pipelines and PID/metadata governance, capture KER/IP for exploitation, and keep aligning with community standards