Connecting BIM and Design: The Heatherwick Process

ALFONSO MONEDERO: Welcome to Connecting BIM and Design by Heatherwick Studio. I'm Alfonso Monedero head of BIM. So Heatherwick Studio is based in London. It's a design firm. And so I was saying my name is Alfonso Monedero, head of BIM. I joined Heatherwick Studio in 2015 as a BIM technician. I'm an architect by training and I have experience in Spain, where I am from, in India, Chile, and UK from 2015.

So Heatherwick Studio is a team of 250 designers based in Central London, and we create buildings and spaces, landscape, master plans, objects, infrastructure. But we are makers. We value craftsmanship and making and never allow ourselves to get lost in the world of digital technology and management. We prioritize projects with the greatest positive social impact, because our motivation is to design soulful and interesting places which embrace and celebrate the complexities of the real world.

We have a very diverse portfolio, like these projects I'm showing you here. Changi Airport, Coal Drops Yard in London, Google headquarters in London, Bombay Sapphire. And I'm going to show you some of the projects we have in the US. This is the Vessel, the Hudson Yards in New York, in Manhattan. Google Bay View. It's a project we did along with BIG in California. It's a Google headquarters. We have Little Island in New York. It's a park floating on the Hudson River.

And very close by we have the London House with Hyde Park Lane crossing through it. And the last one we have done, it's the Harley Davidson Park in Milwaukee in their headquarters. It's a park to get all the riders together and celebrate. But everything we do starts with a conversation, what we are trying to do, and then we have it in a post-it session. And once we have established the question, we go into the ideation.

And we want very loose ideas coming from anywhere or anyone, because if we can put it in a post-it note, it's simple enough, it's powerful enough, and it's one strong idea. So we all start with a brief and we have to go to the end of this stage, but it's not a very linear route. So we might go with option one. Option two, we go back to square one. Option three, we're getting closer to the final of the idea to the submission, and with every iteration we might be getting closer and closer to the goal.

But at some point, we might even go back to square one. And we can repeat this as many times as needed until we have a good idea. So the path is most like this. It's more a winding road rather than a linear straight highway. And now the same process involve engagement from every designer, from making to digital explorations to rapid prototyping. So all members of the studio are able to use the tools after a brief induction.

This is a statement I have in my team, that the quality of the design should never be hindered by software skills or restrictions. So regardless of the tools that you are using, the quality of the design cannot be affected. And every software has its own limitations, whether we're looking into the more CAD modeling and we could put Rhino or 3D Max, to the more BIM and authoring platforms like Revit. But they also have a strength.

So Rhino, it's a modeling tool for abstract elements and it has a very fluid geometry language, while Revit is really good for documentation, having live updates of changes and synchronization between sheets. So in conventional projects, Revit is the obvious answer. And when we go to more singular projects, we could start going into the more fluid geometry that Rhino has. But why not combine the best of both worlds into the same process?

So I'm going to show you how we do it at the Heatherwick in three different steps. In the strategy, the management, and last, the digital, applying the right tool to achieve your strategy. And we're going to see all of these applied into a project. This project is the London Olympia, which is in the heart of London. It's exhibition space. This is the building in its current form. We were asked by the client to review the master plan because it has a lot of cluttered buildings in between.

So we start removing all the clutter and have a proposal of small interventions all over the place, like a theater, the first theater to be built in London since the 1970s. We have a music venue, offices in the Olympia central, two hotels on the public realm, the canopy, which is the jewel of the project. Here we can see the hotel on the left, the public realm on the center. Central is the tower that you can see in front of you, and G-gate is hidden behind it.

Some images of the access to the public realm, the F&B, and the canopy with the towers behind it. And the towers when it's touching the public realm and some views at night. So the strategy, we said it's the first step in having your digital and design combined together. So we started reviewing about BIM maturity stages, and we can identify these four stages.

Stage zero, we work in CAD, we deliver CAD. Stage one, we work in a BIM model, but we exchange 2D information. Stage two, we're exchanging pieces of the model. And stage three, all the consultants are working on the same model, which is an aspiration, and there is no standard on this. But should we always aim for the highest BIM maturity stage possible? Well, it depends on the region. It's not the same working in the US where we're going for a very mature project or working in other regions like China or South Africa.

It depends on your collaborators. Sometimes the landscape usually falls behind, so depending on who you are collaborating with, you would go for a stage one or a stage two. It also depends on the complexities. Not the same during an airport or during a very small pavilion. And then the scope, whether we're doing a visioning, we're doing a concept, or we're doing the full scope of the project. We need to check on the BIM requirements, what the client actually needs, at what stage.

Because if we're assuming usually during a concept stage, Revit stage one, it's a similar stage, they are asking for LOD 100, which is very lean. And when we review the level of detail, it's very well specified in BIM forum, but the level of information, it's sometimes very vague. So in our strategy, when we have a LOD 100, we just specify the category. This is a door, this is a window, this is a wall system, this is a floor, a roof.

When we go to the level of information to 100, we start identifying different types. Like external door double, internal door single, and so on. And when we go to level of information 300, we start putting instance parameter, like fire rating. This door needs a fire rating of 60. This door doesn't need any fire rating. So with this hierarchy, we can start increasing the level of information required. And we need to make sure that what we are submitting is not overdelivering.

We need to identify what is good to have and what is needed to have and try to submit always a lean model. This way we are reducing the waste produced during the submission and the stages. And also at the beginning, identify your BIM goals, because the BIM goals are endless. So for us, we identify design, documentation, and construction stages, and during the different stages, we have different goals. So for design, we want to improve design communication, we want design flexibility, and we want a sustainable and feasible design.

During documentation, it changed to trust and reliability, design efficiency, and feasibility and support. And during construction, we become design guardianships. We want to make sure that the project, when we hand over to the contractor, is not being changed. We want to continue collaboration and coordination and quality assurance. But this is our goals. You need to define the goals that you need for your company. And as the project progresses, we change the design authoring platform.

When we are in the concept stage, the design is being produced from Rhino. When we jump into schematic design, it's starting to shift into Revit. And we might keep some elements, complicated elements like facade, driven by Rhino. But at the end of the SD, everything should be living natively in Revit. By DD and CD we might have a little bit of packages in Rhino, but the vast majority of the project is in Revit, and construction, the same. And along these two platforms we always have a visualization platform like Unreal.

And the exchange between these two platforms is not that simple, so you need to have a interoperability strategy. So for us, when we are in Rhino, we identify three types of geometry. Blocks, elements that they are repeated all over the project, and we can convert them into native families-- for example, the trees. We substitute Rhino trees for Revit trees, which they are very lean and have all the information needed. But sometimes the geometry is complicated. It's coming from Rhino and we need to create direct shape families, like import geometry inside a loadable family, and then it's placed many places in the project.

On the other side of the spectrum, we can have planar geometry that we can decompose in 2D geometry and data. Like if we are creating a wall, it can be defined by the baseline, the build up, and the height. If we're doing a floor, it could be the outline and the build up. And the information that cannot be easily reconstructed in Revit, we can bring it as a direct shape. We have the singular geometry in Rhino and we just import it with a category and the information the LOI required at that stage. And we have different interoperability levels.

So there is no just one way of bringing the information depending on the project. We can go for level zero, which is just CAD. We can have a level one, which is direct import into Revit for quick documentation. We can reconstruct everything with a level 1.5, and we try to maximize the Revit native geometry. Or we go to a level two or level two plus, which the difference is the requirements from the client.

If there are no requirements for the client, we can go for a lean BIM model with no data touch or a level two plus with data touch that is replying to the employer's information requirements. But there is no good or bad interpretability strategy. The same project, depending on the scope and the stage, could start as a level zero if we're doing a visioning study with master plan. During a stage two or concept, we can go to level one, and then we can jump to a level two plus.

So we have this decision tree to start identifying which area would be more suitable for that project. So now if we think about this strategy that is a little bit theoretical and we try to apply it to the project that we were talking about. So we have the project, and in this project, we had the four different packages happening at different times. So we started with a public realm, central, and hotel during the stage two or concept starting in 2017.

By 2018, we submitted the concept and we jumped SD. The hotel we handed over to the executive architect and they were continuing the national hotel, and we were keeping the ownership of public realm and central. And we had to submit SD a couple of times due to value engineering. By 2021, we were jumping into the DD stage, handing over a big majority of the packages. The G-gate, because it's isolated from the existing building, it was having a different timeline.

So we had to submit the concept stage a couple of times, and then we jumped into SD, and with all the knowledge we were learning from the process, we could have a very short and quick SD and then jump into DD at the same time as the rest of the project. So the first time we were submitting the concept for central hotel and public realm, we had in our scope to deliver IFCs, and then the executive architect would be getting our IFCs and producing all the documentation.

This was before I became DD manager and we didn't have any strategy. And even though we were complying with the requirements and the contract, but the geometry was not good enough. It was not up to the standards that we want in the studio that is defining the design intent. So even though it was compliant on paper, on the reality, it was a not good enough solution. So when we jump into the G-gate, which is the orange building on the back, we submitted a couple of times because we were changing the design.

We were having a very black box at the beginning and we wanted to change it to bring the public space to the front, to the road so that people could see from the outside what was happening on the inside. And we were doing some exercises with the facade just to open the curtain wall, as it has curtains. And these are some images of the black box we had at the beginning on the left and then the more open space on the right.

And then the final one is the one on your right, the one that has been coming to planning. These are the images from it and how it is more open. So as I was saying, when we were doing this part of the G-gate, we were getting a little bit more scope. So we got Revit into our scope so that we were in control of the geometry being sent from Rhino into Revit. So we were putting in place the interoperability strategy and we were able to send it. But there were some mistakes, like the one that is highlighted there, just one large stair being identified as one element.

So when we had to do it for a third time, then we were putting the whole strategy in place. And by doing this, we could leverage the data, and that's when the jump into the next stage, the SD, was really seamless and very efficient. So by having a more complicated or more complete strategy, we were having way better results.

Now we jump into the management. And if you remember the BIM goals, we wanted fast iterations and design communication so we could be able to capture all the design changes fairly quickly by using a platform like Revit, because the target is to document every iteration with a minimum input in the software. To do this, we have the set of Rhino files and we mimic with Revit files. So for every Rhino file, we have a Revit file.

We combine all of them into a documentation file and then we can have the different sets being produced, which is the duplicating and changing the files that are linked. And if we see the deliverables, sometimes, most times we know the deliverables way in advance. We know what we're going to be producing by the end of that stage. So when we hit the 50% concept or 50% stage two, we can start jumping into Revit from the Rhino.

And the first iteration we do is very quick. It's just importing the geometry as quick and dirty as possible because we know the geometry is going to change. That's a fact. So we don't want to invest time into having very native geometry when, in the following week, it's going to change. But by doing this, we can set the cartoon set very early into the process. So we can have all the GAs, as you were seeing, all the enlarged plans, all the details, detail arrangements.

And this is simple, but by doing this and having the file strategy, we can iterate all the different design reviews fairly quickly. So if we're having changes in the tower, just with one click, we can have everything documented and printed. And now let's jump into the digital part, the last part of the presentation. So in 2018, we started establishing a digital strategy, what the studio was going to be doing for the following three years.

So we had this problem about sending from Rhino into Revit. So we developed a magic button in-house. It's no magic nowadays, but back then, all the designers were looking at it with awe. Then we got into [INAUDIBLE] in 2019 to Rhino Inside, and this was a really game changer into the process we have in the studio, all the processes and the workflows. And in 2020, we got introduced to BEAM, a new plugin to send both ways, Rhino, Revit, Revit, Rhino.

So by 2020, we felt that the interoperability issue was fully solved. 2021, we started creating a QA/QC in-house plugin because with this interoperability, the models in Revit can degrade very easily. So it's key to keep an eye on how the models are progressing with the health status. 2022, we started looking into automation and how we can automate the cartoon side. And by 2023, we started automating the reviews.

So these two years were the automation part. The most tedious and time consuming tasks, we were able to automate them. 2024, we're looking into Copilot and bringing intelligence to our models and their processes. So let's just start first with the interoperability issue. So I was saying it was not that simple back in the days, in 2018. We had some information in Rhino. We could send it into Grasshopper. We had to send it into Dynamo, and we had three options.

We could save it into a file, we could put it into the cloud, or we could save it into the computer's memory. Then we can load it into Dynamo and then Dynamo into Revit. This is the full steps but some processes were missing one of these steps. So we tried everything in the market. Visual Arc with IFC, Rhynamo, loading directly into Dynamo. Geometry gym, MantisShrimp, Speckle that is continuing to be developed, and Conveyor.

So the first interop strategy, it was including Dynamo into the equation, how we could get information in Grasshopper, send that information into Dynamo, and then send from Dynamo into Revit. But the game changer of Rhino Inside was removing the need from Dynamo, so it's one less conversion on translation that we had to do. So that was more efficient. Usually when we're doing parametric geometry, the script's composition looks something like this. We have inputs in the sense of geometry and we have parameters that is modifying the geometry.

We create basic geometry. So if we're thinking about the facade, it could be points that then they convert into surfaces. It could be curves. If we're looking into floor plates and walls, or it could be surfaces if they are double curved. Then we are creating complex geometry by extruding or creating meshes and doing solids, and then we bake. So we could branch it for the Revit integration with an interoperability conversion and then the Revit nodes that we have inside Grasshopper.

So with one script, we could be feeding two different branches, one for Rhino and one for Revit. So all the geometry is controlled from the same place. So now we have three routes, the blocks, the planar geometry, and the singular geometry. And we have two methods. We use BEAM for the non-scripters, for the people that they are more based in Rhino Revit, and we have Rhino Inside, which is more advanced, but you need computational design skills.

So if we jump into the example of how we were doing the public realm, this canopy that it's hosting all the F&B in the project. So when we have this project or any project, we start analyzing all the geometry that we have in there. And each of these elements would be sent through different routes. So we have the roof. We were sending the points and just placing the adaptive components in Revit. We have the structural columns and the structural framing.

We were sending just the line defining the columns and the beams and applying a profile. The trees we were sending just with a location on the type and pasted as native families. Furniture, just direct shape families. Railings, we can send it just with the line and the build up or the type. Lines we were sending with the wall by outline, singular geometry like the columns. That's a direct shape. Stairs. We don't recreate these stairs. We just paste native families. And the floors, they are created with the outline.

So this is a typical Grasshopper definition that we had in-house. This is the one for the canopy. And then just a little bit we have the central Revit, which is curves and points. So the curves were defining arcs and lines so that we could pay columns and beams, and the points would select in the curtain panel the adaptive component and just paste it. So from the very complicated spaghetti of visual scripting, the Revit integration is fairly simple. We could see here the structural framing when pasted in Revit with the route of Grasshopper to system families.

And the canopy with the adaptive component that was placed in these four roofs. If I show you another example of how we apply this strategy, it's on the central facade. This facade looks simple because it's just a zig zag, a pleated facade. But in some areas, we had every single corner with a different angle. So it was not easy to do it with a curtain wall or a curtain system in order to control the position of the millions. So we had to find another method to do it.

If we identify the family intelligence that we have in Revit, we can say that the adaptive component is the most intelligent part because you have the same family and it allows you to have different set outs on different geometry. Loadable families would be the second tier, where you have an editable family with native geometry, and we can have information attached to the family. And in direct shape, we have two options. We can send a direct shape with type, so it's scalable. You have type property data.

And when you have a copy of that element, it works as an instance. So when you schedule, it would say to instance of this type. And then we can have the direct shape by geometry that you can have instance data. You don't have type properties. It's non-scalable and copies are treated as new families. As similar as when we were talking about the BIM maturity stages, that there is no one solution for all, we need to identify here or ask ourselves which of these four types is going to be the most suitable for our elements.

And the adaptive component is the most intelligent but is not always the best solution. So in this case, we were doing the exercise of checking with the adaptive component, and we have the method on the right-hand side of Rhino Inside and BEAM with URSA, in this case. And the type that we can have in the instance and also in the type. So all the adaptive components could be just one type, and then in the instance, we start identifying them.

And we do the same with a loadable family, and the geometry is not that crazy. Once you open the Edit Family, it can be identified as a free form. So it allows you to a little bit of editability, like placing materials or changing the thickness a little bit. You have type properties and you have instance properties. When we send it with a direct shape by type, you can apply materials to the different geometry inside the element, but you cannot change it.

You can modify the material, but you cannot change the material for another one. So we're starting to lose some editability once we are in Revit. We still have type properties and we still have instance properties. And when we go to the direct shape by geometry, we can have materials applied by color cannot be changed once it's in Revit and we can only have instance information. The good thing about doing it in a proper way is that you can schedule it. You can modify the geometry afterwards, and you don't need to re-import it all the time.

So by testing our strategy on the central facade, we were then able to apply this to the G-gate facade using an eight point adaptive component to give us a fully scalable facade that gave us the total area and count of the facade panels. And it was consistent with the quantities and the cost analysis. That's why the G-gate for the stage three was so short, because we had all the know-how from the previous Tower of Central. I'm going to show you the different tools that we have created.

So the first one is BIM Pulse. BIM Pulse consists in a Revit plugin and a BIM Pulse app. We did this with our colleagues at Rio. They're doing another presentation here in a few. And the Revit integration is fairly simple. It's just that button that is sitting there. You establish it at the beginning of the project. You place all the rules that you want. This is done by the BIM manager or the BIM coordinator. And then every time any person in the project is synchronizing, you are saving a JSON file into the location where you placed.

You are sure that all the information that is being submitted, it's always the latest, because with every synchronization, you are creating a new report. Then these reports, they are being read in an app that it's is a web app where we can create all the projects and all the models that have the QA/QC enable are being read here. And we have here all the information that we are extracting from the Revit file.

The beginning, we were doing this manually with Excel spreadsheets and opening every file and looking for all these metrics, so it was really important to automate this process. And then you can do a comparison report with this red, amber, and green traffic light system and identify which areas need to keep an eye on and which mothers are less healthy than the others. By looking at these reports, we can anticipate future problems like corruption or sync times being very slow. The next project or plugin that we did is the AutoDocs.

So as I was showing before, the workflow was very well-established at the beginning. So we had the cartoon set. We have to create it at 50% in stage two with the geometry might be changing, but the important massing movements have already been decided, and we have the model structure. So we need to add in the documentation file. So the previous workflow we had is by using Revit, we could create empty sheets, we could create views manually, we place the views with the help of IdeateApps, and we could control the parameters through Excel with BIMLink.

We were replicating this. So how the app works is that you open the project. We work now with the scope boxes. You select the views that you want to create from this scope box, like sections, elevations, plans, AXOs, and automatically it's placing it on sheets. So the app looks like this. We select the scope boxes on the left. We select the view settings on the views we want to create, like area plans, floor plans, ceiling plans, sections, elevations.

We select the levels that we want to intersect. We could pick just one. It could be many. And the drawing series, which drawing number this one is going to be applied to like the 00-90 or A 100 series for the GAs. And then we can put individual sheets, one view per sheet, or we can put same series together. We click on Run and automatically it's doing it for us.

So it's a very fast way of doing the first round of GAs. Obviously, we need to then review them and modify some things that might not be fully right, but we can also automate the assembly drawings, how these drawings with different views on the same sheet are being placed. We can do this with multiple scope boxes at the same time.

Now we jump into the digital pin-ups. So the pin-ups, similar to the post-it notes that I was showing at the very beginning, it's a core part of how we design. It is not how we decide, but how we review and how we comment on the design we are producing so that, this way, it's part of the design process. We know 2020 was a difficult year and we had to start doing remote work, and we got introduced to this set of tools like Zoom, Teams, Slack, Notion.

You can name them. But one of the key ones we had, it was the Miro. And Miro board was key in having a seamless transition from working 100% in the office to have a remote work from home. So the Miro board, it was used to place images that we were seeing online but also to start placing all the information coming from the Revit models. So it was a place where we could be pasting all the sheets and start commenting on them from 1 to 100 GAs to details and start having these documentation in here. One place to review everything.

So what we did [INAUDIBLE] that automatically it's exporting all the sheets that you have in Revit and paste them into the Miro board. So the app looks like this. So you have a drawing set selected with a amount of views. You get access to Miro when you paste the URL. You select your grouping parameter, like which parameter is defining your series, you put the resolution, and then you upload it.

And the result is that we have all the packages organized as columns and they are sorted by the drawing number. And then the next time you have a new set of drawings, we could replace the existing drawings or you could paste it on different place on your right-hand side and start comparing the previous submission to the new submission. This is by far the tool that we use most in the studio.

Now we're working into the Copilot. So the Copilot app that we have, it's based on two things. So we have the wiki, all the information that we have been created all these years placed into just one location. In this case, it's SharePoint. And we can access, from Revit, all this information. You can search it and start identifying your BIM standards or how to create a wall or how to import information from Rhino into Revit. So in this way, we have all this wiki within Revit and there is no need for anybody to go to the server and start navigating all that.

The information is at hand. And once we have all the information, it was fairly simple to create a Copilot large language model that is feeding from all this information, so it's tailored from us, but can also get connected to the brother Revit manual that we have in the Autodesk website. And what we're working now on is that is retrieving information from the model itself. Not just from the files that we have stored, but also it could be live with information coming from the model. So we can be asking how many sheets do we have in the model. And it starts giving you answers based on the real live data.

So summarizing the tools that we have been creating. It's for creation, the Rhino Inside, and the BIM. Then we jump into documentation with the AutoDocs. And then we have solved the review method. So by doing this Miro app, the digital pin-up, we are incorporating non Revit users like project managers, project leaders, and technical design leaders into BIM. And that has been a very big change in the way we operate.

So how we establish all these connecting BIM on design and applying to a project, it's always, firstly, strategy. It's the most important thing is to have it clear what you're going to be-- what is the goals and what are the different methods, the management, how we apply these strategy in the field into the project. And lastly, it's the digital. Once you have the workflows, you can start identifying areas of improving and automation. So for us, the digital is very important, but it comes last because it's the workflows that are defining which tools you need to apply to your project.

So it's in this order, and we cannot lose the focus that we all have this strategy, management, digital, because at the end, what we have to produce is projects, that they bring emotions to the passersby like joy or delight, because we all want to be this child going into a building and start exploring what is surrounding them. So thank you. This is all the methods that we have in Heatherwick Studio to connect the BIM and design.