3D Printing in Production
Metal printing delivers engine parts and rocket components in series; printed houses remain a niche.
Open in the interactive tree →GE's LEAP engine, in service since 2016, uses 3D-printed fuel nozzles that replaced 20 parts with one and are 25% lighter; powder-bed laser melting is now used in series production in aerospace. Rocket start-ups print engines too: Agnikul's Agnibaan flew in 2024 on single-piece printed engines, while Relativity Space has moved away from heavy printing for its Terran R rocket. Builders also print concrete houses layer by layer. So far it is economical mainly for small volumes and complex shapes.
As of October 2026
In 2026 the Indian start-up Agnikul fired a cluster of three single-piece printed Agnilet engines (February) and test-fired the Agnite booster engine, a one-metre printed Inconel engine (March); it says it can make a flight-ready engine in about seven days instead of seven months. Relativity Space, once the showcase of rocket printing, now buys fuel-tank domes and fairings for Terran R from outside suppliers. Metal printing is established for small volumes of complex aerospace parts, while printed houses remain a niche.
Open steps
- In-process flaw detection High AI leverageDetect lack of fusion and pores during the build so parts can be certified without full CT scanning; validate against CT data for each machine and alloy.
- Faster parameter sets for new alloys High AI leverageCut the months of trials needed to qualify laser, scan and powder settings for a new alloy or machine.
- Throughput and post-processing Medium AI leverageRaise build rates with multi-laser machines and automate support removal, heat treatment and finishing, which dominate the cost of series parts.
- Printed buildings: structure and codes Low AI leverageProve that reinforced printed walls meet structural codes (rebar integration, layer bonding, durability) so printed buildings can leave the niche.
Where AI could help
Medium AI leverage. AI improves process monitoring, parameter search and design, easing qualification; build speed, powder cost and certification rules limit series use.
- Detect flaws from in-situ sensor data to cut inspection and scrap
- Search process parameters and new alloys with Bayesian optimization
- Generative and topology design of printable parts
- Predict distortion and compensate before the build
Shown so far
- In 2022 Oak Ridge researchers showed that convolutional networks trained on in-situ sensor data detect fatigue-critical lack-of-fusion flaws in laser powder bed fusion, but real-time analysis can slow builds. source
Prerequisites
Sources
More in Manufacturing & Construction · Present
- Robots & Lights-Out Factories2015–2026
- Green Steel2021–2027
- Cement with CO2 Capture2025