Engineering a bridge that designs and builds itself

In the previous articles, we shared the process of designing the bridge, and validating the materials and sensors. We are now almost ready for manufacture.

Print Preparation and Robot Simulation

With a complete design and infill pattern, the part was sliced using ‘Additive’ capabilities in Fusion 360’s Manufacture workspace – the very same method used to slice parts for a desktop ‘FFF’ printer. As explained in Part 1 of this series, although we would print the part on an angled build plate, there was no change in the ‘overhang angle’ rules applied in the slicing software. However, we prototyped a range of new optimizations to improve print quality, especially at the ‘seams’ where the printer starts and stops extruding.

At this point, we had a print file which described over 24 kilometers of toolpath! But, given we would be using a robot for manufacture, we still needed to confirm this entire print path could be followed accurately and safely by our track-mounted robot. We simulated every motion of the print path using a proof-of-concept Robotics Add-in for Fusion 360, which allowed us to simulate the 7 axes of our robot + track combo, and output ABB RAPID code to be run directly on the robot.

Critical Geometry

Despite the many steps of validation and simulation, we still planned for three test prints of key geometry. The worst scenario would be to start printing and find the geometry wasn’t printable!

We chose three sections of the bridge with particularly difficult or unusual geometry. These were (1) the area with the shallowest internal overhang, where quality would potentially be poor if not adequately supported by the gyroid infill, (2) the area with the longest unsupported deck span, to ensure the branches would connect up accurately, and (3) the very tip of the bridge, where the overhang was the greatest.

At the same time, these parts allowed us to uncover any teething problems with the extruder and track setup. Happily, all three parts were completed with success. We made some small visual improvements by adding larger radii where the branches connected to the deck and began preparing for the final print.

The Final Print

The bridge was printed onto a metal build plate, which was coated in a layer of granulates, bonded down with epoxy resin. This ensured a strong bond to the first layer of the print.

At certain stages during the print, we interleaved milling operations using a milling router attached to the robot head. The first milling operation was used to add bolt holes to the part, to give a very strong connection to the build plate and make fully sure the part wouldn’t peel off as it grew heavier.

Other milling operations were used to add channels for sensors and to add holes for cables to be pulled through the inside.

After any milling operation, the part cooled down significantly, so we used Infrared heat lamps to bring the temperature of the part back up to a range where we would get adequate interlayer bond strength again.

Although it was designed to be self-supporting as a cantilever, we decided we couldn’t be too careful, especially given we paused printing over nights and weekends, so we used two boat stands to support the bridge from underneath. To be sure that we were staying within tolerance, we worked with FARO and their Focus Laser Scanner. By taking scans every few hundred layers, we could track if the part was beginning to deflect under its own weight, and indeed if the overall print was true to CAD.

The printing of the bridge lasted roughly 200 hours over the course of 5 weeks in the Autodesk Technology Center in Boston, printing within office hours so that the team could keep an eye on it.

The bridge was completed in August 2022 and exhibited at Autodesk University in New Orleans. The 5m bridge will soon be installed in Dar’s Cairo office in Smart Village, Egypt. Its little sibling, the 2m bridge, will be displayed in Dar’s group headquarters in London UK.

Peter Storey is a Research Engineer at Autodesk.