1. Shaping the Future A PEM fuel cell ontology to facilitate assembly line generation using a semantic approach: A proof of concept WMG Doctoral Research and Innovation Conference 30 th June - 1 st July Mussawar Ahmad mussawar.ahmad@warwick.ac.uk Supervisors Prof. Robert Harrison Dr. James Meredith Dr. Axel Bindel

2. Outline  Fuel cell background  Problem identification – manufacturing know-how  A proposed solution using knowledge representation  Model description  Test and results  Conclusion  Further work

3. Background  Automation Systems Group – WMG – Automation in manufacturing – Process control – Virtual engineering – Tools developed for virtual commissioning  Partnered with Arcola Energy on Innovate UK – Fuel Cell Manufacturing Project  Collaboration with Tampere University of Technology (TUT)  PhD sponsored by EPSRC and High Speed Sustainable Manufacturing Institute (HSSMI)

4. Fuel Cells – General Types Proton Exchange Membrane

5. PEM - Operation

6. PEM Applications and Types Horizon H-series 101005001000200050001000050000100000 Horizon XP-series Horizon AEROSTACKS Horizon MFCs Air cooled Nuvera Orion Liquid cooled Ballard FCgen 1020ACS Ballard FCgen 1300 Ballard FCvelocity 9SSL Horizon Educational The H-Series stacks are not designed with a specific application in mind Power (W)

7. PEM - Power Single Cell More cells = More voltage = Taller stack Larger surface area = More current = “Fatter” stack Fuel cell stack Find the right balance to meet POWER needs

8. PEM Assembly Diffusion layers Membrane “Sub”-cell Also referred to as an MEA (membrane electrode assembly) Gaskets

9. PEM Assembly Open cathode Stack Closed cathode stack Liquid cooled stack More power More complexity But…there is an underlying commonality! If you can make one, can you make them all?

10. Know-how The Problem Fuel cells are great, but… Lack of hydrogen infrastructure Costly compared to incumbent technologies Material costs Manufacturing costs Assembly Component manufacture Assembly costs: 10-30% [1] of labour Up to 50% of total manufacturing [2, 3] [1] J. L. Nevins and D. E. Whitney, “Concurrent design of product and processes,” McGraw-Hill, New York, 1989 [2] U. Rembold, C. Blume, and R. Dillmann, “Computer- integrated manufacturing technology and systems,” Mar-cel Dekker, New York, 1985 [3] S. S. F. Smith, “Using multiple genetic operators to re-duce premature convergence in genetic assembly plan-ning,” Computers in Industry, Vol. 54, Iss. 1, pp. 35–49, May 2004. Equipment Processes Methods Control Criticality Tolerances Sequence

11. The Proposed Solution Know-howKnowledge Capture StoreReuse Knowledge-baseOntology “an explicit specification of a conceptualization” [4] [4] T. R. Gruber, A translation approach to portable ontology specifications. Knowledge Acquisition, 5(2): 199-200, 1993 [5] Jose L. Martinez Lastra, Ivan M. Delamer, Fernando Ubis; Domain Ontologies for Reasoning Machines in Factory Automation; ISBN: 978-1-936007-01-1, 2010; 138 pages Formally describe a ‘domain’ [5] Extensible Scalable Flexible

12. Ontological model - PPR What is needed? How to put it together? What is being made? Resource Domain Volumes Requirements Cost Process Domain Product Domain Enterprise Domain Customer/ Competition Ontology Semantic rules Mapping Axioms Product Characteristi cs Factory commissioni ng Virtual engineering and commissioning tool Use parametric product CAD to quickly assess what changes may be required on the manufacturing system.

13. Ontological Model  Used Protégé - an ontology editor  Uses a semantic language – Web Ontology Language (OWL)  Extension of Resource Description Framework (RDF)  Queried using SPARQL Protocol and RDF Query Language (SPARQL)  Rules can be written in Semantic Web Rule Language (SWRL)  RDF-based models are RDF triples which semantically describe concepts  Mimics and formalises natural language  Model has classes, hierarchies and relationships  These are used to describe real world concepts SubjectPredicateObject FuelCellhasTypePEMFuelCell hasVariantOpenCathodeCathodeStack OpenCathodeStackhasComponentAnodeFlowFieldPlate hasLiaisonWithAnodeGDL

14. What do the domains look like? Product Domain

15. What do the domains look like? Process Domain

16. What do the domains look like? Resource Domain

17. The bigger picture

18. The even bigger picture What is needed? How to put it together? What is being made? Resource Domain Volumes Requirements Cost Process Domain Product Domain Enterprise Domain Customer/ Competition Ontology Semantic rules Mapping Axioms Product Characteristi cs Factory commissioni ng Virtual engineering and commissioning tool

19. Liaisons and Precedence This method allows the modelling of the PROCESS SEQUENCE and thus the ASSEMBLY EQUIPMENT CONFIGURATION

20. Model Only modelled the relationship between the GDLs and the CCM to test…

21. Testing and Results  Queries are written using SPARQL to test the model  Two tests were carried out Resource Domain Process Domain Product Domain 1. Check that the mappings results in the selection of appropriate equipment 2. Check the model technique for precedence works

22. Query 1 Correctly selected appropriate assembly equipment i.e. Robot + gripper

23. Query 2 Correctly ordered and labelled the liaisons between components

24. Conclusion  The concept has been proved – Equipment can be generated – Sequence model works  Designed to allow the addition of more information in the future  Some progress on building a fuel cell assembly KB  BUT – It’s a time consuming process – Concepts being modelled are simple - unforeseen complexity may need a model redesign

25. Further Work XML File Ontology Resource Domain Process Domain Product Domain What is being made? How to put it together? What is needed? Volumes Requirements Cost Enterprise Domain Customer/ Competition XML File Virtual engineering and commissioning tool Use parametric product CAD to quickly assess what changes may be required on the manufacturing system.

26. Acknowledgement Questions Contact Details: Mussawar Ahmad mussawar.ahmad@warwick.ac.uk