We are thrilled to present guest author Michael van Telgen , design automation specialist and structural engineer with Arcadis, who has been doing a lot of work with Dynamo and Refinery. Note that this work was done in the Project Refinery beta which has since graduated to Generative Design in Revit 2021


Exploring Structural Design Options Using Generative Design

Every generation of structural engineers dreams of designing the ‘perfect’ structure. With emerging technologies, reaching that goal is closer than ever before. In this blog I elaborate on the use of structural parametric design at Arcadis, how we use Autodesk Refinery to analyze our designs and how this benefits our clients.


Creating better solutions

Since I was a master’s student at the University of Eindhoven (the Netherlands) almost ten years ago, I am fascinated by the emerging digital design technologies and how they relate to my field of engineering: structural design. When doing a simple Google images search on ‘parametric design’ nowadays, you are overloaded with structural design related content. Student Michael would of course be delighted, but there is one important thing this Google search will not tell you.

Parametric designing of structures is not only about generating complex geometry. It is about creating better solutions.

That doesn’t make this design technique any less interesting though. On the contrary, I only means that our design cases are often much less visually pleasing as Google would let you believe. This lack of esthetics is vastly compensated by the sheer number of possibilities to investigate with parametric design. After a readthrough of this blog I hope you will agree.

Google images search for ‘parametric design’ (April 28th, 2020)


The awesome power of structural parametric design

The company I work for, Arcadis, is a huge design and engineering company that has lately expressed much interest in exploring the design automation domain. But as is often the case, it’s also difficult to bring ideas into practice. Digitization of structural engineering solutions does not prove to be any different.

It’s not the technology’s fault though. Applications such as Grasshopper, Dynamo and the various (already on-the-market) structural connectivity tools have been around for quite a while and we could have leveraged them extensively for years already, but we didn’t. For us, this mostly had to do with the democratization of technology. While we knew about these for a long time, we just didn’t know how to use them effectively.

Two things had to change in order for us to wield the awesome power of structural parametric design

  1. We needed connectivity between parametric design and structural design
  2. We needed a good way to optimize structures and do cross-product analyses


Dynamo to RFEM connectivity tool

Starting in 2018, we invested heavily in the creation of our in-house tool to connect Dynamo with structural engineering software RFEM. This tool was inspiringly dubbed the ‘DynamoRFEM’ tool – we’re still open for naming ideas. DynamoRFEM takes care of translating our Dynamo model into something that the structural software understands and calculates. For instance, Dynamo lines can be converted into beams by assigning properties to them, such as materials and cross-sections. Using this conversion, we can create structural models of any kind, shape, complexity and size.

With the truss generation Dynamo graph, we can create and calculate designs instantaneously

Mock-up for a sound retention wall. This example of a linear structure follows any curve path and aligns the structural elements accordingly

Mock-up for a balcony and tree pot design for a building façade (project WonderWoods, Utrecht, the Netherlands). The balcony and tree pot rest on cantilevering beams.


Project Refinery

One way we help our clients is by ranking different design options. This task can be tedious, especially when reporting the analysis’ results on paper. The better alternative is to either do a cross-product study or an optimization study with Autodesk Refinery. This product has gradually developed during the past years to a suite of tools that enables generative design in Revit 2021. A Refinery study is easy to set up:

Setting up a cross-product study in Refinery

With the cross-product study, you decide which of the Dynamo model’s input fields can be varied. Refinery then runs through all options and reports the results in customizable graphs. In the example below we calculated 240 different options for a truss design. The truss height, the number of divisions and the level at midspan are varied. Those results are then easily compared:

Checking the results of the cross-product study. Analysis time for 240 variants was 30 minutes on a Dell Precision 7520 laptop

Optimization is also possible. Here Refinery really stands out, because you can do minimization and maximization of multiple preferred outputs, even while constraining particular outputs to user-defined minimum and maximum values. In the truss example, we can for instance minimize the truss chord tensile normal force while keeping the Serviceability Limit State (SLS) vertical deflection unity check between 0,8 and 1,0 – the maximum allowed range for deflection of the truss in relation to cost-effectiveness. It is also possible to include cross-sectional checks for example.

One important note though: the quality of the optimization result is heavily influenced by the calculation setup quality and the definition of the optimization goals, so well-defining the problem is essential and a skill in its own right.

Input screen for the optimization study

The optimization results for the truss design. Refinery found one (of many) solutions that fit the optimization goals. Analysis time for optimization was 15 minutes on a Dell Precision 7520 laptop


Case study: lateral deformations for concrete high-rise structures

I work in the city of Rotterdam. Due to the Dutch soil conditions (thick clay layers lying on deeper sand layers) you don’t find too many high-rise dense cities here. Rotterdam is one of the exceptions. With increasing land prices, pockets of high-rise buildings are found throughout the city, the newest – of course – reaching higher than the older. And the city will not stop the emergence of this skyline anytime soon.

In the Netherlands, high-rise office buildings usually have a concrete core for lateral stabilization. Core walls typically contain staircases and elevators and can have a closed floor plan, thereby forming a stiff structure. While a bigger core will makes the building stiffer and less prone to wind deflections, a larger core also means less profit for the real estate developer as the rentable floor area decreases. Beside the core’s shape and dimensioning, the design of the foundation and the building size, shape, and location also determine the required core size. This makes the stability design a complex affair. In the conceptual and preliminary design phases we put a lot of effort into the core design, all to maximize cost efficiency.

To simplify this design exercise, we have been looking at possibilities with two platforms, Packhunt.io and Viktor.ai beside our own solution. Though they don’t offer off-the-shelf solutions for this problem, these platforms can be used to simplify design challenges and roll-out the design tools to our engineers – as said before, democratization of technology is important to us. The DynamoRFEM tool (together with Refinery) is used to benchmark the future Packhunt.io and Viktor.ai tools, aside from being the first simplification of the core design process.

In the example below, an arbitrary core configuration was created in Dynamo, with three elevator shafts, a staircase, and a central hallway. Beside the lay-out of the core, the configuration is fully parametric. We can set foundation properties, such as foundation slab thickness, material strength, the number and layout of piles and the spring stiffness of the piles. Also parametric are the core’s dimensioning and materialization, but we must set the applying wind load ourselves – the mandatory Eurocode calculation was not yet implemented. This could of course be made in a future update. All in all, the Dynamo graph took about 8 hours to set up.

Chosen core configuration

We will target a lateral deflection SLS unity check (UC) between 0,6 and 0,7 to keep enough reserve for future design revisions, under a constant wind load for a building with 20 stories (resulting in a total building height of 64m). We will look for a balance between the core size and smallest (cheapest) foundation.

Typically, the deflection ratio between the core’s deflection and the deflection caused by foundation rotation is about 1:1. We will check this first by changing the point supports from a full support to a spring support. As the deflection increases from 30mm to 85mm the ratio is about 1:2, indicating the foundation is the biggest influence. The found unity check of the deflection (calculated using the Eurocode) equals to 0,23, so our design is safe, but cost optimization can (and should) be made.

Deflections of the core with pinned (left) and spring (right) supports

We will use Refinery’s cross-product study to check several design options. The length of the wall part featuring the access door is varied in length (7 steps from 3m to 6m), as well as the number of piles in X- (4 to 8) and Y- (4 to 6) directions, resulting in 105 different design options.

Variables in the core configuration

Input screen for the cross-product study

We can check the results either using the scatterplot or the parallel coordinates view. As we are only interested in design options with lateral deformation UC’s between 0,6 and 0,7 we can filter those results. When we use the filtered parallel coordinates view it’s clear we can either have a smaller number of piles and a longer core, or we can reduce the core’s length with more piles.

This result seems kind of obvious, but it shows we have flexibility in the design at this stage and can move either way, which is a valuable conclusion. Also, we should not overlook the fact that we are able to estimate the required dimensioning, as well as the spread in dimensioning possibilities with much more precision than before by hand.

When costs are included (you can of course also optimize on costs) the results will become insightful for the client as well as the structural engineer. We have yet to take the step of linking our structural data to a cost estimation module.

Checking the results of the cross-product study. Analysis time for 105 variants was 25 minutes on a Dell Precision 7520 laptop



In our opinion the DynamoRFEM tool, together with Refinery provides a powerful new possibility for in-depth structural analyses and will lead to the creation of better structures. While the core calculation in the example grossly oversimplifies many design aspects, the study does provide new insight in the structural behavior of the core and helps the designer make choices that would otherwise be mere guesses or take much more time to calculate. The opposite could also be argued. As the calculation is more complex than we are used to, who is to say we are not overestimating the design possibilities now and will run into troubles later? As with all design processes, best practice is to use common sense. With the use of these tools during the coming years we expect to get more and more familiarized and confident.

I would like to thank Lilli Smith for the chance to publish this blog on DynamoBIM.org and the Refinery team for their great work. Also, I would like to thank Tom Borst and Toine Piters for their editorial support.