Cosmic Senior Capstone Project
NASA Goddard Space Flight Center, 2025
Background
From August 2024 to May 2025, I worked on Project Daedalus, a year-long senior design project at Virginia Tech. The mission was developed for the 2024 to 2025 Consortium for Space Mobility and In-Space Servicing, Assembly and Manufacturing Capabilities (COSMIC) Capstone Challenge, with the goal of designing a payload that could autonomously perform additive manufacturing, inspection, and assembly within the tight constraints of a small satellite bus.
This work was carried out under the guidance of Dr. Kevin Shinpaugh, professor in the Kevin T. Crofton Department of Aerospace and Ocean Engineering at Virginia Tech, and Dr. Randy Spicer, staff engineer at Northrop Grumman, whose mentorship shaped both the technical depth and systems engineering discipline of our team.
Project Summary
Project Daedalus was designed to show what a student team could achieve in the field of in-space manufacturing. Our payload brought together three technologies: Directed Acoustic Energy Deposition (DAED) to 3D print aluminum in microgravity, a laser welding system to assemble printed parts, and an AI-powered Non-Destructive Evaluation (NDE) pipeline for real-time defect detection. By combining these systems, we demonstrated how future spacecraft could manufacture and repair structures directly in orbit, extending mission lifetimes and reducing reliance on Earth.
My Role & Contribution
- Researched machine learning-based inspection algorithms and prior literature on 3D printing defect detection.
- Designed a CNN(Convolutional Neural Network)-based AI inspection pipeline using optical and IR data for real-time defect classification.
- Built the FDIR logic for fault classification, safe-mode handling, and recovery.
- Defined the command and data handling architecture between payload AI, OBC, and bus.
High-level AI inspection flowchart — core of my autonomy and controls contribution
Integrated system architecture — linking payload, bus, and ground control
As the Autonomy & Controls subteam lead, I was responsible for making sure our payload could operate on its own without relying on constant ground control. That meant building systems that could inspect prints in real time, recover from faults, and manage communication with both the spacecraft bus and ground stations — all within a strict 60-watt power budget.
I designed the AI inspection pipeline, combining optical and infrared cameras with a CNN-based model to detect defects layer by layer. To make this practical for space, I worked on adapting transfer learning methods and supported the trade study that led us to select the NVIDIA Jetson Xavier NX, protected by 2 mm of aluminum shil after our STK radiation analysis. The pipeline performed defect classification on features such as flash formation, voids, and surface roughness, reducing reliance on ground verification and enabling autonomous recovery during in-orbit printing.
Demonstration of the CNN-based AI inspection pipeline I designed — using optical and IR cameras for real-time defect detection.
I also developed the logic for our Fault Detection, Isolation, and Recovery (FDIR) system, modeled after NASA’s High-Performance Spaceflight Computing concepts. This gave the payload the ability to classify faults, enter safe mode, and attempt autonomous recovery before relying on ground intervention.
Finally, I mapped out the command and data handling flow, linking the AI computer, the Sirius OBC, and the BCT bus. I created deterministic flowcharts that defined how the payload transitioned between operational states such as manufacturing, inspection, safe mode, and ground override.
System Architecture and Operations
Payload subsystem integration — DAED printer, NDE system, laser welding, and power management
The payload was designed around four tightly integrated subsystems: the DAED printing system for aluminum manufacturing, the NDE inspection system for real-time defect detection, the laser welding system for assembly, and the command/data handling and power management systems that controlled operations and distributed energy.
Additive Manufacturing (DAED)
Demonstration of Directed Acoustic Energy Deposition (DAED) process with aluminum feedstock.
The DAED process used piezoelectric transducers to generate acoustic waves, which directed the deposition of aluminum wire feedstock layer by layer. Controlled motion from stepper and piezoelectric actuators enabled precise cylindrical and square prints. Our design demonstrated how DAED could repeatedly fabricate parts in orbit while maintaining a manageable power draw of ~60 W.
Concept of Operations
Mission Concept of Operations — step-by-step mission flow from deployment to deorbit
The mission sequence included deployment and initialization, printing of aluminum beams, CNN-based inspection, and either laser welding assembly or storage of prints in a carousel. The final stage included a controlled deorbit burn for disposal.
Laser Welding and Storage
Laser welding and storage operations — assembling two aluminum cylinders or archiving failed prints.
The laser subsystem used a YLM-QCW fiber laser and LDW200 laser wobbler to weld printed cylinders together. Thermal management relied on active fluid loops with Novec 7000 coolant and passive radiators. A storage carousel allowed failed prints to be archived and replaced with prefabricated spares, ensuring mission continuity.
Challenges
The hardest challenge was designing under the tight constraints of the BCT X-Sat Venus-class bus. With only 60 W of available power, strict volume limits, and a capped payload mass, we had to be strategic about every component. The Bill of Materials became a constant negotiation between ambition and feasibility.
Beyond hardware, there was also the challenge of balancing imagination with engineering reality. Aerospace projects invite bold ideas, but we had to ask: could this survive launch loads, function in microgravity, and operate reliably for two years in orbit? Answering those questions required iteration after iteration, trade studies, and compromises that grounded creativity in feasibility.
BCT X-Sat Venus-class bus — constraints that defined our design space
What I Learned
This project reshaped the way I think about leadership, teamwork, and engineering.
When I began leading the Autonomy & Controls subteam, I carried over a strict, military-style approach: assigning tasks, setting hard deadlines, and expecting execution. I quickly realized that in a peer-based engineering team, that approach hurt collaboration more than it helped. I had to shift my style — leading through respect, trust, and shared ownership instead of orders.
I also learned to embrace collaboration over working alone. As an international student, I often felt the need to overcompensate and push harder by myself. But this project showed me that spacecraft are never built in isolation. Our success came from combining diverse strengths, debating openly, and building on each other’s ideas.
Finally, I became a stronger systems thinker. This project forced me to move constantly between the big picture — mission objectives, autonomy requirements, ISAM goals — and the smallest details — shielding thickness, thermal margins, communication bandwidth. I learned how to bridge those levels, ensuring our boldest ideas were backed by engineering rigor.
Project Results
Official coverage: Virginia Tech claims 2nd place at COSMIC Capstone Challenge
- 1st place at Virginia Tech Space Vehicle Design Challenge
- 2nd place nationally at COSMIC Capstone Challenge (presented at NASA Goddard Space Flight Center)
COSMIC Capstone Challenge National Result — Team H.A.D.E.S, 2nd Place
Certificate of Excellence in Space Vehicle Design— First Place at Virginia Tech