Industries / Transport and vehicles / Aerospace and defense
Airframes and materials
This is the work of designing and building an aircraft's structure, meaning its wings, body and tail, and choosing the metals and composites it is made of.
4
Milestones achieved, from 1915 to 2011.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1915
Wings without bracing wires
The Junkers J 1 was built of paper-thin iron sheet, 0.1 to 0.2 mm thick, welded to iron ribs, with thick wings that carried their loads inside instead of on external wires and struts. Test flights began on 12 December 1915 at the army airfield at Döberitz near Berlin. The Hugo Junkers heritage site calls it the world's first airworthy all-metal aircraft without bracing wires and with thick wings. It was fast but heavy and slow to climb, and only one was built.
LessonA metal structure has to be light as well as strong, and the J 1's weight kept it a one-off.
Credit
Hugo Junkers, Junkers & Co
1954
Metal fatigue breaks the Comet
The de Havilland Comet began the first jet airliner service in May 1952. In January and April 1954 two Comet 1s broke up in flight. Metal fatigue is damage that grows a little with every load cycle, here every pressurisation of the cabin. The Royal Aircraft Establishment at Farnborough put a whole Comet fuselage in a water tank and pressurised it again and again. It split after about 3,060 cycles, at the corner of a squarish escape-hatch window. The popular story blames the passenger windows; on the aircraft lost near Elba the crack began at a cut-out for a radio antenna.
LessonA pressurised cabin must be fatigue-tested whole, and about a year after the reports British rules began to require repeated-load tests on a complete pressure cabin.
Credit
Royal Aircraft Establishment; Court of Inquiry chaired by Lord Cohen, report CAP 127 (1955)
1988
Many small cracks join up
On 28 April 1988 part of the upper fuselage of an Aloha Airlines Boeing 737-200 tore away in flight. The aircraft had made 89,680 flight cycles, each one a takeoff, pressurisation and landing. The US National Transportation Safety Board (NTSB) found that the airline's maintenance program missed disbonding (glued joints coming apart) and fatigue cracks in a fuselage lap joint, where skin panels overlap. The FAA's account describes multiple-site damage: many small cracks at neighbouring rivets that link up suddenly into one long crack.
LessonAgeing structures can fail in ways that a one-crack-at-a-time analysis misses, and the US later wrote a rule for widespread fatigue damage.
Credit
NTSB investigation, report AAR-89/03
2011
Half the structure in composites
Boeing delivered the first 787 to All Nippon Airways on 26 September 2011. Composites, here carbon fibres set in hardened plastic resin, make up 50% of the 787's primary structure, including the fuselage and wings, according to engine maker Rolls-Royce's release that day. Earlier airliners were built mainly of aluminium.
Credit
Boeing; launch customer All Nippon Airways
Watch for
A blended-wing airliner
JetZero says its Z4, an airliner whose body and wing merge into one lifting shape, will burn up to 50% less fuel than today's tube-and-wing jets. A full-scale demonstrator is being assembled by Scaled Composites in Mojave, California, to fly in 2027, and the company forecasts entry into service in the early 2030s. United Airlines' conditional order for up to 100 depends on the demonstrator flying in 2027.
Achieved whenthe full-scale demonstrator flies and someone outside JetZero, such as a government test agency or the FAA during certification, publishes a measured fuel-burn figure.
Claimed by
JetZero, June 2026, at the groundbreaking of its factory in Greensboro, North Carolina
Open
01
Public scans of composite flaws
An open, labelled collection of ultrasonic scans of carbon-fibre panels with flaws planted at known places and depths, such as delaminations (layers coming apart), porosity and impact damage, with a baseline model that marks the flaws and a script that scores any model. An ultrasonic C-scan is a map of the echoes that come back when sound is sent into a panel, and it is the standard way to find flaws hidden inside composites.
The briefCredit · unclaimed02
Robot layup on curved moulds
A robot arm that lays a sheet of prepreg, carbon-fibre cloth already soaked in uncured resin, onto a mould curved in two directions without wrinkles, trained from recordings of people doing the job.
The briefCredit · unclaimed
Open 01 · the brief
Public scans of composite flaws
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, labelled collection of ultrasonic scans of carbon-fibre panels with flaws planted at known places and depths, such as delaminations (layers coming apart), porosity and impact damage, with a baseline model that marks the flaws and a script that scores any model. An ultrasonic C-scan is a map of the echoes that come back when sound is sent into a panel, and it is the standard way to find flaws hidden inside composites.
Why it is still open
Robots already move the probes, but a 2025 University of Strathclyde doctoral thesis states that "the analysis of the resulting data remains a manual and labour-intensive task", and it trained its models on synthetic data because real data are scarce. A 2025 study that built 600 C-scan images of flawed panels did not release them as a dataset (its data statement says the data are contained within the article) and reported that the small quantity of data caused overfitting. A search found no public, labelled benchmark of real composite C-scans.
Built with what exists
Panels laid up with planted flaws; a university or service-lab ultrasonic scanner; open segmentation models such as U-Net or Meta's Segment Anything; Python; Zenodo or Hugging Face Datasets to host the files under an open licence.
First step
Write the label format, then ask a university non-destructive testing lab to scan five panels with planted flaws and label them by hand.
Done looks like
At least 100 panels and 1,000 labelled flaws published with their true positions, sizes and depths, and a baseline model's detection rate and false-call rate reported for each flaw type and size, which anyone can reproduce with the published script.
Open 02 · the brief
Robot layup on curved moulds
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A robot arm that lays a sheet of prepreg, carbon-fibre cloth already soaked in uncured resin, onto a mould curved in two directions without wrinkles, trained from recordings of people doing the job.
Why it is still open
A 2022 paper states that "for low-volume production and complex parts, a hand layup is still the main method." Its own robot learned layup motions from a person steering it by remote control, but it was tested only on flat and sloping surfaces, and the authors measured how closely the arm followed its path, not whether the laminate wrinkled.
Built with what exists
A seven-joint arm such as the Franka arm used in that study, with a roller tool, a wrist force sensor and cameras; Hugging Face LeRobot to record demonstrations; an open robot model such as NVIDIA GR00T 2 or Runway Praxis-1 to fine-tune on them. The Notes cover these in "Five demonstrations per task", "Open robot model weights" and "No download".
First step
Make a small double-curved mould and film a laminator laying 20 plies on it, which gives both the training recordings and the human baseline.
Done looks like
50 plies laid on the same double-curved mould, each photographed, with the count of plies that wrinkled or bridged (lifted off the mould in a corner) judged by someone outside the team, set beside a skilled laminator's count and time per ply on the same mould.