March 9, 2022
Sophia Bhaumick
Senior Marketing Manager, Arloid Automation

Creating the Digital Twin: A Short Guide

Every building is different – that’s where the Digital Twin comes in. A bespoke virtual replica of your premises made using live, complex data, the Digital Twin is an integral part of our process.

Creating an accurate virtual model of a building is complicated. And yet, it is essential to achieve total HVAC optimisation. From weather data to occupancy rate to the materials used in construction, there are many factors to consider.

Once complete, the Digital Twin must be effectively calibrated using around 300,000 live simulations. During this time, our AI learns about the building’s thermal behaviour and understands how to achieve peak performance in every microzone. The whole process takes around a week, though this can vary depending on the building size and subsequent connection time.

In this article, Vice Chief Technical Officer Sergey Sherbakov explains how we make our building models.

Step One: Assemble Building Information

The optimal thermal settings will vary between buildings. As a result, the first stage in the creation process is to understand the structure itself. The client will be asked to provide detailed building documentation to ensure total accuracy at this initial stage including architectural drawings and information on HVAC infrastructure locations.

For large buildings, this can result in enormous amounts of data to sift through regarding unique parameters and thermal zones. However, the intricacy of the process is fundamental to the performance of our AI, helping us understand where to locate devices and how to split the site into thermal zones.

The Market for AI and the Barriers to Adoption: In Conversation with Craig Melson >

Step Two: Construct Building Model

The modelling engineer then begins to create a building model using highly accurate data about the building’s architecture and geometry. This model will take a wide range of information into account, including the fabric of the building itself, its specific construction materials and even the pipework!

This stage of the process results in an exact replica of the building with which can interact and learn from. This complex construction phase is integral to the success of the Digital Twin. With an eye for detail, the engineer ensures everything is accurate before the model is allowed to progress to the next stage.

Step Three: Additional Data Collected

Prior to calibration, we collect some final and essential information. Weather data, for example, has a huge impact on temperature and user comfort in a building. We use this data to determine the device parameters to manage and monitor. We also collect data on other factors that will influence the model. Occupancy level is another key piece of information, as the number of people in the building at any time can have a huge impact on the temperature of different zones.

It is here we take into account the function of the building as well. For instance, office buildings have entirely different requirements to hospitals which may need to keep certain zones at controlled temperatures for the storage of medicines and vaccines. This important data collection phase enables to understand how to respond to certain zones and environmental states correctly.

Step Four: Calibration

Once the building model has been created, it goes through an initial simulation to check if any minor corrections are required and ensure it is fully prepared to work with Then, the building model is handed over to machine learning engineers.

The Digital Twin undergoes around 300,000 simulation cycles that span a full year, allowing to understand how to optimise HVAC performance through Deep Reinforcement Learning. Each time, the simulation data is compared to actual data and calibrates itself in response. During this process, our AI learns how to manage each microzone effectively in response to a number of external factors. These include temperature, humidity, airflow, energy efficiency, occupancy rate and pollution levels.

The Power of Deep Reinforcement Learning to Transform the Built Environment >

Step Five: The Result

The completed building model enables our AI to process live data related to a range of unique circumstances and situations. As a result, it is able to determine the best way to behave in each microzone and send commands through our gateway to HVAC devices distributed throughout the model, making minute adjustments and alterations as necessary to restore the optimum settings.

This means it can optimise user comfort, at the same time as minimising CO2 emissions and boosting energy efficiency. That is why the Digital Twin is central to the decarbonisation of the built environment.

Learn more about our technology >

Optimise Your Real Estate with Arloid

Achieving total HVAC optimisation is complex process, and our building models play a crucial role in creating the world’s most powerful AI. Perfect replicas, we collect comprehensive data and intricate information about building composition to ensure pinpoint accuracy. Then, through Deep Reinforcement Learning, gathers information on how to react to a range of environmental states, driving savings and lowering emissions. The end result? AI that is providing real-time performance for clients in a range of sectors, and accelerating the decarbonisation of the built environment.

Get in touch to learn more about our technology today!

Start saving today!

    © 2022 Arloid Automation Ltd | 71-75 Shelton Street, Covent Garden, London, United Kingdom

    ISO 27001:2013 certified

    Complete the form below to talk with one of our experts:

      Achieving Net Zero with