Crimping Quality Control: The Pull-Test Protocol Behind Every Cable We Ship

A bad crimp can look perfect. The conductor sits in the barrel, the insulation grip is placed correctly, the bell-mouth is the right shape, and the part passes a visual screen with no flag. Six weeks later, in the field, the joint develops a high resistance, the conductor pulls out under thermal cycling, and the customer reports an intermittent fault. The crimp die had drifted out of spec halfway through the reel, and visual inspection had no way to see it.

The pull test is the cheap, fast, destructive screen that catches what visual inspection can't:

• Crimp-height drift — a die that's wearing out compresses less, even though the bell-mouth still looks right. Pull-force drops 15–25% before any visual symptom appears.
• Wrong-AWG terminals — a terminal sized for 22AWG accepting 24AWG wire looks fine and pulls out at half the rated force.
• Strand damage — over-stripped insulation that nicks the strands. Visual passes; pull force drops because half the strands are not gripped.
• Inner-conductor recession — the strands have receded inside the insulation, so the crimp grips mostly on insulation, and pull force collapses.

"You can train a good operator to spot four of the eight crimp defect modes by eye. The pull test catches all eight, in two seconds, with a number you can log." — Pioneer Horizon test team

The protocol below is what we run on every cable-assembly programme at the Madurai facility. It's modelled on IPC/WHMA-A-620 acceptance criteria but with sampling cadence and trigger thresholds tuned from our own production data across the last three years.

The pull force a crimp must survive depends on the conductor cross-section. Below these numbers, the joint isn't trustworthy for industrial service. We hold these as minimums, not targets — targets sit 15–20% higher to leave headroom for die wear before the trigger threshold is hit.

Stranded copper conductors (typical industrial cable)

• 30 AWG — minimum 6.7N (1.5 lbf), target 8.0N.
• 28 AWG — minimum 13N (3 lbf), target 15N.
• 26 AWG — minimum 22N (5 lbf), target 26N.
• 24 AWG — minimum 36N (8 lbf), target 42N.
• 22 AWG — minimum 67N (15 lbf), target 80N.
• 20 AWG — minimum 89N (20 lbf), target 105N.
• 18 AWG — minimum 178N (40 lbf), target 210N.
• 16 AWG — minimum 222N (50 lbf), target 260N.
• 14 AWG — minimum 311N (70 lbf), target 360N.

Solid conductors and special cases

• Solid copper — derate stranded numbers by 30%. Solid conductors fatigue at the crimp edge.
• Tinned copper — same as bare copper; tinning doesn't materially change pull strength.
• High-flex stranded (e.g., DLO, locomotive cable) — derate by 15%; the higher strand count slips slightly more in the crimp.
• Silver-plated for high-temp service — same as bare copper.

How the test is run

A digital pull-tester (we use an Instron-style fixture with a 500N load cell) grips the conductor 25mm from the back of the terminal and pulls at 25mm/min until separation. The peak force, displacement at peak, and the failure mode (conductor break vs. pull-out vs. insulation slip) are all logged to the sample's serial entry in the MES.

A pass requires both (a) peak force above the minimum, AND (b) failure mode is conductor break, not pull-out. A pull-out failure below maximum is automatic line hold even if the number alone passed — the failure mode tells you the crimp didn't grip.

Pull-test is destructive. You can't run it on 100% of production. The question is how often to sample, and how to escalate when something drifts.

Per-shift sampling

1. First-off after die change or reel change — 5 units pulled, all must pass with peak force ≥ target (not just minimum). If any single sample is below target, the die or reel is re-validated before production continues.
2. Hourly thereafter — 1 unit pulled per crimp-station per hour. Logged to the operator + die serial number.
3. End of shift — 3 units pulled, peak force trend reviewed against the shift's running average. A 10% downward drift across the shift triggers a die inspection.

Per-reel sampling

Every new reel of terminals or wire gets 5 first-off pulls, AQL-1.0 sampling thereafter, and a final 3 units at reel-end. Any drift across the reel — i.e., end-of-reel numbers more than 8% below start-of-reel — flags the reel for incoming-quality follow-up.

Escalation triggers

• Any single sample below the AWG minimum: line hold, last hour's production quarantined, die inspected and re-validated.
• Three consecutive samples below target (above minimum but below the 15% headroom band): die inspected, no production hold but operator alerted.
• Any pull-out failure mode below the minimum: line hold, entire reel quarantined, supplier notified.
• A 15% drop from the rolling 24-hour average: die change-out and the removed die is sent to the metrology bench for crimp-height verification.

What this catches

Across the last 18 months on our cable-assembly line, this cadence has caught: two die-wear events (replaced before any out-of-spec product shipped), one mis-labelled reel (terminals marketed as 20AWG but sized for 22AWG), one operator error (wrong applicator on a turret crimper), and zero field-return clusters traceable to crimp quality. The previous cadence — sample once per shift — caught two die-wear events after out-of-spec product had shipped, and was the reason the current cadence exists.

The pull test is the truth, but the visual check is the early warning. An operator who knows what to look for can catch a drifting die in the hour before the next scheduled sample. The signs are subtle and consistent.

The five visual signals — in order of how often they show up

1. Bell-mouth asymmetry — the flared opening at the wire-entry end of the crimp should be symmetric around the conductor axis. Lopsided bell-mouth = the die halves aren't meeting symmetrically, usually because one half is wearing faster.
2. Insulation grip rolled or torn — the insulation grip should be smoothly indented onto the cable jacket. Tearing means the grip is over-compressing; rolling means it's under-compressing.
3. Conductor extension — strands should extend 0.5–1.5mm past the conductor crimp. Recessed strands = mis-strip; over-extended strands = strip-length drift.
4. Witness mark — most quality terminals have an inspection hole or witness mark. Strands visible through it = crimp depth correct; strands not visible = crimp too shallow.
5. Crimp-height callout dimension — measured with a micrometer at the centre of the conductor crimp, against the manufacturer's spec for that terminal. ±0.05mm tolerance on most industrial terminals.

The two signals that mean stop the line immediately

• Cracked terminal plating visible under 10× magnification on the crimp barrel — die pressure is too high and is fracturing the terminal. Pulls will pass; long-term reliability collapses because crack propagation continues in service.
• Conductor strands visibly broken at the front edge of the conductor crimp — die geometry has shifted or the wire is over-stripped. Visual pass on this sample, but the next reel may go full pull-out.

How operators get trained

Every cable-assembly operator at our facility goes through a 4-hour module with a sample board of 24 crimps — 8 good, 16 deliberately bad — and has to call out each one and explain why. They re-certify every 6 months. The pass rate on first attempt is around 70%; the goal isn't to pass first time, it's to leave the module knowing what to look for. Combined with the pull-test cadence, this is the system that's kept our cable-assembly DPMO under 200 across the last two years.

The crimp die is consumable. Every die we run has a serial number, an installation date, and a cumulative cycle counter. Above a threshold count, the die is removed from service regardless of whether it's still passing pull tests — die failure tends to be sudden, and we'd rather replace a die that has 5% life left than discover at hour 41 of a shift that it failed at hour 40.

The die-life numbers we hold to

• Standard production die (carbide-faced) — 100,000 cycles, then replaced; sent to vendor for inspection / reface.
• High-volume turret die — 250,000 cycles per applicator station, with rotational re-allocation across the four stations.
• Hand-tool die — 25,000 cycles or 12 months, whichever first.
• Engineering / low-volume die — 50,000 cycles, with mandatory crimp-height verification every 5,000.

What gets recorded

Every die in the facility has a QR-tagged maintenance record:

• Manufacturer, model, terminal series compatibility.
• Install date, install cycle counter reading.
• Every pull-test sample taken with this die, force value, pass/fail.
• Every visual flag, every line hold, every operator who used it.
• Retirement date and disposition (reface vs. scrap).

The replacement-plan economics

A carbide-faced production die runs ₹18,000–₹28,000 to refresh. A line-hold event with quarantined product traceability runs ₹40,000–₹80,000 depending on what's caught in the hold. Replacing a die at 95% of its rated life costs us 5% of a die; replacing it at 110% costs us a die and a line-hold. The arithmetic is straightforward; the discipline is to honour the schedule even when production is busy.

"The day the line is most busy is the day you most need to not skip a die change. Skipping it because the schedule is tight is how you get a Friday-evening line hold and a Monday-morning containment meeting." — Pioneer Horizon NPI lead

If you build cable assemblies in-house and want to compare your pull-test protocol against ours, share your acceptance criteria and we'll redline against the IPC/WHMA-A-620 baseline plus our internal cadence. The redline takes about an hour and tends to reveal one or two gaps that matter.