Rhino + Forma: More Forms, Less Carbon

KEVIN WALSH: My name is Kevin Walsh. I'm an architect at Nikken Sekkei in Tokyo. And I'm joined by Ellis Herman, Product Manager with Autodesk. To begin, I want to take you back to almost a year ago when I presented at AU for the first time. That's my co-speaker, Chris, on the right. We spoke about Rhino and Forma and how to get them to work together.

Now, Chris was responsible for a plugin that allowed geometry to be moved from Rhino to Forma, and that was the main subject of our talk. I started out by explaining who I work for, a company called Nikken Sekkei in Tokyo. I explained a bit about our history, our size, and the kind of international work we do all over Asia and the Middle East. I also touched on the work I do, which is mainly focused on early stage design.

And you can see Chris here is not saying much, but he seems kind of interested. He used to be an architect, so maybe this is still kind of interesting for him. But when I went on to start explaining Nikken Sekkei's approach to decarbonization, you might notice Chris's body language is changing slightly. He looks like he's starting to lose interest here.

In fact, at this point right here where I'm talking about carbon in the design process, he looks out at the audience. I can almost read his thoughts. "Why is Kevin talking about this?" The thing is that the Rhino Forma plugin we're supposed to be demonstrating had almost nothing to do with carbon.

I was getting really detailed, talking about when in the design process, decisions affecting carbon are made. What kind of decisions are quantitative versus qualitative? Why it's important to maybe prioritize upfront carbon reduction embodied carbon versus operational carbon. And Chris is getting really bored.

Why was I talking about carbon so much? One of the things I wanted to show in my talk is that architects are visual and need visual feedback to understand the design space intuitively during design. And at the end of my talk, I had a question for Chris and for Autodesk. But first I pointed out that the Rhino plugin we were demonstrating didn't work for lots of analyzes in Forma, like this operational energy analysis.

My question was, is there any way architects can access some kind of embodied carbon analysis for early stage design in a way that's visually rich and in a way that accommodates various geometries from various software? So in his response right here, you can see Chris is actually announcing that an embodied carbon analysis tool will soon be launching in Forma. But he neglected to say whether I would be able to access it with Rhino geometry.

The problem was with many types of analysis in Forma, when we imported something from Rhino, we got an unsupported element message like this. Architects and designers, we grab inspiration, particularly geometry and form, from everywhere. And we need to be able to think about all of these kind of diverse forms in terms of carbon as well as GFA, cost, function, and any of the other metrics we use to evaluate our architecture.

So as I finished my class at AU last year, I knew that if an embodied carbon tool were actually released by the time this year came around, I had basically talked myself into doing another AU class on this subject, if I could. So in many ways, this year's discussion is a continuation of what I started last year.

It also turned out that the day before my talk, I met Ellis, who may be the person most responsible for the embodied carbon tool in Forma that Chris announced last year. So, Ellis, maybe you can tell us about yourself and your work on embodied carbon.

ELLIS HERMAN: I would love to. Thanks for introducing me. So I started at Spacemaker about six years ago, and then joined Autodesk through the acquisition a couple of years later. And through that time, I have worked on the solar panel analysis on our microclimate analysis, and now on our embodied carbon analysis.

And basically through these three things, I've been motivated by the fact that these super early phases of design were the ones with the most potential for impact. This is when you can still make those really fundamental, high level changes to designs for cheap. And also, I was motivated by the challenges of getting architects to incorporate a new tool into a part of the design process that generally does not have that much time or money allocated for it.

So very quickly, a little basic background on Forma. It is a fully cloud-based tool that is focused on early stage planning, and it lets you quickly and easily set up a fully geolocated project with a lot of contextual data that comes for free with that geolocation. And maybe most importantly, most of you probably already have access to it because it's included in the Autodesk AEC collection.

So with these easy to use design tools with all of this contextual data and with this easy to understand analysis suite, Forma tries to bring knowledge of those outcomes that Kevin mentioned as early in the design process as possible. So it provides sun, daylight noise, microclimate, area metrics. Right now, I'm going to focus quickly on our wind analysis because I think it illustrates some of the value we think remains on the table for embodied carbon.

So our wind analysis is a full computational fluid dynamics wind simulation. So it doesn't let you get into the weeds as much as the software that experts would use. But that's not who it's for. And we don't think that it's going to replace those experts. What it does do is allow any architect with essentially no technical training to make informed decisions and to iterate quickly without those long feedback loops with specialists, and to show up to the conversations with those experts with a vocabulary that you need and proposals that are already in good shape.

So here's the embodied carbon analysis. You start out by choosing one or multiple buildings. You choose a program, a couple of envelope parameters, cladding material and window to wall ratio, and then a primary structural system.

So basically, this analysis only takes a couple of clicks to get started. You can see here it runs in just a few seconds. And it lives alongside Forma's other analysis. So it's sun, wind, area metrics. So that you're able to understand the trade-offs between all of these multidimensional aspects of early stage design.

And here, the focus, especially for this first implementation of the analysis, is really on accessibility. The tool is only a couple of months old. Forma itself is only about a year old. And we're growing and improving fast. But we wanted to start out with something that was accessible to all early stage designers, regardless of carbon expertise, and then add that complexity and customizability from there.

So again, you're hearing me say accessible a lot, and that really was our first goal here. We found that most of the embodied carbon tools available in the market today were focused on people with more carbon expertise. Kevin, if you go to the next slide here. Secondly, we really wanted a tool that started the conversation between those early stage designers who generally did not have that carbon expertise and the carbon experts who do get deeper into the weeds as the project moves along.

And thirdly, we thought this tool needed to enable goal setting from day one to give whole building carbon results that let you understand where you're starting out, and then understand as you move forward how those decisions are affecting or compare to that baseline you started with.

KEVIN WALSH: So this tool that Ellis and his team have been working on, it goes a long way to addressing one of the needs I asked Chris for last year-- making an early stage tool available that evaluates specific forms created by the user in terms of embodied carbon.

However, a big problem is that if we try to import our designs from other software like Rhino, we still can't actually use this tool. Currently, the only way to use this tool is to model a building directly in Forma, which maybe not everybody wants to do. For example, in my company, if we're already doing the modeling in other software, we don't want to remodel again in Forma, even for simple volumes.

If I was to use the Rhino Forma plugin, which I demonstrated last year, to bring geometry into Forma, the embodied carbon tool just doesn't work. We get this generic message down here. So I want to go back in time again now to April this year. I still had it in the back of my mind those questions that I had left hanging at the end of my talk last year. So when the call for proposals for AU opened up, I started thinking about them again.

This is the proposal submitted, and it contains what I want to cover today. Basically, architects have the ability to make a wide variety of forms, but estimating carbon is still a complex and slow process that doesn't really happen fast enough or early enough in the design process.

And I've outlined here what I hope you all will know or understand by the end of the class. Things like how to prepare data in Rhino. Good ways to iterate designs with Rhino and format in terms of embodied carbon. And especially I want you all to know how fast we can do this now. But I just want to tell you a little bit about my second sentence here, where I describe what I'll be doing.

It turned out this phrase, "novel workflow," ended up being a bit of a problem for me. To be honest, I wasn't particularly confident that this proposal would actually be accepted, so I didn't fully consider how difficult it might be to actually execute this workflow. The problem arises because the embodied carbon tool only works for a specific type of geometry that at the moment can only be made natively in Forma. This is a geometry type called a basic building. A basic building.

So when I was submitting the proposal, I emailed Ellis to see if he'd be interested in presenting with me, and this is his email. He immediately pointed out that the embodied carbon tool only works on basic buildings, so maybe he understood the challenge in a way I did not. But anyway, my proposal did end up being accepted.

So this is the novel workflow in question. I made this diagram as I was starting to grapple with how I could get this Rhino to Forma workflow, working with embodied carbon analysis and avoiding any kind of import error messages. So the steps, theoretically, are make the geometry in Rhino, and then find some way to transfer it to Forma, and then run the embodied carbon analysis.

So obviously, the transfer to Forma part is what really needed to figure out here. I tried several methods, and I'll just jump through them now. First, the Rhino Forma plugin that I discussed last year. It didn't work. As I already demonstrated, it brings in a mesh which is converted to a generic object which can't be used.

I also experimented with third-party tools like Speckle and ShapeDiver, but they had the same issue. They brought in a mesh that couldn't be used with the embodied carbon tool. I actually had high hopes for Dynamo because I realized there actually was a Dynamo plugin for Forma, but it seemed that the only nodes available for it were for analysis and for data extraction, and nothing really useful for creating geometry in Forma. So that was a dead end, too.

So my final hope was accessing geometry creation through the former API. But I'm not a programmer, I'm an architect, and the API documentation and getting started with the API-- there were really beyond the amount of time and really the skill that I could throw at this. So I had a problem.

To describe the problem in a bit more detail is this idea of 3D versus 2.5D. So in 3D, geometry is defined by points in three dimensions. This is how Rhino defines geometry, with points connected to create lines and faces and closed meshes and poly surfaces.

But other software defined geometry in different ways. For example, Revit uses levels to represent floor levels, and geometry is often represented by two dimensional footprints like a floor plate or a wall base. A wall in Revit is defined by a polyline, basically, and a height, and then maybe the relationship of the top and the bottom to the different levels. So it's not fully 3D, but it's not really 2D either. So this is sometimes called 2.5D. And Forma also uses this 2.5D and some of its geometry types.

This approach is close to how buildings are usually built with repeated floors, so it can be pretty handy. The disadvantage is the difficulty of representing more complex geometry, like the kind of geometry we often create with Rhino. For the embodied carbon tool, we can't use these imported meshes because the database that the embodied carbon tool uses to compare our designs with called C-Scale requires specific information that tools like Rhino typically do not embed with geometry. So maybe, Ellis, you can tell us, what is C-Scale?

ELLIS HERMAN: Sure. I would love to. So C-Scale is a whole-life carbon calculation engine, and it estimates emissions from the construction, renovation-- the operation of buildings. And it uses a large database of fully designed, fully built buildings to predict the material quantities from the very early stages of design. And this was built by the C-Scale team. So that's actually an external team from Autodesk that we've been partnered with over the last year.

KEVIN WALSH: So at the risk of oversimplifying, I'm going to say it uses buildings similar to yours to predict an estimate of your building's carbon footprint by measuring the degree of similarity, basically. So I think crucially, rather than measuring quantities, it compares very high level statistics about your design with similar statistics for buildings whose embodied carbon has been accurately measured.

And if you go to the website for the Epic tool, which is a way to access the database directly, you can see that a lot of the input information is related to floor area and number of floors, facade area, window to wall ratio, that kind of thing. So if we were to export a simple blobby mesh from Rhino, we're not going to have that kind of information, and it's not going to be possible to use it in the embodied carbon tool. So you can get these great analysis at the Epic website, and it's very useful.

ELLIS HERMAN: Yeah. And essentially, the information that you've pointed out here, project location, basic structural system, primary use, floor area, all of that stuff, are the inputs to the predictive machine learning model. And what that means is that these things-- location, structural system-- are the features that have been pulled out of all of the buildings in that training data. The features that the C-Scale team has found to be most predictive of a structural bill of materials from these very early, early days of design.

And the value of a predictive machine learning model in this case is that it helps to fill in this huge space between an early stage massing model and a final bill of materials. So when you say that you have a five story, 10,000 square foot residential building in Cambridge, Massachusetts with this structural system and this envelope system, we can use all of these real-life bills of material to help complete that picture.

And then, as you fill in more details about your building, our longer term goal is that that result starts to reflect more and more the specific case that you're designing instead of needing to learn from how it's been done before.

KEVIN WALSH: So that's the background of the problem. And I was starting to understand the technical challenge. But it took me about a month to get this far to understand all of that. And to be honest, I was actually running out of time to solve this problem. I tried to arrange a call with Ellis and someone I knew in the former team at Autodesk.

So I was calling Oslo and Boston and trying to coordinate these different people's schedules, and they were able to put me in touch with a guy called Hobart. And Hobart was working on the Dynamo for Forma plugin, and he had just finished work on a basic building node. So this turned out to be the missing piece I had been waiting for.

So the final workflow transformed from this into this. So the first step, making our geometry in Rhino and Grasshopper and then transferring it to Forma, becomes a process of preparing our geometry in Grasshopper, converting it to point and height data, and writing it to a CSV file, which is then read by Dynamo and converted into a basic building by running the Dynamo Player in format.

And then step three is running the embodied carbon tool in Forma. And we can organize it and compare data using ChatGPT if necessary. So now I'm just going to show a quick demo of how we can use this workflow for a specific use case.

Often at the beginning of a project, we need to find where is the sweet spot in terms of embodied carbon emissions and massing? So this is a fundamental task that we take to see what is the optimum building height for a given GFA So in this demo, I'm going to test four different volumes that all are the same GFA, and I'll export them one by one into Forma as separate proposals.

So let's start by creating the geometry in Rhino, using a Grasshopper script that checks the GFA. It's a basic script that cuts floor plates at specific levels and then sums up the floor area to ensure we have the correct GFA. And then these sliders here will change the building height and the building footprint to automatically balance the GFA areas.

Now you can see this window at the bottom right. That's showing that the Grasshopper script is automatically writing the data to a CSV file. And this CSV file is going to be read by Dynamo in the next step. So then in Forma, I load the Dynamo Player into the project. This can take a minute, but you only need to do it once per project.

And what I use this for is to call an open Revit or a Dynamo file on another screen. So once this is up and running, I just hit Run. After I zoom in to the site, I just hit Run here. And what it's doing now is it's running the Dynamo Player separately. And then you can see a preview of the geometry there. And I just hit Add, and it will add the geometry into the formal proposal.

And then I just need to go through the steps that Ellis already showed. So what I'm doing here is just adding the embodied carbon tool. I think maybe by the time this video is on the AU website, this embodied carbon tool might actually be embedded in the Analysis bar.

So I'm just going to set these building parameters that are necessary-- building program, facade types, and structure system. And once that's all done, I should be able to run the analysis here. And you can see it's quite fast. And I'll just grab a screenshot of this. And I'm going to create a new proposal, and then go back to Rhino and adjust the form so that I have a different height to GFA ratio.

So here I'm back in Rhino. I just quickly changed that slider there. I go back to Forma, run the Dynamo Player once again, and just hit Add once I've confirmed that the preview is correct. So this is all real-time. You can see it's quite fast. It takes less than a minute to move from Rhino into an embodied carbon result. So the one pain point maybe that still remains is that we have to manually add these building parameters. But still, it's pretty fast.

So there we take another screenshot maybe of this. And I'm just going to go very quickly through the next two options. This is sped up quite a bit. So just two more options. This is the tallest option, and there's the embodied carbon result. So I'm grabbing screenshots of this as I'm going through, and I've made a GPT to read these screenshots and generate a comparison report. This is an open GPT you're welcome to use.

I think the Compare tool might be active by the time this video is available. But for now, I've found this step was helpful. So we can just add in our screenshots, and GPT will spit out the results and then make a quick graph. So we can see that maybe proposal 2 is actually around the optimum height to footprint area ratio. So this is quite interesting that we can get these kind of results so fast.

So I just want to talk about one kind of extreme case and get Ellis's input on it. This is a building that my company, Nikken Sekkei, completed last year in Dubai. It's called One Za'abeel. The design features two towers with a horizontal tower in between that cantilevers out to one side. The central link of this structure, it actually holds the record for the longest cantilever in the world.

So what I've done is model the building and Rhino and then convert it into Forma using the workflow I just demonstrated. But before I load the embodied carbon tool, Ellis, can you predict what will happen?

ELLIS HERMAN: Yeah. I think that cantilever is going to ring some alarm bells.

KEVIN WALSH: OK, let's see. So yeah, there was an error message here. It's to do with the cantilever. Well, Ellis, can you explain what this is?

ELLIS HERMAN: Yeah. So if we remember my description of this machine learning model, it's predicting based on real buildings. Completed bills of material. And so what that means is that we should trust the model's results more when there are a lot of similar buildings in the training data to learn from, and we should trust it less when there are fewer.

So you said this is the longest cantilever in the world. So not only do large cantilevers generally require sort of idiosyncratic structural solutions, this one literally has no peers.

KEVIN WALSH: OK. So bearing that in mind, in the earlier demonstration I showed, you might have noticed that I was hiding some of the more technical details. So I'm going to go a bit deeper here and explain how exactly to set this workflow up on your own system. All these files, I think, will be available as part of the class handout that you can download, and they should be easy to install on your system as long as you have Grasshopper and Dynamo.

So first, I want to go through the Grasshopper setup. This Grasshopper script you should be able to use without any problems, even if you're not too experienced with Grasshopper. The only thing you need to do is to set a local location for your CSV file. So I have down here a small window. So if you double-click here and enter a local location for your CSV file, just specify any folder.

And I think you should only need to do this once when you start using this file. You're not going to need to set this every time you use the workflow. It'll work across multiple sessions. There's a button, I think, to delete old files. But the script is designed so that it only grabs the most recent CSV file, so you can just leave all those files there forever.

So this is the folder where the CSV files are written. And the script, it runs continuously. There's no push button, so you just need to have Grasshopper script open and running. And you put Rhino geometry onto the export layer, and it becomes split into floors and extruded in a way that the embodied carbon tool can read.

And this is the CSV file that it writes to. So you can see all the relevant data from the geometry is converted into a string of numbers which basically describes a polyline and building height.

So if I go to the next slide, I want to show now the Revit and Dynamo section. So you got to have Revit 2024 or, I think, Dynamo Sandbox 3.1 or higher to use the former nodes. But once you open Dynamo, you need to set the directory where you saved your CSV files again. And once you do that and you hit Run, you can see that these are the parts where you're importing the CSV. And over here, it's translating all the string of numbers into points and stories and floor to floor height.

And then going into this great node that Hobart made for me over here, the basic building nodes. So if you just hit Run down there, you can see it's reading the CSV file now. And if I go into preview, you can see in the background there is the polyline. That polyline forms the base of the geometry.

So one key thing is to set this to Periodic down the left hand side rather than Manual or automatic. That allows you to just minimize Dynamo and Revit and let it run in the background. So you can just move back and forth between Forma and Rhino for the rest of your session.

And then if I go back to Forma, you can see that it's the same system I showed earlier where we're just going to be loading the Dynamo player and hitting Run. I have noticed that sometimes the Dynamo Player gets stuck, and I would recommend just refreshing the Forma browser window. And it seems to generally start up again, once you do that.

So once we go back in, it will pick the open Dynamo file. You just hit Run, and you get the preview. And you can hit Add. So what I'm going to do here is go back into Rhino briefly and-- sorry. Not to Dynamo. To Rhino and make some arbitrary changes to this geometry and jump back over to Forma and hit Run once more.

So you can see that geometry is picked up pretty quickly. There's not a lot of lag time. Writing a CSV is generally pretty fast. I have noticed with some more complicated geometry, it can take two or three seconds to write and read the CSV, but the lag time is pretty low.

So again, to actually run the embodied carbon analysis, we do need to set these building parameters here. So this is a manual step in the process that's still necessary. But yeah, I'm not sure if this will be ironed out, but I've noticed it's sometimes necessary to switch to one analysis and then back to embodied carbon analysis just so that you can actually press the Run analysis button. So it's a small little bug for you there, Ellis. And there we get the embodied carbon analysis tool working.

So actually, I ended my talk last year by highlighting these two areas where Forma wasn't really there yet in terms of embodied carbon analysis and geometry import. Autodesk have addressed one of these by introducing the embodied carbon analysis. But for me, I was a little disappointed they haven't moved quickly enough to support importing geometry from Rhino, which is a key issue for me and my company.

However, they have started introducing scripting tools in Forma, like with Dynamo, that are actually indirectly helping to address this gap. And it's actually pretty exciting to see the potential for users like us here at Nikken Sekkei to build our own versions of tools without really needing extensive programming skills and without really needing to directly interface with Autodesk. We can start building things by ourselves, which is how we're used to working as computational designers working with Grasshopper.

We're quite accustomed to augmenting software to push boundaries. So it is really interesting to see Autodesk products becoming more flexible like this and open to this kind of creative adaptation. I'm hoping that in the future, these kind of parameters like setting structure and materials and window to wall ratios will also become accessible from within Dynamo, and even maybe the opportunity to set proposals.

Or even define Forma proposal locations or project locations from within scripting tools like Forma would be a great way that I think the tool can develop. But, Ellis, do you have any kind of future developments that you're looking forward to, maybe especially in relation to advancements in AI technologies?

ELLIS HERMAN: Yeah. I mean, we of course, have specific features we plan on adding to the embodied carbon analysis to Forma. This broader geometry support, like you mentioned. A lot more customizability of those carbon intensity factors. Stuff like that.

But I think the most interesting thing is this broader view of embodied carbon and AI. So our industry is really in the very first stages of caring about embodied carbon, which means that a lot of the data that we're basing tools like this off of is tough to come by. Over the next few years, that data is going to become a lot more complete. A lot more standardized. A lot more public and transparent.

And that will make tools like this even better. It'll make them more accessible. And really importantly, it'll make them more comparable, which is what will help create these project-long workflows that we need for informing and for building on those decisions that impact carbon.

KEVIN WALSH: Yeah, it's really interesting. And actually, I hope that all these new directions that you take in Forma will actually give me a good idea for my next AU talk. So thanks, Ellis, for making the tool. And also, thanks for all your input on the talk today.