This was one of my most intense projects thus far. With a team of about 20 people, we developed an automated robot for use in the NDA (non-destructive assay) of pipes used in nuclear diffusion piping in about 4 months.
Overview
Starting early 2017, we began discussions with the Department of Energy (DOE) about how we could help in the cleanup and close of a nuclear facility in Portsmouth, Ohio. Previously used for weapons, energy, and eventually commercial purposes, this massive site was built very quickly during the Cold War. Currently, the DOE is looking to shut down and demolish the facility, but it’s not simple. The pipes used in gaseous nuclear diffusion (a process used to enrich uranium) are still lined with some radioactive material. Before the facility can be destroyed, all pipes lined with this material need to be taken out and cleaned. Unfortunately, this process is not easy. Workers manually record radioactive measurements section by section, and therefore it is costly, because there are miles of pipe to evaluate. What if there was a way to automate this process, making it quicker, more accurate, and safer for workers by removing them from the radioactive environment?
Design
Our proposed solution was PipeDream, a robot that would evaluate the deposits on the inside of the pipes volumetrically. The design criteria that we came up with were:
- Suitable for 16″—30″ pipes
- Detect down to 1.7-mm deposis
- Overcome 1/4″ deposits/obstacles
- Measurements from inductive and triangulation sensors
- 3-point suspension for centering
- Tube body for structural integrity
- Continuously rotating sensor disk

Pipes varied in size throughout the facility, but the amount of deposit per foot required for pipe removal remained the same. This meant that for the larger pipes, this robot would need to detect a smaller thickness of deposit. If we could validate this system on one of the larger sizes, say 30″, this method of NDA would be valid for smaller sizes. We wanted the robot to be modular for various sizes so it would be more useful and wouldn’t require redesign later on. For simplicity, we decided on a 3-point suspension system for the front and back of the robot (think an inverted umbrella). A singular body tube would house all electronics for structural integrity, as opposed to discrete rectangular pieces connected later. Finally, a singular rotational disc would house all the sensors. A disc allowed the sensors to be fixed relative to one another (as opposed to individual spokes that they could be mounted on) and perpendicular to the pipe (as opposed to being mounted directly onto our suspension). The continuous rotation meant that we were not concerned about backlash in the gears when taking measurements.
Rotation Module
As one of the mechanical engineers on the team, my tasks included designing the rotation module. Some features included:
- 8” ID thin section X-type bearing providing large thrust and axial support (future iterations would require heavy loads on the sensor disc)
- Gearbox and pinion-gear providing 180:1 gear reduction
- Motor encoder and index pulse providing angular position
- Cap housing shaft bearing and wiring
- Seals for dust and light HF gas exposure
In addition to designing this module, I was involved in motor selection, reached out to manufacturers for prices and lead times, and took part in the machining and the assembly of the device. The final assembly can be seen below:
Wheel Module
The second mechanism that I designed was the wheel module. Some features included:
- Teflon (chemically resistant) thrust bearing
- PEEK(chemically resistant) bearings
- Tires easily changable after contamination
- Design for quick assembly and disassembly
Similar to the rotation module, I designed this mechanism, and was also was involved in component procurement, machining, and assembly. The assembly process is shown below:
Summary
This was a grueling project that defined my summer. I’ve never worked with this many people to meet such an aggressive deadline. Seeing such a complex system come together was extremely rewarding. Currently, the two robots we built are back at CMU. Although, the hot test (hot meaning in uranium-coated pipes) at the DOE facility in Ohio was a successful proof-of-concept, CMU continued with a separate robot built in tandem to this, RadPiper, which measured radiation instead of volume. Without the effort on PipeDream, however, we would not have gotten to a place where RadPiper would even have been an idea.
In the future, I’d like to do some more analysis on my rotation module, and perhaps write a paper on it; hopefully I will be able to before graduation in May 2018. As a team, we did publish a paper on the robot, which has been accepted and will soon appear in the Symposium on Waste Management: “Robotic Measurement of Holdup Deposit Volume in Gaseous Diffusion Piping to Quantify U-235 Content”.
Enjoy some more media below!

