A real part, redesigned for a digital route
This was a group project in MEEN40030 Manufacturing Engineering II, part of my MEngSc in Robotics and Intelligent Manufacturing at UCD. Five of us worked on it in the autumn trimester of 2026. We took a real component, redesigned it for a digital manufacturing route, and worked out the process, inspection, cost, sustainability and Lean improvements for a fictional company.
The company was AeroForge Engineering, an Irish Tier-2 aerospace supplier. It owns a metal laser powder bed fusion (LPBF) printer, a CNC mill, a CMM and a surface profilometer, and its customers buy in small batches. We handed in a 6-page report on 9 November 2026 and presented on 19 November.
The component was a bleed-air elbow from a turboprop air-conditioning system. It carries air at 230 to 250 °C and 220 kPa. The original is three parts: a bent AISI 321 stainless pipe and two machined flanges, joined by two TIG welds.
It cracked at the weld
A published failure study (Raghavendra et al., 2022) found one of these elbows cracked by fatigue after 7 years. The crack started at the flange-to-pipe weld, where the wall stepped abruptly from 4.2 mm to 1.4 mm. A misplaced, softer weld made it worse.
Spares are slow too. Welded ducts take 45 to 192 days to arrive after ordering.
One piece, no welds
We printed the whole elbow as one piece in 316L stainless steel, with a gradual transition at the flanges instead of a sharp step. The wall is 3 mm for fatigue margin. The printed inner surface can't be machined and is rougher, and pressure stress is only about 4 MPa, so fatigue decided the wall thickness, not pressure.
The geometry is a 50 mm OD, a 110 mm bend radius, 6 mm flanges with six bolt holes and a groove for a Parker metal E-ring seal. On our like-for-like comparison the printed part is about 7% lighter, 1.395 kg against 1.504 kg.
The certification route is EASA Part 21G production with the EASA CM-S-008 guidance for printed parts (expected Class C). Proof and burst tests run at 1.5 times and 3 times operating pressure, with AS9100D quality management and AS9102 first article inspection.
What the rest of the team worked out
Teammates covered materials, build layout and simulation, the baseline and its justification, CAD and CAM, and inspection and safety. In short: 5 elbows go on each build plate. A build takes 112 h, 103 h of it printing, and uses 8.3 kg of powder including 0.935 kg of supports. A distortion simulation predicted a maximum displacement of about 0.42 mm.
After printing comes outsourced stress relief (72 h with transport), wire EDM to cut the parts off the plate, support removal, then CNC finishing of the flange faces, bolt holes and seal groove. That is 44.82 min per elbow, 30 min of it setup. Every elbow gets CMM, surface roughness and pressure and leak tests, 45 min in total. The first 3 also get a CT scan and a full first article inspection. We assumed a 5% reject rate.
The first 5 elbows are ready about 202 h, roughly 8.4 days, after manufacturing starts. The main safety hazard is the fine 316L powder, which can be inhaled and can ignite. It is controlled with an ATEX wet-separator vacuum, P3 respirators and oxygen-depletion alarms for the argon, at about €6,500 a year.
About €1,657 per elbow, and where it goes
My part was cost, sustainability and Lean. At 80 elbows a year the model gives the figures below.
| Item | Basis | € per elbow |
|---|---|---|
| Printer time | €1.5M over 8 years, 6,000 h a year = €31.25/h | 700 |
| Argon | 128 m³ per build at the BOC cylinder price | 372 |
| Labour | 4.2 h at €35.32/h (CSO) | 150 |
| Powder | 1.66 kg at €87.50/kg | 145 |
| Electricity | 453 kWh at €0.2552/kWh (Eurostat) | 116 |
| Heat treatment | €476.60 per plate ÷ 5 | 95 |
| CNC | 44.82 min at €40/h | 30 |
| Scrap and rework allowance | 3% and 2% | 49 |
| Total | about 1,657 |
The welded comparison is a market selling price of €914 to €984. It includes the supplier's profit, so treat it as a rough benchmark, not a like-for-like cost.
The gap closes in two steps. Buying argon in bulk (wholesale about US$0.64/kg) brings the cost to about €1,300. Selling spare printer time as well brings it to about €1,070, which is 9 to 17% above the welded price. A change of 20% either way in the printer rate and argon price moves the total between €1,440 and €1,870.
Mostly electricity and argon
Scope 1 is about zero. The site is all-electric and argon is inert. Scope 2 comes from the printer: 22 kW × 103 h is 2,266 kWh per build, or 453 kWh per elbow. At the SEAI 2024 grid factor of 0.224 kg/kWh that is about 102 kg CO₂ per elbow, or 8.1 t a year. This counts printer electricity only.
On the plus side, no welding consumables are needed, about 87% of unused powder is reused and scrap is recycled. The fixes we proposed were a certified renewable tariff, bulk liquid argon, re-orienting the part to need fewer supports, and tracking powder reuse.
The biggest waste is an idle printer
The printer is busy only 1,790 of 8,760 h a year, about 20%, and that is also the biggest cost driver. Other wastes: 72 h waiting at heat treatment, CNC setup taking two-thirds of the CNC time, 60 min of manual support removal and 100% inspection.
Next steps are to sell spare printer hours, cut CNC setup time, re-orient parts and move to sampling inspection after the first articles. Batching heat treatment in 20s would cut it from €95 to €24 per elbow. We rejected it because it slows delivery, and speed is the selling point.
I also set up the shared inputs spreadsheet with its Baseline tab and collected the data each member needed. I made the cost, sustainability, Lean and conclusion slides, and wrote the executive summary and the AI-use appendix.
Use it for spares and urgent orders
Printing removes both welds, which is where the original cracked, and cuts lead time from 45 to 192 days to about 8.4 days. It costs more today, but it gets close to the welded price with bulk argon and better printer use. We recommended it for spares and urgent orders, where speed and reliability matter more than unit cost.
Limits: no physical part was made or tested. This is a design and planning study, and the cost figures are estimates. The printer price and the CNC rate are the weakest inputs.
Lessons
Cost per part depends far more on how busy the machine is than on the material. A cost and a selling price are different things, so say which one you are comparing. In a five-person report, getting everyone's numbers to agree (print time, masses, argon volume) took as much work as the analysis, and fitting it into 6 pages forced hard cuts.
On AI: it sped up finding sources and checking sums, but every figure had to be checked against the original source. The group declared all AI use in an appendix.