AGV and AMR Caster Maintenance: What to Inspect, How Often, and When to Replace

Setting a cadence when no OEM interval exists

You will not find an OEM-mandated interval per component in this evidence set, so waiting for a published schedule stalls the routine. What the sources do share is the shape of the programme: a short visual check at high frequency, a longer hands-on inspection at lower frequency, and fastener torque on its own cycle¹⁸.

One maintenance guide frames the whole programme around a 90-day inspection¹; another sets weekly visual checks, monthly hands-on inspection and quarterly fastener torque¹⁸. A third suggests rotating wheels every three months and clearing debris from the swivel raceway¹⁵.

Those periods are not interchangeable, and none is stated as an AGV/AMR OEM requirement. Treat the structure as settled and the period as yours to calibrate: start from the shortest cadence your duty cycle can absorb, then stretch it where inspection records show nothing found.

Re-greasing is the exception to calendar thinking — swivel raceways and wheel bearings should be greased by duty cycle and environment rather than by the date¹⁸. If your fleet runs washdown, grit or long loaded shifts, that is the argument for a shorter period, not a longer one.

Swivel and bearing symptoms that take a caster off the fleet

These are the observable symptoms the sources tie to swivel and bearing degradation. Use them to decide whether a caster leaves the fleet for service, and note that no source gives an adjust-versus-replace threshold — the decision rests on what you find.

• ✓Swivel no longer turns freely — suspect raceway contamination, corrosion or dried lubrication
• ✓Rough feel when rolling — check for debris around the axle, worn tread, poor lubrication, corrosion, axle damage, side loading, or a bearing that no longer fits the wheel
• ✓Audible shimmy or vibration at travel speed — a raceway play symptom on AGV duty
• ✓Visible play at the swivel head — compare against the AGV axial-play figure before accepting the caster back into service
• ✓Loose kingpin or walked fasteners — listed among the failure signals a PM inspection catches
• ✓Axle wear and tread cupping — inspect at every hands-on cycle

The play limit that separates AGV duty from manual-cart duty

Raceway play is where an AGV caster and a manual-cart caster stop being the same component. Manual casters tolerate 0.005–0.015 in of axial play; AGV casters need ≤0.002 in to hold docking repeatability⁶.

The mechanism is speed: swivel shimmy that is invisible at walking pace becomes a docking error at 3 mph⁶. For your inspection routine this changes what a pass looks like.

A caster that feels acceptable on a push cart can still be out of tolerance on a platform that must repeat its dock position. The same source reports that standard industrial casters on AGV duty run 8–14 months between replacements against 24–36 months for precision forged-raceway casters⁶.

A supplier dataset, so treat the interval as that supplier's observation rather than a fleet benchmark. When you write the PM record, capture the measured play, not just a pass or fail, so the trend is visible before the limit is crossed.

When a wheel is replaced on tread condition

Tread is the part of the inspection you can judge by eye and by ear. Flat spots form when a wheel sits parked under heavy load for long periods, most often on polyurethane and rubber compounds, and you feel them as a thumping vibration when rolling slowly⁹.

Replace wheels when flat spots, cracks or tread separation appear¹⁸. Chunked tread points at capacity or shock loading rather than normal wear, so a chunked wheel is a signal to recheck the load case before reordering the same part⁹.

Material choice sets how quickly that day arrives. In one 5" x 2" heavy-duty series, capacity per caster runs 800 lb on polyurethane, 1,000 lb on phenolic, 900 lb on steel and 400 lb on rubber⁵.

Rubber cushions rough floors and thresholds but flat-spots if parked loaded⁵; phenolic is hard and can chip on debris⁵. If your fleet parks loaded for long stretches, the parking pattern is part of the tread decision, not just the floor type.

Debris, cleaning and the seals that keep grit out

Debris ingress is the most consistently cited route to a dead swivel: grit packs into the raceway, washdown chemicals strip lubrication, and the caster stops turning⁹. Raceway seals and thread guards exist to block that path, so the cleaning routine and the seal spec belong in the same decision.

• ✓Clear debris from the swivel raceway at each hands-on cycle
• ✓Inspect for dirt or debris around axles and wheels
• ✓Keep wheels clean and free of debris, especially on rough floors
• ✓Specify raceway seals where grit, washdown or process debris reaches the caster — they keep dirt out of the swivel bearings and improve grease retention
• ✓Specify thread guards where hair, thread or shrink wrap wraps the hub and axle; they require a bearing, so plain-bore wheels cannot take them
• ✓Check seal temperature capability against the application — standard raceway seals are rated to 140°F and high-temp seals to 350°F

Condition triggers that pull a caster forward between inspections

A calendar alone misses the failures that develop between cycles, and the sources describe signals an operator can catch at the machine. A loose kingpin, axle wear, tread cupping, raceway play and fastener walk are the failure signals a PM inspection is meant to surface¹.

Rough feel traces back to debris at the axle, worn tread, poor lubrication, corrosion, axle damage, side loading, or a bearing that no longer fits the wheel⁴. Thumping at low speed is the flat-spot signature⁹.

Give operators a short reporting list built from those signals and let any one of them open a work order ahead of the scheduled inspection. Fastener torque is worth its own trigger: torque quarterly or after any collision¹⁸, because an impact can walk fasteners without producing an obvious handling fault.

The point of the trigger list is to shorten the gap between a symptom appearing and someone acting on it.

Weighing failure cost against replacement cost

The cost of a caster failure is not the cost of the caster. A heavy-duty caster runs roughly $50–300, the labour to replace it $20–60, and the floor damage from a failed caster $500–5,000⁹.

That gap is the argument for diagnosing before you replace like-for-like: reinstalling the same part on the same schedule reinstalls the same failure⁹. On AGV duty the replacement decision also carries a specification question.

Precision forged-raceway casters are reported to run 24–36 months between replacements against 8–14 months for standard industrial casters on the same AGV, at a 40–70% higher caster cost with a 6–12 month payback on labour and uptime⁶.

That is a supplier dataset, not a guarantee, but it reframes the premium as a maintenance-frequency trade rather than a unit-price one. If your records show casters coming off well ahead of the expected interval, the next purchase order is the place to change the spec.

Checking the published rating against real fleet duty

A caster's published load rating assumes lab conditions — smooth floor, about 70°F, 3 mph, no impact. And every condition that moves the application away from that reference reduces what the caster can carry³⁰.

Speed, cracked or uneven floors, continuous 24/7 duty and swivel operation under load all de-rate the figure³⁰. A rating accepted at face value therefore overstates what your platform can carry on a real route.

Before you accept a caster into the fleet, check the rating against the duty the AGV actually runs, not the duty in the catalogue footnote. Then record what the inspection found: measured play, tread condition, debris state and any trigger that opened a work order.

Those records are what let you stretch or shorten the cadence on evidence rather than opinion, and they are the file you take to the OEM when you ask for its own interval table.

Where the sources disagree on intervals and play

The cadence figures below come from different contexts and none is an AGV/AMR OEM requirement. Keep both sides visible and confirm against your own OEM before adopting either.

Before you request a quote

• Public sources do not settle this — put it to the supplier: no AGV/AMR OEM maintenance manual or per-component interval table appears in these sources — ask the platform OEM for its caster inspection schedule before fixing your cadence.
• Public sources do not settle this — put it to the supplier: no swivel resistance torque or bearing play rejection limit is published; the ≤0.002 in axial play figure is an AGV tolerance, not a stated rejection threshold — ask the OEM what measured play retires a caster.
• Request the complete specification sheet with units, test conditions and configuration assumptions.
• Public sources do not settle this — put it to the supplier: no minimum tread thickness, minimum wheel diameter, or flat-spot/chunking size limit is stated — the replacement triggers in these sources are qualitative (flat spots, cracks, separation).
• Public sources do not settle this — put it to the supplier: no source quantifies how much a given floor condition should shorten the inspection interval — set the period from your own inspection records.
• Confirm named service coverage, response expectations and warranty terms for your destination.
• Public sources do not settle this — put it to the supplier: no bearing-kit cost, tooling requirement, labour time per replacement, or fleet-size sourcing threshold is given.
• Public sources do not settle this — put it to the supplier: no AGV-duty caster load/speed rating table, IP rating requirement, or OEM-required inspection checklist or record format is established — confirm the record format with your OEM and quality system.