Computational Caving

1676981078223 v2

As an engineer, what do you do when you’re challenged to deliver a swimming pool in a stunning limestone grotto in a constrained site?

Option 1, finding a conveniently located cave, was swiftly ruled out following a not-particularly thorough desk study of the site, so we quickly pivoted to Option 2; creating an artificial cave with a series of vaulted limestone arches buried into the side of a hill, to form a stunning partially buried pool discretely embedded in the landscaping.

The easy solution would have been to create a square concrete box, and internally clad it to create the desired natural form. However, we rejected the structural dishonesty and high embodied carbon of this approach, instead proposing a series of post-tensioned stone arches and stone vaults, arcing across the pool, carefully shaped to create the required clear heights while avoiding the many root protection areas on site. Insulation and waterproofing were then provided to the exterior of the shell, before being buried in landscaping, allowing for a fully conditioned and waterproofed space while fully expressing the structure.

The inspiration was originally developed from a balloon, tied down at discrete locations with string. Similar to the somewhat-more-famous Sagrada Familia, this provided a method to demonstrate a shape where the shell (balloon) was fully in compression, relying only on discrete lines of support, demonstrated by the string. By using this efficient geometry, we were confident that the overall proposal was achievable and required limited material and we could focus on refining the concept with the architect.

Modelling and analysing such a complex form 'traditionally' would prove too time-consuming, so instead we developed a Grasshopper script, letting us generate (and alter) the complex form in Rhino for visualisation, while offering easy transfer to our analysis models. This combined workflow meant that the design and analysis could be developed in parallel, essential for such a complex geometry.

The key challenge was to develop a 'common' geometry for the individual elements to allow for easier fabrication, while also creating the variation and interest required visually.

As with all our grasshopper modelling, the script was broken into a series of blocks, tasked with different elements of the work. Inputs defining the geometry are provided in block one, defining the basic parameters such as the width and height of the arches, but also the inclination of the arches.

The next two blocks define the curves of the arches and the edge walls, creating the skeleton of the framing.

Further blocks add increasing detail, creating shells for the stone vaults and walls, and sweeping the beam profiles, until the final 3D model is produced.

By automating the generation of the model, the design was easily iterable, and could be directly imported into Revit allowing for architectural design development, including acting as the basis for the architectural renders presented to the client. It provided a single model coordinating the structure and architecture, simplifying coordination, that was then used to form the basis of the manufacturing mode, allowing for clear demonstrations of the build-sequence and key interfaces, allowing for rapid progress on site. It was also a good basis for trialling out TwinMotion to produce some structural renders straight from the Revit model.

By using such an efficient shape and avoiding overcladding an RC box, the embodied carbon of the design was significantly less than the baseline, both due to the low embodied carbon of stone, but also due to the more efficient structural form.

In the end, we think our pre-fabricated limestone grotto was a practical solution to the challenge!

Liam Bryant

Liam is an Associate at Webb Yates Engineers, and a member of our internal computational group; a team involved in improving our company-wide efficiency through the development of custom tools and scripts.