Most robot projects that stall in Indian factories do not stall because of the robot. They stall because the shed was never checked: the floor is thinner than anyone remembered, the spare 3-phase capacity is a few kilowatts, and the earth pit is a pipe behind the panel that nobody has measured since 2011. Two questions decide whether your building is ready — can this floor take the loads, and can this supply run the controller without nuisance trips.
What the robot actually asks of the building
The robot vendor’s product manual is the only document that matters here, and it is usually ignored in favour of the integrator’s quotation. Read three sections of it before you sign anything: Requirements, foundation, Loads on foundation, and the controller’s Technical data. Everything below is a translation of those sections into Indian site terms.
The order of work is fixed. Measure the floor and the supply first, price the robot second. If you do it the other way round you end up buying a robot that fits your budget but not your building.
The floor: a robot pushes harder than it weighs
Take the ABB IRB 760, a 450 kg payload robot weighing 2,300 kg. Its manual gives the forces the base puts into the foundation during normal operation as 26.7 kN downward and 3.6 kN upward, ±9.1 kN horizontally, and a bending torque of ±28.9 kNm. On an emergency stop those become ±17.7 kN horizontal and ±38.5 kNm of bending torque.
The robot’s own weight is about 22.6 kN. So the vertical endurance load is already larger than the weight, and the bending torque is the number that actually sizes the anchors. As a rough check, a 28.9 kNm moment resisted across a base bolt pattern about 0.9 m wide is a couple of roughly 32 kN, split across the bolts on the tension side. With six bolts carrying tension and a safety factor of three, each anchor is being asked for something like 16 kN of pull-out — on a base weighing 2.3 tonnes.
Two conclusions follow. First, never let an anchor supplier size the bolts from the robot’s weight alone. Second, the floor is being asked for a fatigue load, not a static one: the coupling reverses direction every cycle, tens of thousands of times a month.
What thickness and grade
ABB does not publish a slab thickness, because it is the pedestal and slab together that must be stiff enough. What ABB does publish for this robot is a foundation flatness of 0.3 mm at the mounting surface and a minimum resonance frequency of 15 Hz for the robot mass sitting on the pedestal. The manual warns that a lower resonance frequency “may affect the manipulator lifetime”. That is the real specification: a stiff, level base, not a thick one for its own sake.
FANUC’s published figure, as restated in an integrator’s installation note, is more concrete: rebar-reinforced concrete, minimum 4000 psi, at least 200 mm thick, with 200 mm anchor hole depth. Two cautions on that number. It is a floor for a mid-size robot, not a universal rule for a 2.3-tonne palletiser. And 4000 psi is a cylinder strength, about 27.6 MPa. Indian cube grades are measured differently — at the usual 0.8 cylinder-to-cube ratio that is roughly 34 MPa cube, which is M35 territory, not M25. If a contractor quotes you “M25, same as the shed floor” for a 4000 psi requirement, that is a real gap. Ask for 28-day cube test results rather than a grade name.
For Indian sheds the practical issue is that many factory floors are 100–150 mm of plain cement concrete, sometimes with a mesh, laid over compacted murum. That is fine for a forklift and wrong for a robot base. The usual fix is to cut out a square under the base, dowel into the surrounding slab, and pour a 250–300 mm reinforced pad, or to build a steel riser that spreads the load. Which one is cheaper depends on how much you have to break.
Level, vibration and joints
Three things ruin an otherwise good pad. The first is levelness: 0.3 mm across the base is a machining tolerance, not a masonry tolerance, so the pad needs a steel base plate set and grouted, not shimmed on a trowelled finish. The second is vibration from a neighbouring press, power press or generator; if the floor rings when a press strokes, the robot’s resonance check will fail and the finish on your welds or your pick accuracy will suffer. The third is expansion joints running under the base — a joint is a hinge, and a hinged foundation moves.
If your cell shares a bay with a press line, an isolated block with a gap around it is often the only reliable answer. That is a structural engineer’s call, not the robot integrator’s, and the cost belongs in the project budget before you compare robot prices.
The floor is only one half of the footprint question. Reach, clearance, guard position and service access are covered in the robot cell layout guide, which assumes the building can take the cell in the first place.
Anchors: the part everyone gets wrong
There are two families and they fail in different ways.
| Mechanical wedge anchor | Chemical (resin) anchor | |
|---|---|---|
| How it holds | Expansion against the hole wall | Bonded to the concrete along the embedment |
| Sensitive to | Hole diameter, hole depth, edge distance, slab thickness | Hole cleaning, concrete moisture, curing temperature, cure time |
| Vibration and reversing loads | Can loosen; re-torque is part of maintenance | Better under fatigue if correctly installed and cured |
| Typical use in a robot cell | Light robots, temporary installs, steel risers | Standard choice for a fixed industrial robot base |
| Failure mode if rushed | Pulls out with a cone of concrete | Spins or pulls out because the hole was dusty or the resin was cold |
Whichever you use, four rules apply on an Indian site. Drill to the depth the manual states — 200 mm is the FANUC figure quoted in that integrator note — and use a depth gauge, not the drill bit’s flutes. Clean the hole properly; resin will not bond to dust, and a blower plus brush is not optional in a shed with concrete grinding dust in the air. Respect the cure time, which lengthens in cold weather and in monsoon humidity. And torque the anchors to the manual’s figure only after the cure, then record it; a base that was torqued while the resin was green will be loose within a month.
One more Indian reality: if the pad is 250 mm and the anchor needs 200 mm of embedment, you have almost no margin below the anchor. A hole that breaks through the underside of a thin pad anchors nothing. Measure the pad, not the drawing.
Power: single phase or three phase, and why it matters more than the robot
This is where cobots and industrial robots part company completely.
A Universal Robots arm is specified for a single-phase 100–240 VAC supply, and UR’s own guidance is 100–240 VAC, 50–60 Hz, single phase, with grounding. A UR10e draws 350 W at moderate settings and 615 W at maximum. In Indian terms: a cobot can genuinely run off a 230 V circuit, which is why small units in Pune and Coimbatore can install one without touching their sanctioned load.
An industrial controller is a different animal. The KUKA KR C4 technical data specifies a rated supply of 3 x 380, 400, 440 or 480 V, a permissible tolerance of −10 per cent below rated, mains frequency of 49–61 Hz, mains-side fusing of at least 3 x 25 A, and a system impedance up to the controller terminals of 300 milliohms or less. That last number is the one nobody checks. By my arithmetic, 415 V across 300 milliohms of loop impedance implies the supply can push roughly 800 A into a bolted fault at the terminals. If your feeder cannot, the controller will see dips on every acceleration and will fault at the worst possible moment.
The grounded-neutral problem
The KUKA manual is blunt: the controller may only be operated on a grounded-neutral supply, and if no grounded neutral is available, or the voltage differs from the rated values, “a transformer must be used”.
This catches Indian sites constantly. Plenty of small industrial connections are 3-phase, 3-wire, with a local earth pit and no distributed neutral — closer to what IEC calls a TT arrangement than to the TN system the controller expects. Add a diesel generator with a floating neutral and a 4-pole changeover that switches the neutral at the wrong moment, and you have a cell that runs perfectly on mains and faults every time the grid drops.
Three fixes exist, in order of cost: confirm the supply is genuinely grounded-neutral and bond it properly; fit a 4-pole changeover so the neutral is switched with the phases; or fit an isolation transformer with a grounded secondary for the cell. The transformer also fixes voltage mismatch and stiffens the supply, which is why it is the usual answer when the site is marginal. It is also the item most often missing from the robot quotation.
Voltage sag, generators and what trips first
A servo drive pulls a short pulse of current at every acceleration. The controller’s internal supply tolerates a narrow band — the KUKA figure is a 10 per cent drop — and nothing else on your feeder may notice a dip of that size. Lights flicker slightly, a motor grunts, and the robot faults out mid-cycle with an undervoltage alarm.
So the diagnostic order is: measure the voltage at the controller terminals during a production cycle with a logging meter, not a multimeter glance; measure loop impedance at the same point; then look at the feeder. Sharing a distribution board with a welding set, a large compressor start or a stone crusher is the usual finding.
On generators, two things matter. Sizing: the set has to carry the drive’s starting demand, not its running load, and the controller’s rated current on the nameplate — not the robot’s — is the number to give the genset supplier. Changeover: the controller should not ride through a break in supply. Expect the cell to stop, and plan for a controlled stop sequence rather than hoping the robot survives the transfer. I have not found a manufacturer document that promises ride-through, and you should not assume it.
RCCB and protective devices
The KUKA manual recommends, if a residual-current device is used, a 300 mA universal-current-sensitive, selective device per robot controller. That is Type B in IEC terms, not the Type AC unit your electrician stocks.
The reason is practical. Servo drives and long shielded motor cables produce leakage with a DC component and high-frequency content that a Type AC device reads as a fault. Fit a 30 mA Type AC RCCB on a robot circuit and you will spend a week chasing nuisance trips before someone replaces it. Fit the right device once, and remember that a 300 mA device protects equipment and cables, not people — the operator protection comes from the safety circuit and the guarding, not from this breaker.
Earthing: one bar, measured, recorded
IS 3043:2018, the Indian code of practice for earthing, is the governing document, and the robot manual adds its own connection requirements. Two habits separate a working cell from a mysterious one.
First, give the cell its own earth connection back to the main earth bar, sized for the fault current, rather than daisy-chaining onto whatever the previous machine left. The controller’s protective conductor, the panel’s earth bar, the guard frame and the machine frame should all land on a common reference, exactly as the manual describes for equipotential bonding.
Second, measure. An earth tester reading taken at the pit, written into the commissioning file with the date and the instrument used, is worth more than any certificate. If the reading drifts between seasons — and in dry months it will — you have a maintenance item, not a one-time job.
I am deliberately not quoting a target resistance in ohms. IS 3043 sets values that depend on the installation type, and the robot manufacturer sets connection rules rather than site resistance. Ask your integrator what they will accept, get it in writing, and measure against that.
What to check before you release the purchase order
| Check | How to measure | Typical pass condition | Who signs |
|---|---|---|---|
| Slab thickness and reinforcement | Core drill or existing drawing; rebar scanner | Meets the robot maker’s figure, or a new pad is designed | Civil engineer |
| Concrete strength | 28-day cube tests on the new pad | As specified, converted to Indian cube grades | Civil engineer |
| Floor level at the base | Straight edge and feeler gauge over the mounting face | 0.3 mm for a mid-size ABB, or grout a base plate | Integrator |
| Vibration from neighbouring machines | Vibration meter at the base location, machine running | No dominant peak near the robot’s resonance limit | Integrator and civil engineer |
| Supply voltage and phases | Logging meter at the controller point, one full cycle | Within the controller’s stated tolerance | Electrical contractor |
| Loop impedance | Loop impedance tester at the controller point | At or below the manual’s limit (300 milliohm for a KR C4) | Electrical contractor |
| Earth resistance | Earth tester at the pit, dry season | The figure your integrator accepts, recorded | Electrical contractor |
| Spare capacity | Compare cell demand with sanctioned load and contract demand | Headroom without a DISCOM enhancement | Owner |
The last row is the one owners underestimate. Adding a 10–15 kVA cell to a connection that is already close to its contract demand means an application to the DISCOM, not a new cable. That process has its own timeline and should start before the robot ships. The rules and limits differ by state and by DISCOM, so check your own bill and your own state’s regulations rather than relying on a number from a neighbouring unit.
What this costs in India
The honest answer is that civil and electrical site work is priced locally and cannot be benchmarked from a blog post. What I can give you is the quantity model and one worked example, so that you can sanity-check a quotation instead of accepting it.
The line items that appear on a site-readiness quotation are almost always these: breaking out and disposing of existing floor; a reinforced pad in concrete; a steel base plate, grouted; anchors; a new earth pit and conductor; a dedicated distribution board with the correct breaker; cable and glanding; and, where the supply is marginal or ungrounded, an isolation transformer. Labour and supervision are usually quoted separately from material.
For a worked example, take a 10 kg-class six-axis cell in an existing shed, on a 1.2 m x 1.2 m x 250 mm pad, with a dedicated 3-phase circuit about 15 m from the panel. Every rate below is an assumption I have written down so you can replace it with a local quote. I have not verified current market rates, and I am not presenting these as market prices.
| Item | Quantity | Assumed rate (₹) | Amount (₹) |
|---|---|---|---|
| Break out existing floor and dispose | 0.36 m³ | 3,000 / m³ | 1,100 |
| Reinforced pad, M25 or better | 0.40 m³ with wastage | 9,000 / m³ | 3,600 |
| Chemical anchors, M20 class | 6 | 1,200 each | 7,200 |
| Earth pit with chamber and strip to panel | 1 | 12,000 | 12,000 |
| 4-core 6 mm² armoured cable, glanded | 15 m | 450 / m | 6,750 |
| 4-pole 32 A Type B device with enclosure | 1 | 6,000 | 6,000 |
| Subtotal without transformer | ~36,700 | ||
| 10 kVA isolation transformer, if required | 1 | 45,000 | 45,000 |
| Subtotal with transformer | ~81,700 |
Two things this table is trying to show. The civil work is small money next to the robot; the electrical fix is not, and the transformer line is the one that decides whether a marginal site is viable. And the spread between the two subtotals — about ₹45,000 in this example — is roughly the cost of not checking the neutral before you order.
If you want a public anchor for the civil rates rather than a contractor’s word, the CPWD Delhi Schedule of Rates is the reference most Indian estimators start from, adjusted for city and year. I looked at a 2021 copy; treat it as a basis for comparison, not as your price.
For the wider budget picture — what an integrator’s quotation does and does not include — see robot integration cost in India. Site readiness is one line in that budget; this article is the detail behind that line. On the funding side, MSME automation subsidy in India covers what is claimable and what is not.
How to decide
Work down this ladder and stop at the first line that fails.
- Is it a cobot? Single-phase 230 V, a level floor and a sound earth may be all you need. Check the datasheet’s supply line, not the sales deck.
- Is it an industrial robot? Then you need 3-phase at the rated voltage, a grounded neutral, and a feeder stiff enough for the controller. If any of those is missing, budget the fix before you compare robot prices.
- Is the floor thinner than the manufacturer’s figure, or unreinforced? Cut and pour a pad, or design a riser. Do not let an anchor supplier talk you past this.
- Is the connection near its contract demand? Start the DISCOM process now; it is usually the longest item on the critical path.
- Is the site marginal on all three? Then a used robot on a new pad in a different bay is often a better project than a new robot in the wrong one.
Once the building passes, the next document you need is the energisation and acceptance sequence in the robot cell commissioning checklist. If you are still choosing between a cobot and a traditional arm, the cobot buying guide for India starts from the same supply question this article ends on.
What I could not verify, and what you should not take from this article: current local rates for civil and electrical work; the DISCOM limits and charges in your state; whether a specific controller will ride through a generator changeover; and the earth resistance figure your integrator will accept. Each of those has an owner who can answer it in writing, and that written answer is what belongs in the cell file.
Sources
- ABB, Product manual — IRB 760, document 3HAC039838-001, revision AB. Sections 2.2.2 “Loads on foundation” and “Requirements, foundation” (flatness 0.3 mm, minimum resonance frequency 15 Hz, force and torque table), and robot weight 2,300 kg.
- Hamrick Packaging Systems, Concrete Recommendations for Robot Riser (installation note restating FANUC’s concrete specification: rebar-reinforced, 4000 psi minimum, at least 200 mm thick, 200 mm anchor hole depth).
- KUKA, KR C4 operating instructions, document BA KR C4 GI V11 (2015). Section 4 “Technical data” and 6.3 “Connection conditions”: rated supply 3 x 380/400/440/480 V, −10 per cent tolerance, 49–61 Hz, system impedance ≤ 300 mΩ, mains fusing 3 x 25 A, grounded-neutral requirement, RCCB recommendation.
- Universal Robots, UR10e e-Series datasheet (08/2023): 350 W moderate / 615 W maximum power, tool flange EN ISO 9409-1-50-4-M6.
- Universal Robots, support article Good practices on powering energy to your cobot: 100–240 VAC, 50–60 Hz, single phase, grounding requirement.
- Bureau of Indian Standards, IS 3043:2018 — Code of practice for earthing (second revision). Cited for title and edition; the clause values were not reproduced here.
- Central Public Works Department, Delhi Schedule of Rates (2021 copy consulted): used only as the standard public reference for civil work rates, not as a source of any figure quoted above.
Common questions
Do I need a 3-phase connection for a robot cell in India?
It depends on the robot class. Cobots such as Universal Robots run on single-phase 100-240 VAC, so a 230 V circuit can be enough. Industrial controllers are three-phase: KUKA specifies 3 x 380/400/440/480 V for the KR C4, with mains fusing of at least 3 x 25 A. Check the controller nameplate, not the robot brochure.
How thick should the concrete floor be under an industrial robot?
FANUC’s guidance, as restated by a US integrator, is rebar-reinforced concrete of at least 200 mm with a minimum 4000 psi strength, and 200 mm anchor hole depth. Treat that as the floor for a mid-size robot only. ABB publishes flatness and stiffness limits instead of a thickness, because the pedestal design matters as much as the slab.
Can I bolt a robot to my existing shed floor?
Sometimes, but measure first. You need the slab thickness, whether it has reinforcement, its 28-day cube strength, and the distance to the nearest joint or edge. A 100-150 mm unreinforced floor typical of small Indian sheds is usually cut out and re-poured locally under the base.
What earth resistance does a robot cell need?
IS 3043:2018 is the governing Indian code and the robot maker’s manual sets the connection rules. Neither gives one number that fits every site. Ask the integrator for the target they will accept, get it measured with an earth tester and recorded before handover, and keep the reading in the cell file.
Why does the controller trip on undervoltage when other machines run fine?
Servo drives draw a short, heavy pulse and the controller tolerates only a small dip. KUKA allows a 10 per cent drop below rated voltage and requires system impedance at the connection point of 300 milliohms or less. A feeder shared with a welding set or a large motor start often fails that, even though lights and motors seem unaffected.



