Gear Motor Quality Control: IQC, SPC and Traceability


Quality inspector measuring a gear motor output shaft with a digital micrometer at an incoming inspection station

Incoming inspection decides the batch before assembly does: a shaft diameter or magnet grade caught here costs a phone call, not a scrapped gearbox.

When a gear motor shipment arrives with a 0.3% failure rate instead of the agreed 0.1%, the instinct is to look at the end-of-line test data. That data will be clean — every unit passed. The problem happened upstream, two or three weeks earlier, in a magnet batch, a grease drum, or a press-fit that slowly drifted out of its window. End-of-line testing tells you what already went wrong; it cannot tell you why, and it cannot stop the next batch from repeating it.

Quality in mass production lives in three places: what you let through the door (incoming inspection), how the process behaves while nobody is watching (statistical control), and whether a failure can be traced back to its cause (batch traceability). This article walks through each gate from a buyer's perspective — the checks that matter for a gear motor specifically, the sampling math behind them, and the records you should be able to demand. Our own controls are summarized on the quality control page; what follows is the reasoning a supplier should be able to reproduce in an audit.

Why End-of-Line Testing Cannot Hold Quality In

End-of-line (EOL) testing has a legitimate job, and we described it in detail in our article on what buyers should audit in end-of-line testing: it catches units that are already defective, and it protects your production line from receiving them. Its limits are just as specific.

EOL testing is a filter, not a cause-finder. A motor that passes no-load current at the limit and fails in the field three months later tells you nothing about whether the cause was a weak magnet, a dry bearing, or a gear with a hard spot. And because it samples the output rather than the process, it cannot see drift: a process can pass 100% of units on Monday and be producing a 4% failure population by Friday, with no change to the EOL test limits at all.

There is also a cost asymmetry. A defect caught at incoming inspection costs a phone call to the gear vendor. The same defect caught at EOL costs a scrapped assembly plus the labor already invested in it. The same defect caught in your factory or in the field costs a shipment, an RMA, and a customer conversation. The earlier the gate, the cheaper the catch — which is why a serious QC system spends most of its effort before assembly even starts.

Gate 1 — IQC: Incoming Checks That Actually Matter for a Gear Motor

Incoming quality control (IQC) is where the fight is won or lost, because a gear motor concentrates several failure modes into a small bill of materials. These are the checks we consider non-negotiable, and the instruments behind them:

ComponentWhat is verifiedInstrumentTypical rejection trigger
Permanent magnetsRemanence (Br) spot check, dimensions, coating integrityGaussmeter, caliperBr below batch spec; coating chips
Gears (steel / PM)Tooth profile, runout, hardness, density (PM)Gear rolling tester, hardness testerRunout beyond window; hardness below HRC 58
Output shaftsDiameter tolerance, surface hardnessMicrometer, hardness testerDiameter beyond ±0.005 mm; soft shaft
BearingsRadial clearance grade, noise gradeClearance gauge, noise benchWrong clearance class; audible roughness
Copper wireEnamel class, conductor diameterMicrometer, spark testEnamel class below agreed; undersize conductor
GreaseConsistency (NLGI), dropping point, batch certificatePenetrometer, supplier CoADropping point below spec; unknown batch
Housing / flangeBore diameter, concentricity, sealing faceBore gauge, runout fixtureConcentricity beyond window; porosity
DM-28PBL2838 28 mm BLDC planetary gear motor, front view with output shaft side

DM-28PBL2838: 28 mm BLDC planetary gear motor, 6-24 V, 3-1,500 rpm, up to 15 kgf·cm, used across compact automation builds.

Gear and magnet checks deserve their reputation as the two that cause field failures. A gear with correct outer dimensions can still have a tooth-profile error that produces whine at a specific meshing frequency, which is why gear inspection runs on a rolling tester against a master gear rather than on calipers — the same principle behind our gear inspecting equipment. A magnet that measures fine on the production line but was sintered outside its temperature window can lose a few percent of remanence permanently, which shows up months later as motors that run warm and slightly slow.

The Sampling Math: How Many Pieces to Inspect

Checking every incoming piece is neither possible nor useful; the value is in a defensible sampling plan. The standard reference is ISO 2859-1, and the arithmetic is worth walking through once, because it is the language your supplier's IQC bench should speak.

Take a delivery of 1,500 gear sets. Under general inspection level II, a lot of 1,201–3,200 pieces maps to sample size code letter K, which means a sample of 125 pieces. With an acceptance quality limit (AQL) of 0.65 for major defects, the acceptance number is Ac 2 / Re 3: two or fewer major defects in the sample accepts the lot, three rejects it. For minor defects at AQL 1.5, the same sample accepts up to 5 and rejects at 6. Truly critical characteristics — the ones that would make a motor unsafe rather than just noisy — are not sampled at all; they run at 100% or on a tightened plan agreed in the quality agreement.

Two buyer-side habits make this table real instead of decorative. First, ask which lot size and code letter the supplier used for your shipment — if the answer is "we checked 5 pieces," the plan is theater. Second, ask what happens on rejection: a switch to tightened inspection on the next lots, per the switching rules, is the standard response, and a supplier without a switching rule has an IQC system that never learns.

Gate 2 — SPC on the Six Characteristics That Predict Failure

Between incoming material and end-of-line testing sits the longest stretch of the process, and the only practical way to watch it continuously is statistical process control (SPC). The idea is simple: for the characteristics that predict field failure, measure a small sample at a fixed interval, plot the behavior of the process rather than the unit, and act on drift before it produces rejects.

For a gear motor, six characteristics carry most of the predictive weight:

CTQ characteristicInstrumentTypical windowSubgroup planCpk target
No-load currentMotor performance testere.g. ≤0.40 A at rated voltage5 pcs / 2 h≥1.33
No-load speedSame bench, tachometere.g. ±3% of nominal5 pcs / 2 h≥1.33
Gearbox backlashBacklash fixturee.g. 0–1.5° at output5 pcs / shift≥1.33
Output shaft runoutDial indicator, V-blocke.g. ≤0.05 mm5 pcs / shift≥1.33
Insulation resistanceInsulation testere.g. ≥100 MΩ at 500 VDC100% (safety)100% pass
Noise at fixed distanceAcoustic boothe.g. ≤45 dB(A)5 pcs / 2 h≥1.33

The bench behind the first two rows is the same equipment shown on our motor performance test page; run at subgroup 5 every two hours, it produces the X̄–R charts that make the process visible.

A Worked Example: Why Cpk 1.00 Is Not Good Enough

Capability index answers a different question than pass/fail. Cpk asks: given how this process behaves on a normal day, how close is it to producing rejects? The formula is the distance from the process mean to the nearest spec limit, divided by three standard deviations:

Cpk = (USL − mean) / (3σ)

Consider a 24 V planetary gear motor line with a no-load current limit of 0.40 A. Suppose the process runs at a mean of 0.31 A with a standard deviation of 0.03 A. Every unit still passes — but:

Cpk = (0.40 − 0.31) / (3 × 0.03) = 0.09 / 0.09 = 1.00

A Cpk of 1.00 puts roughly 1,350 parts per million outside the limit. That is one motor in 740, which on a 20,000-piece program is 27 units shipped over the limit — and because the process is centered only 3σ from the wall, a small shift (a colder morning, a new grease drum, a different operator's press technique) pushes a visible fraction of the day's production over.

To hold that limit properly, the process needs Cpk ≥ 1.33, which for the same mean means σ ≤ 0.0225 A — about 25% less variation. In practice that reduction comes from specific, findable causes: tighter spread in magnet remanence between batches, consistent brush seating and run-in, controlled bearing preload, and gear-mesh preload set by fixture rather than by feel. This is the conversation worth having with a supplier — not "is your no-load current passing" but "what is your Cpk on it, and what did you change the last time it dropped."

Backlash tells the same story from the mechanical side. A window of 0–1.5° with a process mean of 0.85° and σ of 0.18° gives Cpu = (1.5 − 0.85)/(3 × 0.18) = 1.20 — technically capable, practically marginal, and audible in the field as unit-to-unit variation in whine. The usual fixes live in the gearbox assembly step: carrier plate press-fit windows, spacer selection discipline, and grease quantity. When backlash variation drops, noise consistency follows, and so does the lifetime cycle test result, since loose meshes wear faster.

DM-28PBL2838 BLDC planetary gear motor side view showing gearbox and rear wiring exit

Gearbox backlash and shaft runout are two of the six characteristics tracked on a control chart rather than inspected at the end of the line.

Gate 3 — Batch Traceability and the 24-Hour Drill

The first two gates reduce the chance of a problem; traceability determines what a problem costs after it happens. If a field failure can be traced to a magnet batch, the corrective action is a contained lot of 800 magnets. If it cannot be traced, the corrective action is 20,000 finished motors.

A workable traceability system for small gear motors is simpler than most buyers assume:

  1. A lot code on the housing or label — typically production date plus line identifier, applied at final assembly.
  2. Four upstream mappings behind that code: the magnet batch, the gear lot, the winding lot, and the grease batch used on that production date. These live in the MES or batch records, not on the motor.
  3. Defined retention — build records kept for a stated period (three years is common in supply agreements) in a form that can actually be retrieved.

The audit move we recommend to every buyer is the retrieval drill, and it costs ten minutes: during a factory audit or a sample review, pick any serial number from a recent shipment and ask for its complete build record — incoming inspection results, process data, EOL test values, the four upstream batch numbers. A serious system answers within 24 hours with raw values, not a pass/fail summary. A supplier who needs a week, or who returns a report with no measured data on it, has just told you what a field failure investigation will look like.

The record itself should carry at least these ten fields, per unit or per lot: serial or lot ID, drawing revision, magnet batch, gear lot, winding lot, grease batch, no-load current, no-load speed, backlash, EOL pass with measured values. If a supplier's report has room for all ten but arrives with six filled in, ask about the missing four — omissions are rarely accidental.

Change Control: The Five Substitutions That Need Your Approval

Between batches, the biggest quality risk is not a bad part — it is a different part. Suppliers substitute for ordinary commercial reasons: a magnet vendor raises prices, a grease brand has a lead-time problem, a gear supplier offers a better deal. None of that is wrong; doing it silently is.

Five substitutions should always trigger written approval from the buyer before the next production run, because each one can shift motor behavior without changing any dimension on the drawing:

  1. Magnet supplier or grade — remanence and coercivity spread change current draw and torque consistency.
  2. Grease brand or type — viscosity and consistency change noise, efficiency at temperature, and long-term wear.
  3. Gear material or heat treatment — hardness and wear rate change backlash drift over life.
  4. Winding wire or enamel class — thermal class and dielectric strength change reliability at temperature.
  5. Bearing brand — clearance and noise grade change runout and acoustics.

Anything else — a packaging change, a color, an internal fixture — is a form-fit-function judgment the supplier can make and document. The mechanism is the same engineering change control described in our article on the custom DC gear motor development process: written proposal, sample on the new configuration, first-article inspection, approval, then production. Buyers who write this list into the quality agreement at the sample stage never have to argue it during a price-pressure quarter.

What to Ask For: The Evidence Package

Everything above compresses into a short list of documents a buyer should be able to request — at RFQ for a high-volume program, and with the first production shipment at the latest:

  • IQC results for the material lots in your shipment, with the sampling plan (lot size, code letter, AQL, Ac/Re) stated
  • SPC summary for the production period: Cpk on no-load current, no-load speed and backlash, with the chart for at least one characteristic
  • EOL test data with measured values, not just pass/fail
  • The traceability mapping for your shipment (four upstream batch numbers per lot)
  • A statement of which, if any, engineering changes occurred since sample approval

One caution on reading capability numbers: a Cpk study is only as honest as its sample. A meaningful study uses at least 20–25 subgroups of 5, collected across different days and shifts — roughly 100–125 consecutive production units — not 25 units pulled from one good hour. When a supplier reports Cpk 1.67, ask when the study was run and how many subgroups it covered. For the statistical background, the NIST/SEMATECH engineering statistics handbook is the standard reference, and the definitions of acceptable quality limits and process capability indices are worth a skim before the first quality negotiation.

FAQ

What Cpk should I demand from a gear motor supplier?

Cpk ≥ 1.33 on the characteristics that predict field failure — no-load current, no-load speed and backlash as a baseline — and Cpk ≥ 1.67 for anything safety-related. More important than the number itself is knowing the study behind it: how many subgroups, over how many days, and on which production line. A Cpk value without a study is a claim, not a measurement.

Is AQL 0.65 sampling strict enough for automotive or medical programs?

AQL 0.65 on major defects is a common commercial baseline, but automotive programs typically run tighter: AQL 0.25 or 0.40 on majors, tightened inspection during ramp-up, and PPAP-style documentation instead of a simple sampling record. The sampling plan is one tool; for regulated programs, process capability and full traceability carry more weight than incoming sampling ever can.

How many units are needed for a meaningful capability study?

At least 100–125 consecutive production units, gathered as 20–25 subgroups of 5, spread across different days, shifts and — ideally — material batches. A study on 25 units from a single morning will almost always look better than the process deserves, because it captures none of the day-to-day variation that produces field rejects.

Can I verify these controls without visiting the factory?

Partially. A remote audit can cover the evidence package: sample records with raw values, the traceability drill on serials you choose, and a live video walkthrough of the IQC bench and one SPC station. What a remote visit cannot confirm is whether the charts on screen are the ones the line actually uses — for volume programs, one on-site audit before the first large release remains worth the trip.

A gear motor program inherits its quality from decisions made before the first unit ships: which materials get through the door, which process characteristics are watched statistically, and whether a failure can be traced in hours instead of weeks. None of these add meaningful cost to a motor; all of them are invisible until you ask. The list in this article is the ask. If you are setting up a new program, our team can walk through the evidence package for your application — start from the quality control overview or contact us directly with your volume and application.

 

Shenzhen DongMing Motor Co., Ltd                                                                                                                    https://www.dmdcmotor.com

Latest Blog


Nov 17,2026

Gear Motor Quality Control: IQC, SPC and Traceability

How a gear motor supplier holds quality in mass production: incoming inspection sampling, Cpk on six critical characteristics, and batch traceability a buyer can audit.


Nov 13,2026

Gear Motors for Smart Home Devices: A Quiet, Compact, Low-Power Selection Guide

A quiet, compact, low-power selection guide for smart home gear motors — noise budgets, battery life math, torque margin rules and a ten-item RFQ checklist.


Nov 10,2026

Gear Motor Output Shaft and Flange: Specifying the Mechanical Interface

How to specify a gear motor output shaft, flange register and tolerances: shaft end choice, overhung load limits, sealing and a 12-line RFQ checklist.


Nov 06,2026

Gear Motor End-of-Line Testing: What Buyers Should Audit

An end-of-line test station screens every motor against a golden sample: no-load current, speed, commutation waveform, insulation resistance and a short load check. This article gives the test sequence in order, the acceptance limits that actually catch winding, gearbox and brush defects, the 100% versus AQL split, and the test-record fields a buyer should audit before approving a lot.