Robot maintenance cost and schedule.

The arm itself asks for surprisingly little. Daily checks take minutes. A couple of times a year someone cleans the controller vents, checks cables and bolts and looks for oil at the seals. Once a year the encoder batteries get changed. Every few years the reducers get fresh grease. On a collaborative arm the list is shorter still. Universal Robots' service handbook lists monthly safety-function tests and periodic visual and bolt checks, with no grease or oil change in its inspection plan.

The cell around the arm needs more attention than the arm. Gripper fingers and vacuum cups wear. Sensors get knocked out of alignment. Fixtures collect chips. The vision lens gets dirty. Most of the maintenance tickets we see on a running cell are about the tooling, not the robot.

So budget in two buckets. One is the predictable stuff in the table below. The other is a downtime number that tells you how much it's worth spending to avoid a stop, which is where the rest of this page goes.

The single most common scheduling mistake we see is treating the calendar interval as the only one. The FANUC table says "3 years or 11,520 hours, whichever comes first," and the manual spells out that if you run more than 3,840 hours a year you shorten the intervals in proportion. A cell running around the clock hits 11,520 hours in well under two years.

The ABB and UR numbers come from different kinds of documents, so read them as a sense of the spread rather than a ranking. ABB's figures come from its 2014 launch material for the IRB 6700, which also claimed double the service intervals of the IRB 6640 it replaced. Always work from the product manual for your exact model and revision.

The procedure, from the FANUC M-20iD manual

1. Move the robot to the greasing posture for the axis you're doing, from the table in the manual.
2. Turn off controller power.
3. Remove the bolt and seal washer from the grease outlet.
4. Remove the bolt and seal washer from the grease inlet and fit a grease nipple.
5. Grease slowly with a manual pump, about one stroke every two seconds, until new grease pushes the old out of each outlet.
6. Release the remaining pressure from the grease bath using the manual's procedure, then close the outlet.
7. Clean up every bit of excess grease from the robot and the floor.

The manual's cautions are worth taking literally. It says to avoid air-powered grease pumps where possible and, if you must use one, to stay at or under the specified gun-tip pressure, 0.1 MPa or less for these reducers. It specifies one grease, and warns that another type may damage the reducer. And it warns that hot grease can eject suddenly when the outlet is opened, so bag the outlet and wear gloves and eye protection.

On that model the job takes 1,250 g of grease for J1, 830 g for J2 and 300 g for J3, with the wrist gearboxes on oil instead. Other models differ, so don't reuse these numbers on a different arm. What does carry over is the habit of looking at what comes out. Old grease that's dark and full of fine metallic particles is telling you about wear inside the reducer. Note it, and compare it with the next change.

The number that drives everything is margin per hour of the stopped cell. Use contribution margin, the price of a part minus its variable costs, not revenue. And be honest about recovery. If your plant runs five days with slack on Saturdays, a lot of lost hours come back at overtime rates. If it runs 24/7, nothing comes back, and every hour is pure loss.

For scale, Siemens' 2024 True Cost of Downtime report estimated that an hour of downtime at a large automotive plant costs $2.3 million, and that the world's 500 largest companies lose about $1.4 trillion a year, 11 percent of revenue, to unplanned downtime. It also put the top end for smaller manufacturers at $150,000 an hour. Siemens sells maintenance software, so read those as large-plant ceilings, not your number.

Your number is probably closer to ours. A single machine tending cell feeding one machine at a job shop might be worth a few hundred to a couple of thousand dollars an hour stopped. A cell that paces an assembly line can be worth far more, because the whole line stops with it.

Once you have the rate, it prices everything else on this page. It tells you whether a spare reducer is worth stocking, whether a 4-hour service response is worth paying for over next-day, and how much downtime a used arm can afford. Our used vs new robots guide runs that last question with the same numbers.

The failure probability is the soft number, and you won't find a good one on a datasheet. Use your own history if you have it, the robot maker's guidance if they'll give it, and a sensitivity check if you don't. If the decision flips between 2 and 5 percent, ask the maker or a parts house about an exchange program or a guaranteed lead time instead.

Standardizing on one robot family is what makes pooling work. Six cells with six different arms need six spare reducers. Six cells on the same model need one. That's an underrated argument for a plant standard, and it's one of the things we weigh when we recommend an arm. Our industrial robot arms guide covers the main families.

A starter spares kit for one cell

• A full set of gripper fingers or vacuum cups, and the fasteners for them.
• Each type of proximity sensor, photo-eye and part-present switch in the cell.
• Fuses, a spare I/O module and the relays used in the panel.
• Encoder batteries for the arm and controller, in date.
• A current program backup, stored off the controller, and the mastering data.
• The grease specified for your reducers, and a manual grease gun.

Service plans from robot makers and integrators vary a lot, so compare them on three things. First, response time, written as hours to a technician on site or on a call, not "priority." Second, what's included: PM visits, parts, travel, software updates, remote diagnostics. Third, who covers the cell, because a robot maker's plan covers the robot, and your gripper, vision and PLC may belong to someone else.

That third point trips people up. When a cell stops, the first hour is usually spent working out whether it's the robot, the tooling, the safety circuit or the machine it's tending. A contract that only covers the arm can leave you with three companies each saying it's not theirs. One owner for the whole cell is worth paying for if you can get it.

The break-even is the same downtime math as the spares. If a faster response saves 12 hours a year at $836 an hour, that's about $10,000 of value, and a contract that costs less than that is a good buy. If your cell can sit for a day without hurting anyone, a contract is mostly insurance for peace of mind, which is fine, but price it that way.

Training is the cheapest alternative. A maintenance tech who can read the alarm log, swap a sensor, restore a backup and re-teach a point fixes most stops before a service call would even arrive. Our training covers exactly that for the cells we build.

FANUC describes Zero Down Time as an application that collects robot data, looks for abnormalities in mechanical health such as reducers and servo motors, and sends alerts before failure so maintenance can be scheduled. Other makers and third parties offer similar monitoring. The idea is simple. A worn reducer usually draws more current and runs hotter for weeks before it breaks, and a trend line catches that where a person walking past doesn't.

The value lands right in the downtime math. In our example, turning an unplanned 40-hour stop into a planned 8-hour swap cuts the downtime cost from $31,660 to $4,910, about $27,000 per event. On a single cell with a low failure rate, that may not justify a subscription. On a fleet, or on a cell that paces a line, it often does.

There's a cheaper first step, too. Most controllers already log alarms, and many can log motor current or torque for a test program. Run the same short motion once a month and keep the numbers. It's not a product, and it won't catch everything, but a rising trend on one axis is the kind of early warning that turns a breakdown into a work order.

For the vision side of a cell, where lenses get dirty and lights age, our machine vision maintenance checklist covers the same idea for cameras.