Industries / Industrial equipment and services / Industrial services / Pump and valve repair
Machining and parts
Machining and parts is the work of restoring worn parts and making new ones: welding hard alloys onto valve seats, copying parts whose maker has gone, and casting, printing or machining replacements.
6
Milestones achieved, from 1922 to 2022.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1922
Hard alloy laid by welding
Haynes International, which descends from the business Elwood Haynes set up in Kokomo, Indiana, in 1912 to make his cobalt-chromium Stellite alloys, dates the invention of hard-facing to 1922, when the business belonged to Union Carbide: a welding process that lays a layer of hard alloy over a cheaper, weaker metal. Early uses included oil-well drill bits and the valves and valve seats of engines, and cobalt-based Stellite 6 became the usual hard face on the seats of power plant valves.
Credit
the Stellite Division of Union Carbide, formerly the Haynes Stellite Company
1989
Refurbished valves sold as new
In 1988 the US Nuclear Regulatory Commission warned nuclear plants about refurbished valves sold as new and about possibly substandard valve replacement parts, among other counterfeit and fraudulently marketed products. Its Generic Letter 89-02 of March 1989 conditionally endorsed EPRI's guideline NP-5652 (NCIG-07) for what is called commercial-grade dedication: before a part bought on the open market goes into a safety system, the plant verifies by engineering the characteristics that matter for its job.
LessonA replacement part's paperwork does not prove what the part is; the buyer has to check the features that matter.
Credit
US Nuclear Regulatory Commission; the Electric Power Research Institute (EPRI)
1991
Cobalt-free valve seats tested
Cobalt worn off Stellite faces in reactor valves and pumps can be carried into the core and turned into radioactive cobalt-60, which a 2018 paper calls the largest contributor to the radiation doses nuclear plant workers receive. EPRI developed iron-based hard-facing alloys called NOREM, and at Atomic Energy of Canada Limited (AECL) gate valves faced with four cobalt-free alloys, and one with Stellite 6, were opened and closed 2,000 times over about a year under simulated pressurized water reactor conditions; reported in August 1991, all the iron-based alloys did better than Stellite 6. Later laboratory tests at 343 °C disagreed for one grade: Ryan Smith and co-authors found that NOREM 02 galled, a tearing wear between metal surfaces sliding under load, at 70 MPa against 200 MPa for Stellite 6, and developed a nitrogen-rich alloy, Nitromaxx, that matched Stellite 6 up to 350 °C in their own tests.
LessonAn alloy that does well in one wear test can do badly in another, so a replacement seat material has to be tested under the loads and temperatures it will meet.
Credit
EPRI (H. Ocken) and AECL (E. V. Murphy, I. Inglis); R. Smith, M. Doran, D. Gandy and co-authors, Materials & Design, 2018
2006
A wrecked impeller copied
In November 2006 a pump that builds hydraulic pressure to start a steam turbine at Owensboro Municipal Utilities in Kentucky reached Rotating Equipment Repair (RER) of Sussex, Wisconsin, with half of its impeller's shroud destroyed and the tops of the vanes sheared off. A pattern maker rebuilt one damaged vane in clay, Advanced Design Concepts scanned it with a measuring arm, and its modeler copied the vane around the axis in CAD and rebuilt the top shroud from the bottom one, in a day and a half; Signicast made three investment castings in a week, and the pump shipped by the end of Thanksgiving Day. RER's engineer Eric Kirschling estimated that the method saved 6 to 10 weeks.
Credit
RER, Advanced Design Concepts (Justin Ebbe and Chris Mulhall) and Signicast
2017
Printing a part nobody makes
In January 2017 a printed metal impeller, 108 mm across, went into a fire protection pump that runs continuously at the Krško nuclear plant in Slovenia, replacing one in service since the plant was commissioned in 1981 whose maker had gone out of business. Siemens's team in Slovenia reverse-engineered it from the old part and Siemens printed it at Finspång, Sweden; the plant operator, NEK, tested it with Siemens over several months, and an outside institute ran further tests, including a CT scan. Siemens called it the first successful commercial installation of a printed part in a nuclear plant, a claim World Nuclear News reported without checking.
Credit
Siemens and Nuklearna Elektrarna Krško (NEK)
2022
A refinery qualifies printed parts
In early 2022 Shell fitted 3D-printed impellers, made with Baker Hughes, to a centrifugal pump at its Energy and Chemicals Park in Rotterdam, after the two companies worked out a repeatable way to qualify suppliers of printed parts. In 2023 DNV, a certification body, certified a gate valve that Shell and Bonney Forge printed in 316L stainless steel (UNS S31603), also meant for the Rotterdam site.
Credit
Shell (Angeline Goh's team), Baker Hughes and Bonney Forge; certification by DNV
Watch for
Printed impellers for offshore pumps
KSB says it printed 14 impellers, each 409 mm across, in Inconel 625 for two high-pressure HVR pumps on Equinor's Snorre platform in the Norwegian North Sea, in about nine weeks where castings typically take six to eight months; the customer had asked to change the material from duplex steel. DNV certified KSB's printed impeller family in Noribeam 625 at its highest level, AMC3, in December 2023, but the lead times come from KSB staff.
Achieved whenEquinor or another party outside KSB reports the pumps back in service with the printed impellers, with their hours of running.
Claimed by
KSB, in Formnext Magazine, June 2026
Open
01
Rebuild worn vanes from scans
Software that takes a 3D scan of a worn or broken impeller, fits the surfaces of the parts that are still whole, and rebuilds the missing vane tips, edges and shroud into a CAD model a shop can machine or print from, with a map of every place it had to guess.
The briefCredit · unclaimed02
How long printed impellers last
A public endurance test in which reverse-engineered impellers made three ways, printed in 316L stainless steel by laser powder bed fusion, printed in a carbon-fiber-filled polymer, and the maker's cast original, run side by side in identical small clean-water pumps for 2,000 hours, with flow, head and power logged and each impeller scanned and weighed every 500 hours.
The briefCredit · unclaimed
Open 01 · the brief
Rebuild worn vanes from scans
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
Software that takes a 3D scan of a worn or broken impeller, fits the surfaces of the parts that are still whole, and rebuilds the missing vane tips, edges and shroud into a CAD model a shop can machine or print from, with a map of every place it had to guess.
Why it is still open
Shops do this by hand. In 2006 a repairer had a pattern maker rebuild a damaged vane in clay before it was scanned, and copied that one vane around the axis in CAD (Digital Engineering, 2008). Stanasel and Buidos (2016) set out the manual steps for modeling a pump impeller from a worn one. Maiorova and co-authors (2026) described a multi-step scanning procedure, tested on an impeller, whose final model held "within the range of ±0.05 mm", without comparing it with an unworn original. Jet engine repair has related methods: a 2019 Georgia Institute of Technology study by Myong Joon Kim, Maxwell Praniewicz, Thomas Kurfess and Christopher Saldana measured worn blades on the machine and bent the blade's design model to fit them before machining a repair, but that work starts from the maker's model, which a pump shop copying an obsolete impeller usually does not have. No published method was found that rebuilds lost impeller vane geometry automatically and reports its error against an unworn twin.
digitalengineering247.comdoaj.orgjournals.uran.uapar.nsf.gov
Built with what exists
A structured-light scanner, such as a Revopoint or Creality model, or a measuring arm; Open3D to align scans and compare surfaces; CadQuery to build the CAD model from code, using the fact that the vanes repeat around the axis; a coding agent that checks its own work, such as Inkling 2 Small (see the Note "Checked without being asked").
First step
Buy a new and a worn impeller for the same small pump model from a supplier or scrap dealer, scan both, and measure the gaps between them in CloudCompare.
Done looks like
20 worn impellers from at least three pump models, each with an unworn twin or the maker's drawing; the rebuilt vanes lie within 0.2 mm of the unworn reference over at least 95% of their surface; the time taken compared with a skilled modeler doing the same parts; scans, code and results published.
Open 02 · the brief
How long printed impellers last
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public endurance test in which reverse-engineered impellers made three ways, printed in 316L stainless steel by laser powder bed fusion, printed in a carbon-fiber-filled polymer, and the maker's cast original, run side by side in identical small clean-water pumps for 2,000 hours, with flow, head and power logged and each impeller scanned and weighed every 500 hours.
Why it is still open
Published comparisons are short performance tests on one pump. Popa and co-authors (2024) found that impellers made by laser melting matched a machined one on a test bench, and Papageorgiou and co-authors (2025) measured about 15% higher efficiency for a printed carbon-fiber polymer impeller than for the cast iron original. Users in industry, such as NEK at Krško and KSB for Equinor, report qualification tests but publish no running data. The longest run found is a printer maker's case story: Nikon SLM Solutions says a printed Inconel 718 impeller for Wärtsilä ran for more than 500 hours in a laboratory engine with no wear or defects found (2025). C. Jiang, B. A. Fleck and M. G. Lipsett (2020) printed plastic copies of a slurry pump's steel impeller on purpose so that garnet slurry would wear them quickly, to study where wear occurs, not how long printed parts last. A search of the Directory of Open Access Journals and the web for endurance tests of printed impellers against cast ones found none.
doaj.orgdoaj.orgworld-nuclear-news.orgnikon-slm-solutions.comdoaj.org
Built with what exists
Three to six identical small end-suction pumps; a structured-light scanner; a metal printing service for 316L parts and a desktop printer for carbon-fiber nylon or PPS; a lathe and a balancing service for finishing; a test loop on the open-source design of Semenzin and co-authors (HardwareX, 2020), with a flowmeter, pressure sensors and a power meter logged by a Raspberry Pi; Open3D to compare scans.
First step
Buy two identical small pumps, scan one impeller, and order one printed copy in each material.
Done looks like
Each impeller runs 2,000 hours; head and efficiency at the best-efficiency flow measured every 250 hours; the change in head, efficiency and mass of every impeller published with the raw logs and scans.