Apex-iCore · Heat Boomerang
APEX-iCORE · 3D model

A machine built to harvest waste heat

The clash-free Apex-iCore build: tunnel, airlocks with split doors, hook rail and trolleys, settling box, four scrap pits, ducts, dampers and bag filter. Drag to orbit, scroll or pinch to zoom, right-drag to pan. Press Animation to run one 10-second index cycle on loop (real cycle: 2 minutes).

Loading the 3D model…

Apex-iCore · Demo video
The Apex-iCore machine walkthrough · 2:17Open the 3D foundry tour → · Read the handbook ↓
Apex-iCore · Yuva Yodha Challenge 04 · Smart Manufacturing

Apex-iCore Handbook

Hot castings give their heat to air. The air preheats the scrap, and the scrap goes straight into the induction furnace. The hourly pit cycle is the Heat Boomerang.

Furnace electricity avoided52.5 kWh/t
Delivered into the melt39.35 kWh/t
Net saving a year (1 t/h)₹20.3 L
Build cost · payback₹48 L · 2.4 yr
CO₂ avoided a year174 t
Scale it to 5 t/h
₹1.02 crore saved every year
12,59,030 kWh of furnace electricity avoided · ≈ 870 t CO₂ a year · payback under 2 years

Team: R Priya Dharshini and M S Jayamouthikhan. Open the 3D foundry tour.

3D · Foundry tour

Walk through the foundry in 3D

15 narrated chapters, furnace to furnace, with temperatures from the Apex-iCore physics model. Needs an internet connection for the 3D engine.

01 · The design

How the machine physically works

Hot castings preheat the next charge of scrap. Air carries the heat, the scrap stores it, and the furnace uses it. No fuel, no steam, no pressure.

1Shakeout500°C castings, 30 sets of 33.3 kg an hour
→
2Jib + airlockset hung on a hook, pedal opens the outer door
→
3Tunnel 20 m140 min; air 35 → 398°C, counterflow
→
4Settling boxsand drops out of the air
→
53 pits + spareair flows down through 1 t baskets
→
6Hopperhot basket ≈ 326°C into the furnace
Castings in → out500 → 120°C
Air in → out of tunnel35 → 398°C
Air through the pits396 → 80°C
Hot basket to furnace≈ 326°C
Exhaust36–84°C

Air through the three pits in series: 396 → 253 → 125 → 80°C.

Subsystems

SubsystemKey figures
Recovery tunnel20 m, 0.9 × 0.9 m inside, 300 mm rock wool
Fins, hot 10 m100 × 2 mm plates every 50 mm, ≈ 3× the wall area
Stirring fans5 fans, 2 m/s, tilted 12° toward the hot end so the air corkscrews along the tunnel
Hook loop≈ 77 trolleys on a rail inside the tunnel; one push every 2 min moves the line 0.286 m
Airlocks2 doors each, never open together; tunnel held at −8 Pa
Settling box9 m/s duct opens into a 0.3 m/s box, so sand falls out
Scrap pits4 pits, 1.6 × 1.6 × 2.6 m, sand-trough seals; 3 in series, 1 spare
Dampers12 manual (H, L, X); 6 levers each hour
ExhaustBag filter, 0.75 kW fan with VFD, whole line under suction

Operating cycle: two rhythms

Every 2 minThe tunnel
0:00Inner doors open; the line moves one hook
0:15Pusher returns; inner doors close
0:30Exit door opens; a set rolls out at 120°C
0:40Unloader bins the castings, returns the hook
1:00Jib hangs the next 500°C set on a hook
1:40Pedal: outer door opens, trolley enters
Minute of the hourThe pits
0–16 damper levers: air to the next basket
1–3Crane lifts the hood off the hottest pit
3–5Hot basket, ≈ 326°C, to the hopper
5–6Lid opens, bottom drops, scrap falls in
6–9Empty basket to the yard; hood back on
9–55Charger feeds hot scrap to the furnace
11–25Yard refills the empty basket
30–36Refilled basket into the spare pit

Pit rotation: the air moves, not the baskets. Each hour every role (hottest, middle, coolest, spare) shifts one pit along. That takes 6 levers and 1 minute, instead of 3 hot basket lifts. From a cold start the scrap warms from 28 to 42 kWh/t over a 16-hour day; the year average is 41.1 kWh/t into the scrap.

The physics behind 52 kWh per tonne

Heat balance per tonne of castings, Apex-iCore model, kWhReleased by castings61radiation ≈ 38 · convection ≈ 23Lost through insulation−6Picked up by the air54Basket steel, pits, exhaust−12Into the scrap (year average)41.1Delivered into the melt39.35Electricity avoided (÷ 0.75)52.5
Figure 1 · Heat balance per tonne of castings, kWh, from the Apex-iCore model. 1D transient: 70 segments, 4 s steps, 3 simulated weeks with real shifts; the energy ledger closes. Uncertainty ±20–25% until the pilot measures it.
Counterflow, matched
  • Castings 0.17 kW/K against air 0.15 kW/K.
  • So the air leaves within about 100 K of 500°C.
Why downflow
  • Bed friction 0.1–0.2 Pa against hot-air buoyancy 0.3–2 Pa.
  • Gravity steers the air, so it must flow down through the scrap.
Settling (Stokes)
  • 100 µm sand falls at 0.44 m/s.
  • The box air moves at 0.3 m/s, so the sand drops out.
Fans and pressure
  • ≈ 1,000 Pa budget, 700 of it the filter.
  • Exhaust fan ≈ 0.29 kW; all running power 6,294 kWh a year.

Safety: every failure cools it down and lets air in

HazardDetected byPrevention and response
Wet scrapYard check, weightPreheat dries it before the melt; baskets drain
Oily scrapCO/VOC sensor, filter ΔpSegregate at the yard; air capped at ≈ 400°C
Over-heatThermocouplesAlarms at 420°C in the tunnel, 110°C at the filter
Fan stopsManometer, motor currentPedal locked; doors stay shut; castings cool
Stuck damperPit in/out temperaturePanel lamps show the pattern; freed by hand
Pit fireOutlet hotter than inletClose its H damper to starve it of air
Door crushDoor position switchesInterlocks, guard rails, E-stop
Power cut—Nothing pressurised; doors spring shut
PLC decides
  • Airlock sequence, pusher, suction loop, alarms and logging.
AI advises
  • Change-over timing, anomaly warnings, savings reports.
  • Never overrides an interlock.
Human override
  • Manual dampers, pusher hand valve, key-switch doors, E-stop.
  • Runs with the PLC off.

Interlocks: one airlock's two doors never open together · the pedal is ignored without suction · the pusher runs only with the inner door fully open.

02 · Money and impact

₹20.3 lakh a year from one 1 t/h foundry

A year at 1 t/h, 2 shifts, 300 days, at the Tamil Nadu HT tariff.

Net saving every year₹20,30,895
Simple payback≈ 2.4 yr
Simple return a year≈ 42%
CO₂ avoided a year174 t
10-year NPV at 10%₹77 L
A year at 1 t/hValue
Iron melted4,800 t
Heat delivered into the melt39.35 kWh/t
Furnace electricity avoided (÷ 0.75)2,51,806 kWh
Worth at ₹8.859/kWh₹22,30,844
Running power, 6,294 kWh−₹55,687
Extra demand charge−₹8,261
Maintenance−₹1,36,000
Net saving every year₹20,30,895
Build cost, ₹ lakhLowHigh
Tunnel, fins, fans11.516.2
Hooks, rail, pusher, jib3.14.8
Settling box, ducts, filter, fan3.56.0
4 pits, hoods, baskets, dampers9.913.5
Controls and sensors0.91.6
Site: foundations, wiring1.22.4
Engineering + contingency6.312.8
Airlocks + inlet fan≈ 1.5≈ 1.5
Total (mid ≈ 48)37.858.6

What moves the payback

CaseNet a yearPayback
Base case₹20.3 L2.4 yr
Tariff ₹7/kWh₹15.8 L3.0 yr
Tariff ₹10/kWh₹23.1 L2.1 yr
Furnace efficiency 0.85₹17.7 L2.7 yr
Furnace efficiency 0.65₹23.7 L2.0 yr
70% output₹13.6 L3.5 yr
One shift (estimate)₹8–9 L≈ 5.5 yr
Three shifts (estimate)≈ ₹30 L≈ 1.6 yr
5 t/h, 2 shifts≈ ₹1.02 Crunder 2 yr
Rule of thumb
  • Every ₹1/kWh of tariff moves the net saving by about ₹2.5 lakh a year.
Scale
  • 5 t/h: ≈ ₹1.02 crore a year, 12,59,030 kWh, ≈ 870 t CO₂ avoided.
  • 1 Mt of iron a year on Apex-iCore: ≈ 52 GWh and 37,000 t CO₂.
ADEETIE (BEE)
  • 5% interest subvention for micro and small units; foundry is covered and Coimbatore is a listed cluster.
  • ≈ ₹2.4 lakh a year less interest on ₹48 lakh.
  • 52.5 kWh/t is about 8.8% of a 600 kWh/t melt, so pair it with small measures to clear the 10% savings line.

The size of the problem

Heat in each tonne of castings, 500 → 35°C71 kWh
To melt iron (the iron itself needs 396)560–600 kWh/t
Melting power at 1 t/h₹2.5 Cr/yr
CO₂ from that power1,970 t/yr
Castings made in India, 2024–2515.86 Mt
Foundries, about 90% MSMEs≈ 5,000

Tariff: ₹7.50/kWh + 25% time-of-day on 8 of 16 hours + 5% tax = ₹8.859/kWh (TN HT) · demand charge ₹608/kVA/month · CEA grid factor 0.710 t CO₂/MWh · industry figures from IFC and IFEX 2026. Build costs are planning estimates; replace them with quotes.

03 · Materials, fan speed and time in the tunnel

One tunnel, twelve alloys: set the fan, keep the strength

Locked design: 20 m tunnel, 0.9 × 0.9 m, 1 t/h, 33 kg sets, 28 mm equivalent section.

Fan speed and time inside the tunnel

Main fan rated 0.25 kg/s (≈ 785 m³/h) at 50 Hz; flow is set on the VFD. The conveyor is the same for every material: 0.143 m/min, one 33 kg set every 2 min, 0.29 m hook pitch, 140 min inside. Airflow matches each material's heat-capacity rate (about 1.05×).

MaterialEntersMain fan (kg/s · m³/h · Hz)Stirring fansReaches 150°CExitsHeat to scrap
Grey iron500°C0.161 · 508 · 32 Hz2 m/s (25 Hz)116 min111°C43.0 kWh/t
Ductile iron500°C0.161 · 508 · 32 Hz2 m/s (25 Hz)116 min110°C43.0 kWh/t
High-Cr white iron500°C0.153 · 481 · 31 Hz2 m/s (25 Hz)117 min113°C41.2 kWh/t
Carbon steel (WCB)500°C0.161 · 508 · 32 Hz2 m/s (25 Hz)120 min117°C42.3 kWh/t
Ni-Cr-Mo alloy steel500°C0.161 · 508 · 32 Hz2 m/s (25 Hz)120 min117°C42.3 kWh/t
Stainless CF8M500°C0.152 · 478 · 30 Hz2 m/s (25 Hz)125 min127°C39.5 kWh/t
Duplex CD4MCuN500°C0.151 · 476 · 30 Hz2 m/s (25 Hz)124 min125°C39.5 kWh/t
Aluminium A356400°C0.238 · 748 · 48 Hz4 m/s (50 Hz)120 min107°C42.9 kWh/t
Aluminium LM6400°C0.235 · 740 · 47 Hz4 m/s (50 Hz)120 min106°C43.0 kWh/t
Gunmetal LG2450°C0.113 · 354 · 23 Hz2 m/s (25 Hz)116 min114°C27.2 kWh/t
Leaded brass450°C0.114 · 358 · 23 Hz2 m/s (25 Hz)115 min112°C27.7 kWh/t
Aluminium bronze450°C0.129 · 406 · 26 Hz2 m/s (25 Hz)113 min108°C31.5 kWh/t

Air speed in the tunnel: 0.17 m/s (cold end) to 0.38 m/s (hot end) for irons; 0.26 to 0.55 m/s for aluminium. Aluminium needs more air and 4 m/s stirring because its emissivity is low (~0.2).

Metallurgy check: strength and properties unchanged

MaterialCooling rate to 150°CTime in risky rangeEffect on strengthVerdict
Grey and ductile iron181°C/hNone; entry well below eutectoid (~730°C), matrix fixed in mouldNoneYes
High-Cr white iron179°C/hNoneNone; hardening done afterwards anywayYes
Carbon steel (WCB)175°C/hNone (below A1)None; normalise/temper afterwards per specYes
Ni-Cr-Mo steel175°C/h27 min in 375–500°C temper-embrittlement bandEmbrittlement needs hours; Q&T afterwards resets itYes (Q&T after)
CF8M stainless167°C/h5 min in 450–500°CSensitisation needs hours; solution-anneal afterwardsYes
Duplex169°C/h19 min in 400–500°C475°C embrittlement needs hours; solution-anneal afterwardsYes
Aluminium A356125°C/h (from 400°C)97 min in 200–400°C over-ageing bandLost only if supplied as-cast or T5; T6 re-dissolves itT6 orders only
Aluminium LM6125°C/hNone (not heat-treatable)NoneYes
Gunmetal, leaded brass155–157°C/hNoneNone; slower cooling reduces residual stressYes
Aluminium bronze160°C/hNone (entry below 565°C eutectoid)NoneYes
Magnesium AZ91––Fire riskNo
Zn, Pb, Sn alloys––Too low-melting for this processNo
Gentler than the floor
  • The tunnel takes 133–137 min to reach 120°C vs 75–98 min on the floor (about 1.6–1.8× slower), so cracking risk drops.
Thin-vs-thick difference
  • 131°C in the tunnel vs 250°C on the floor for irons; stress falls from 0.5–0.6 to 0.25–0.3 of UTS.
Tightest cases
  • Stainless, aluminium and gunmetal reach about 0.5–0.6 of UTS in the tunnel (0.8–1.0 on the floor), so keep section changes gradual.
Hooks
  • Aluminium at 400°C is very weak: use mesh trays, no piercing hooks, no heavy hanging loads.
Leaded alloys
  • Lead is liquid above 327°C; avoid shock or heavy hook load until below that.

Thick castings (75 mm equivalent): half-throughput mode

At normal settings a 75 mm section exits at 195–214°C and recovery falls to ~33 kWh/t. For thick castings run half throughput with half the airflow.

MaterialFan (kg/s)Hook intervalTime insideReaches 150°CExits
Irons0.077–0.0814 min280 min243–245 min123–126°C
Steels0.0814 min280 min253 min132°C
Stainless / duplex0.0764 min280 min266–269 min142–144°C
Aluminium0.118–0.1194 min280 min245 min117–119°C
Copper alloys0.056–0.0654 min280 min236–246 min121–128°C

Thin sections (10 mm) exit at 59–73°C and recover about 45–48 kWh/t. Below about 20 Hz use a damper or bypass instead of running the motor slower.

Panel settings and limits

On the panel
  • Recipe: alloy, fan Hz, stirring m/s, hook interval. Exit IR pyrometer confirms ≤ 150°C.
Operating rules
  • Maximum entry temperature about 550°C.
  • One alloy family at a time, with separate scrap baskets.

Limits: model predictions, not measurements. Metallurgical time limits are handbook order-of-magnitude values, not a kinetic simulation. Stresses use room-temperature modulus and UTS (conservative, especially for aluminium and copper). Actual section thickness decides which mode to use.

04 · Pilot, team and vision

From model to measured savings in 15 months

Maturity today: system TRL 3, parts TRL 8–9. Designed, modelled and simulated; not yet tested with real castings and scrap.

  1. Months 0–3 · Bench test2.5 m tunnel section with real castings: radiation, fins, fans.
  2. Months 2–4 · Single hot pitReal scrap and a 400°C heater: downflow, bed pressure drop, sand seals.
  3. Months 3–6 · Pilot designPartner foundry, layout, ADEETIE energy audit and detailed project report.
  4. Months 6–12 · Pilot buildFull Apex-iCore at 1 t/h in a Coimbatore foundry.
  5. Months 12–15 · MeasurePaired heats on the furnace meter, verified by an auditor.
Success criterionTargetModel
Furnace electricity saved≥ 45 kWh/t52.5
Heat delivered in scrap≥ 35 kWh/t39.35
Casting exit temperature≤ 150°C≈ 120
Tunnel suctionbelow −3 Pa always−8 Pa
Uptime, production hours≥ 95%—
Payback on measured data≤ 3 years2.4

The narrated 3D tour walks through a 1 t/h foundry in 15 chapters, about 12 minutes: furnace, pouring and shakeout, the jib, pedal and airlocks, a thermal camera inside the tunnel, the fins and tilted fans, the settling-box swirl, every pit part, the hourly crane change-over and a live energy meter. Every colour and temperature comes from the physics model.

▶ Open the live 3D tour

Team

Controls, data and businessR Priya Dharshini
  • PLC interlocks, VFD and sensors
  • Data logger, digital twin, M&V savings reports
  • Costing, payback and ADEETIE finance
  • Partner foundry and pilot
Thermal and mechanicalM S Jayamouthikhan
  • 1D physics model and heat balance
  • Tunnel, fins, fans, hook loop and airlocks
  • Scrap pits, baskets, dampers and frame
  • 3D foundry tour

Why Apex-iCore should continue

Market
  • ≈ 5,000 Indian foundries, 90% MSMEs, 15.9 Mt a year; clusters from Coimbatore to Rajkot.
Next uses
  • Forgings heating billets, heat-treat loads, die-casting, ceramic kilns.
Product path
  • Manual build → motorised, PLC-controlled version → digital-twin service.
  • Licensed to cluster fabricators, ESCO or ADEETIE-financed.

Castings leave the mould at 500°C. Apex-iCore sends that heat back: 52 kWh/t, ₹20 lakh a year, about a 2.4-year payback, built locally.