Now that my head is briefly back above water following Autodesk University 2022 and some time off with the family, I thought I’d revisit the Dar 5m Smart Bridge project to share some of the things I learned about it during AU in New Orleans.
As a reminder, I was not heavily involved in the building of the bridge itself – although I did track it with great interest from a distance – as my role really only came into play in the final stages of the project, when we were looking to get the data from the bridge into Dasher (for both historical and real-time display).
But I did spend close to 3 days talking about the bridge alongside members of the team who built it, so I do feel that I picked up some information that’s likely to be of interest to people.
The first thing to note is that this is the second bridge that Autodesk Research built for – and with the strong participation of – Dar. The first was a 2m bridge that was 3D-printed from carbon fibre-reinforced ABS in our Birmingham Technology Centre in the UK. That particular bridge weighed in at around 90kg and had a very different design from the larger bridge: it was effectively a table-like structure that was printed upside-down. While having a different design, material and printing process, a lot was still learned from this initial phase of the project that was applied later on.
So what was different about the 5m bridge?
Let’s start with the obvious: it was 2.5 times the length of the first one. What’s sometimes less obvious to people is that an expansion of (say) 3X in each spatial dimension actually means an increase of 9X in terms of the overall material volume, and this is effectively what we’re talking about, here: the 5m bridge has roughly 10-11X the weight of the 2m one. (They have different designs and were made from different materials, but the scale-up in material is basically in-line with this rule of thumb.)
The 5m bridge weighs in at just under a metric ton, at 960kg. It has been designed to support 2.5 tons of load, way less than we were allowed to put on it at AU.
The design of the 5m bridge also made more use of Fusion 360’s generative design capability: while the design of the 2m bridge relied on GD for about 40% of the work (vs. 60% manual) this was closer to 90% for the 5m bridge. Fusion 360 was also used for the printing of the 5m bridge.
One of the things I like most about the 5m bridge is invisible in the final product: it has a gyroid (in fact a cycloid) infill that gives the structure significant strength relative to its weight. (In contrast the 2m bridge had a solid fill albeit with hollow parts.)
The material used for the 5m bridge was recycled PET-G with glass fibre reinforcement. Just like the 2m bridge (which was in contrast made from non-recycled material) the 5m bridge could in theory be ground-up and used to make other (non-structural) objects in the future.
The design of the 5m bridge is a single cantilever that needed to be able to support its own weight during the printing process.
While in a controlled environment it was printed in an upright position to approximate the robot printing in-situ across (say) a body of water.
The cantilever was only for the printing, however: the completed bridge was always intended to be supported at both ends – e.g. on both banks of the river.
The printing took approximately 200 hours over a 5-week period, equating to roughly 10 hours per day of printing. The printing was monitored closely, both by the research staff and by sensing devices that included thermal cameras. Sensors in the extruder made sure the temperature of the material was within the appropriate range: hot enough to melt but not hot enough to crystallize and clog the extruder.
When I first saw that heat lamps were used during the printing I assumed it was to dry the material quickly, but actually it was the opposite: they were used to keep the existing material soft enough to adhere properly with the newly deposited material. (And yes, this is one more example of why I’m focused on building software rather than figuring out real problems in meatspace. :-)
Two types of sensor were embedded in the bridge: once again we used Fiber Bragg Grating (FBG) sensors for the decking, but for the supports we used traditional strain gauges. Both types of sensor ultimately report strain, and having two types of sensor did add complexity to the system, whether in terms of installation (the soldering for the traditional strain gauges), the data management (we had middleware from both FiSens and HBM running) or the visualization (we had socket messages coming through for the realtime display that only contained partial data payloads which then needed special handling inside Dasher). It was also a bit of a pain to callibrate the two sets of sensors so that they reported strain in the same range.
I’m still very impressed by FBG as a technology: it’s based on optical fibres that refract light differently when under strain. They’re much simpler and apparently more robust that traditional electronic sensors. One interesting note is that once again (we had already found this with the 2m bridge at BIM World Paris) we had to zero the sensor readings in the middleware very regularly: I assume this drift is mainly down to thermal expansion of the bridge, but there may be other reasons, too.
This shot of Ghassan from Dar on the bridge is a nice demonstrator of the data coming off it and being displayed in real-time.
Here’s a picture taken by my old friend Debashis Dhar from the AutoCAD team. It’s a pretty typical view of me during this year’s AU, pointing at sensors and perhaps the PC running the bridge’s “OS”.
Anyway, talking to customers – and many curious exhibitors and Autodeskers, too – about the bridge was a whole lot of fun.
On the last day of AU I was interviewed for an Autodesk Research video talking about the event. Because my segment didn’t make the cut – they assured me it was due to timing (it was one of the last segments shot) rather than content ;-) – Dan Ahern from our communications team kindly put together a clip with me in it. (Thanks, Dan!)
Congratulations, once again, to all those involved in this project. Here are the members of the team who made it to AU in New Orleans: Gonzalo Martinez, Peter Storey, Ghassan Zein (Dar), Taylor Tobin, Stefanie Pender, James Donnelly. Those who weren’t able to make it (or weren’t in the photo, anyway): Fikret Kalay, Nic Carey, Fope Bademosi, Brandon Cramer, Franck Messmer.
This project is significant for many reasons: it has proven the ability to use large-scale additive manufacturing to create smart infrastructure using sustainable materials. As the design made use of Fusion 360’s generative design capability – and we’re using sensors to collect real-world performance data – it’s feasible to “close the loop” and use this data to create a more optimal design during the next iteration. I’m glad to have had the chance to be involved in a project of this importance.