05/08/21
This project combined computational fluid dynamics (CFD) with data-driven machine learning (ML) techniques to produce a system that can give detailed, intuitive information about one space in a large complex building.
CFD models flows within a space and can produce representations of the effects of different interventions (heating, cooling, occupancy) in ways that are easy for engineers and building operators to understand. However, attempting a CFD model of an entire modern building is not feasible.
The data available from a modern smart building, however, lends itself to ML approaches that can provide predictive models of the overall system. These predictions of the overall building behaviour can be fed as ‘boundary conditions’ to a CFD model of an individual space of interest to provide the best advantages of both techniques.
The project was aimed at three of the Pitch-In barriers:
Trust and uncertainty – using data from the University of Sheffield’s Diamond Building and its IoT sensor network, this project will ensure that the models produced are validated against a complex but measured real building; this is only possible due to the increased prevalence of IoT sensors. As well as ensuring trust in the model, the intuitive output from the model provides better understanding for building engineers, using IoT data to reduce uncertainty in building management.
Developing IoT and ensuring fitness for purpose – this work demonstrates how to get value from a heterogeneous mix of IoT devices. In particular, it helps target the placement of IoT sensors that could be retrofit into buildings without an existing smart building infrastructure.
Building IoT capacity and collaborations – although Covid-19 limited the opportunities for dissemination and engagement workshops that were planned, the tripartite project has created opportunities for future collaborations between the Universities of Sheffield and Cambridge, and Twin Dynamics Ltd., the industrial partner in the project. Also, we presented the work at various online workshops, thereby raising awareness of our organisations’ capabilities in the IoT domain.
Using data and 3D models from the University of Sheffield, Twin Dynamics developed a CFD model of Workroom 3 – a flexible teaching space in the Diamond building.
The model was designed to take boundary condition data at a number of points and predict resulting thermal behaviour in the room. Their CFD work produced a set of model parameters that allowed a computational model to execute rapidly and produce accurate predictions of the space’s behaviour.
The Cambridge and Sheffield teams used further smart building data to produce a predictive model of the boundary conditions based on predictable external factors, for example weather and timetables. This was then integrated with the rapid CFD models to produce an overall system that can take input parameters and produce predictions for the thermal behaviour of the space in less than 10 seconds – which makes its use viable for real time predictions.
The resulting integrated model can predict thermal behaviour in the space in a number of different scenarios and for a reasonable range of boundary conditions. This has demonstrated not only the potential for the mixed-fidelity approach, but also developed some practical workflows for developing each of the components.
Since the models have to be developed uniquely for any given space, it is these workflows and experience which is the most transferable and generally useful result from the project. That it was completed within the short timeframe of this project is a good demonstration of the potential for the technique to be applied realistically in commercial settings.
Twin Dynamics is intending to file a patent application based on the technologies developed during and in support of this project.
The work was presented at both the Pitch-In seminar series, and the Sheffield IoT Meetup, each of which had national audiences, as the pandemic-enforced adoption of an online format effectively removed the local attendance limitation.
A follow-on project for Twin Dynamics to expand the modelling to humidity and moisture droplets – particularly, tracing those produced by room occupants, a phenomenon whose significance has been brought home by the Covid-19 and its modes of transmission – has been funded by the Innovate UK Sustainable Innovation Fund (value of grant: £98,294).
The partners have been awarded a Sheffield Innovation Partnership (SIP) grant to help develop the opportunities created by the project.
As described above, it is intended that the case study form the basis of an academic publication, and also that it be continued as a demonstrator project within the University of Sheffield.
The funded follow-on project will expand the functionality and applicability of the work, as would the more extensive project currently being reviewed.
Although there were some bureaucratic hurdles, the sharing of data with the project partners worked well. Hopefully the provision of data-release agreements and meta-data can serve as a basis to speed any future work with campus or building data and external partners.
Working together on a deliberately mixed-modelling project has given all the partners an appreciation for the benefits of combining approaches to complex system modelling, and has encouraged a move away from a competitive approach to the application of modelling techniques.
The Covid situation was unanticipated, and some parts of the project would have been much easier if we could have simply met together in some of the spaces being discussed and seen things in person. Nonetheless, we were able to communicate sufficient information to achieve a shared understanding of the spaces, but we certainly have a better understanding of the complexities of communicating the interactions of different building systems that aren’t always apparent from documents such as engineering drawings or floorplans.
We were a little cautious about using shared data and file repositories due to a range of concerns about IP and proprietary data. A way to share documents, data, and code between partners but with sufficient security and ownership guarantees to satisfy all the stakeholders (not just the project partners, but their parent organisations and/or contractors) would have aided things at several points.
This project provided exactly the ‘pump priming’ that we had hoped for – funding a proof-of-concept case study which we have used successfully to apply for follow-on research support in the form of development funding from Innovate UK.
The project was genuinely collaborative from the get-go, with all partners benefiting both from what was achieved and from successful follow-on. This highlights the importance of genuine co-creation.
Dr Ramsay Taylor – University of Sheffield