How parametric design tools can be used to reduce embodied carbon in data centres

On our journey towards zero carbon design in data centres, we have found that early decisions have a significant impact on the embodied carbon on a structure. Data centres are typically very large structures that are regular and repetitive in nature, so even small design decisions are amplified to have a large impact on the structure’s overall embodied carbon. It is, therefore, important to size elements from an early stage in the design to provide data to inform these decisions. Using parametric modelling, a design approach that uses parameters and their relationships to create models that are easily adapted and modified, the structure can not only be modelled quickly buy also updated in real-time. This is important during the early stages of design, when arrangements often change rapidly.

Historically, a mixture of hand calculation and analysis software has been used at the early stages of design to inform preliminary sizes and decision-making. However, these processes are not parametric, meaning it is slower to make changes. Working out key outputs such as embodied carbon would be done manually outside of the software. This means it is not well suited to early-stage optimisation when the design is flexible, and change happens quickly. Therefore, there was an opportunity to develop a tool that can be used for this stage of design.

The steel optimisation design tool aims to influence the early-stage design of a structure by demonstrating how various factors impact key outputs such as steel tonnage, embodied carbon and number of elements. This has been achieved using parametric modelling, giving the user complete control over not only the geometry of the structure but also other factors like loading, deflection criteria, span direction, etc. The design tools take these inputs and optimises each element’s size so that it is as light as possible while still passing both capacity and deflection checks before calculating key outputs.

On one data centre project, the client asked us to explore how steel tonnage (and by extension embodied carbon) could be reduced. We investigated the impact of three variables: span length, load, and deflection criteria. A baseline was established using the project’s geometry of reference, along with a client’s deflection criteria and a typical suspended load. From here, each variable was altered to understand the magnitude of change and what a potentially fully optimised structure could save in terms of tonnage. The diagram below presents these savings. In response to this, the client then relaxed their deflection requirements which allowed us to investigate span lengths in more detail during the concept design stage, given the potential reductions.

The client instructed us to investigate how altering the span length would impact the key outputs as they looked to find the cheapest and most sustainable solution. Throughout the concept design, the geometry of the building would change. Since the parametric modelling script was set up, these changes were easily accommodated, enabling an analysis of the impact of the change to be understood and then optimised. For example, the client changed where columns could be placed internally within the building, changing the available span options. Through the tool, each option was assessed with the key outputs presented to the client so they could get both cost estimates and contractor input. This helped to inform the client which option was the most feasible whilst reducing cost and embodied carbon. The options and outputs are shown below.

In another data centre research project, there was a greater degree of flexibility over where columns could be positioned internally, meaning that there were multiple span options compared to previous uses of the tool. To get a better picture of what span lengths would be best for the research project, a genetic algorithm was introduced to the script that would optimise for both steel tonnage and the number of elements, finding a balance between the two. A genetic algorithm is a process derived from natural selection where a population of solutions is created, and they are scored with the best solutions retained and a new population is created from them. Eventually, a range of optimal solutions will be provided. This presented the client with the best options that they could then assess from a cost and programme perspective to select the optimal solution within the red circle.

The client then asked us to investigate a column-free data hall. This would require roughly a 30m span, so it appeared more economical to investigate the viability of a truss. Using the same principles of optimisation for a whole data centre, another Grasshopper script was set up that could also utilise a genetic algorithm to find the optimal truss solution, providing a balance between depth and tonnage. This enabled us to assess hundreds of truss configurations in minutes, selecting the best solutions to present back to the client and minimising cost and structural depth.

Implementing steel optimisation along with an embodied carbon calculation into a single tool allows us to bring the issue of the environmental impact of the structure to the forefront at the point where the most influential changes to the structure can be made. Currently, the tool has only been used in data centres due to their regular repetitive structure, but the tool can easily be adapted to other building types with more complex geometries when there is an opportunity. The tool is a powerful method of optimising our designs to reduce embodied carbon to achieve net zero.