Industries / Industrial equipment and services / Industrial machinery / Boilers and pressure vessels
Forming and welding
This work turns steel plate and ingots into shells, heads and rings, by rolling, pressing and forging, and joins them with welds as strong as the plate.
4
Milestones achieved, from 1951 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1951
Welding metres-thick metal
The E. O. Paton Electric Welding Institute in Kyiv developed electroslag welding in 1951. In this process the weld metal is melted by current passing through a pool of molten slag instead of an open arc, and by the institute's timeline it joined metal up to 2,000 mm thick. The institute names water turbines and presses as early uses; its timeline does not mention boilers or vessels. Electroslag remelting of steel followed in 1952.
Credit
E. O. Paton Electric Welding Institute
2010
Presses that remove welds
Japan Steel Works commissioned a second 14,000-tonne forging press at Muroran in early 2010, able to forge ingots of about 650 t by the World Nuclear Association's table (670 t in its text). One-piece forgings of reactor vessel rings and heads replace welded joints, which the association notes would otherwise need checking through the life of the plant. JSW's capacity rose from 4 reactor vessel sets a year up to 2007 to 12 by early 2011.
Credit
Japan Steel Works
2012
Pressed powder becomes code metal
Powder metallurgy with hot isostatic pressing (PM-HIP) packs metal powder into a sealed can and presses it at high temperature and gas pressure into a solid part, with no casting or forging. EPRI research from 2009 produced a 3,600 lb vessel nozzle this way, and David Gandy of EPRI reported in 2016 that four ASME Code Cases, for 316L stainless, Grade 91 and duplex stainless steel, had been issued since 2012.
Credit
Electric Power Research Institute (EPRI), David Gandy; ASME
2025
A vessel welded in minutes
Electron beam welding melts the joint with a beam of electrons in a vacuum and can fuse a thick wall in one pass. In May 2022 Sheffield Forgemasters and Cambridge Vacuum Engineering welded a 200 mm thick ring of SA 508 steel, about 10 m around, in about 140 minutes, and ultrasonic testing found no reportable flaws under ASME's acceptance criteria. In July 2025 Kairos Power installed a 14 ft reactor vessel, welded by electron beam at the AMRC in South Yorkshire, in its non-nuclear test unit in Oak Ridge, Tennessee; the AMRC says welds that took a week on the earlier vessel took about 10 minutes. EPRI's David Gandy and Marc Albert report a 1.82 m mock-up welded in 47 minutes against 7 to 8 days by narrow-gap arc welding.
Credit
Sheffield Forgemasters; Cambridge Vacuum Engineering; Nuclear AMRC and the AMRC, University of Sheffield; EPRI; Kairos Power
Watch for
Electron beam welding in a US shop
EPRI's David Gandy and Marc Albert write that the first US heavy-section modular electron beam welding system is being commissioned at BWXT in Ohio and is to weld several reactor pressure vessel shell rings from 2026. An EPRI joint industry project is gathering data for ASME on welding without preheat and on repair and inspection of such welds.
Achieved whena production shell ring weld at BWXT passes code inspection, or ASME publishes a code case for electron beam welding of reactor vessel steels.
Claimed by
EPRI authors in POWER, April 2026
A printed steel vessel
Oak Ridge and Idaho national laboratories printed a closed steel vessel about 3 ft by 5 ft by wire-arc additive manufacturing, which builds a part from weld beads laid by robot, on ORNL's three-arm MedUSA system, with the aim of vessels checked as they are printed. No mechanical test data or code status was reported.
Achieved whentensile, toughness, inspection or burst results for the vessel are published, and a code case or stamp covers such a vessel.
Claimed by
Oak Ridge National Laboratory and Idaho National Laboratory, August 2026
Open
01
Heat-treatment charts read by machine
A program that reads the thermocouple records of post-weld heat treatment, the slow heating and holding that relieves stress in a finished weld, from logger files or scanned paper charts, and flags any hold that ran too cool, too short, or heated or cooled too fast against a rule table the user sets.
The briefCredit · unclaimed02
Flaws found in seam scans
An open model that finds weld flaws in the raw phased-array ultrasonic scans (ultrasound from many small elements steered electronically) of submerged arc welds, the welds used on vessel seams, and is scored against the flaws the scans' authors located.
The briefCredit · unclaimed
Open 01 · the brief
Heat-treatment charts read by machine
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A program that reads the thermocouple records of post-weld heat treatment, the slow heating and holding that relieves stress in a finished weld, from logger files or scanned paper charts, and flags any hold that ran too cool, too short, or heated or cooled too fast against a rule table the user sets.
Why it is still open
In the 1965 Immingham failure described under Testing before shipping, the weld was specified at 620 to 660 °C for 6 hours but the furnace reached only an estimated 520 to 610 °C, according to TWI's case study, and the shortfall went unnoticed until the vessel broke. A search in November 2026 found commercial heat-treatment recorders but no open tool that checks the records against a specification.
Built with what exists
Python with pandas; WebPlotDigitizer or an AI vision model to read paper charts, every reading checked by hand; a rule table in YAML that can be edited.
First step
Write the rule table for one common steel, then build ten synthetic records, five sound and five faulty.
Done looks like
A published test set of at least 50 records, real and synthetic, with known faults planted in a stated share of them; the tool catches every planted fault, including a re-creation of the Immingham shortfall, and its false alarms are counted.
Open 02 · the brief
Flaws found in seam scans
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open model that finds weld flaws in the raw phased-array ultrasonic scans (ultrasound from many small elements steered electronically) of submerged arc welds, the welds used on vessel seams, and is scored against the flaws the scans' authors located.
Why it is still open
Tapia Suárez and colleagues published the "Submerged Arc Welding Open Repository" on Zenodo in March 2025 under a CC BY license: 12.7 GB of phased-array files, X-ray and ultrasonic reports and a spreadsheet of flaw locations for each weld bead. A search in November 2026 found no published model or baseline that detects these flaws from the raw scans.
Built with what exists
The Zenodo repository; Python with h5py to read the scan files; PyTorch for a detection model on one graphics card.
First step
Download the repository and match each scan file to its rows in the flaw spreadsheet.
Done looks like
Open code that finds at least 90% of the listed flaws in held-out welds, with its false calls per metre of weld reported.