How to inspect stamped parts for repeat production

A bracket that passed first article at the start of the shift can fail at 60,000 strokes on the same press, same tool and same operator. Stamped part inspection has to be built around three variables that a machined part does not have: tool wear counted in strokes, coil-to-coil variation in thickness and hardness, and a strip that holds four or five parts in process at once.

Most press shops inspect as if they were machining. They measure the first piece, the last piece and a few in between, write the numbers on a sheet and file it. That catches a gross error such as a missing hole. It misses the two failures that actually reach the customer: a die that has worn 0.06 mm since the last regrind, and a coil running 0.08 mm under nominal at the thin end.

What must stamped part inspection control in repeat production?

Stamped part inspection is a check on the tool, not on the person. The dimensions move because the cutting edges wear, the punch shortens after each grind, the press ram tilts slightly under an off-centre load, and the incoming strip changes grade.

Stroke count since regrind, coil lot and thickness, two or three critical characteristics, and gauge or fixture identity are the four controls that cover nearly all of it:

• Stroke count since regrind. This is the real clock. A die at 80,000 strokes is a different process from the same die at 8,000.
• Coil lot and thickness. Record the lot on the first-off sheet. A 0.05 mm thickness change moves hole size, burr height and bend allowance at the same time.
• Critical characteristics. Two or three dimensions that decide whether the part assembles. Everything else is a general tolerance.
• Gauge and fixture identity. A checking fixture that has not been verified is not evidence, it is opinion.

Which characteristics belong on the inspection sheet?

The sheet should be short enough that an inspector completes it properly. Pick the features that affect assembly, function or the next operation, then add burr height, which is the earliest physical signal of wear.

Two rules keep the sheet honest: a characteristic with no tolerance basis will be argued about at goods-in, and a capability statement is not a substitute for the drawing. Where the Ganesh column above and the customer's drawing disagree, the drawing wins and we raise the question before the tool is cut.

How do you set a sampling interval in strokes?

Work backwards from the wear allowance the die has left. A die with 0.075 mm of working allowance and a history of 0.015 mm of wear per 10,000 strokes gets a check every 0.075 ÷ 0.015 × 10,000 = 5,000 strokes; how that regrind history is kept, and how the allowance is set, is the subject of the progressive die maintenance guide.

At 30 strokes per minute a 5,000-stroke interval is roughly 2 hours 45 minutes of running, so a check every two to three hours fits the interval without adding work.

The same arithmetic works on a progressive die with four stations. Wear at the piercing station usually runs faster than at the forming station, so the interval is set by the fastest-wearing feature, not by the average.

Where the interval is shorter than the time an inspector needs to reach the press, the process needs a different control: a sensor in the die, a mistake-proofed gauge on the line, or 100% inspection at the next operation. Adding inspectors to a die that wears quickly does not fix the die.

What does a first article have to prove?

A first article proves the process, not the part. One good piece can come from a tool that is about to drift out of tolerance, so the release decision needs a short capability study: 30 consecutive parts, one characteristic at a time, mean and standard deviation.

Take a mounting length of 50.0 ±0.3 mm. If the 30-piece mean is 50.05 mm and the standard deviation is 0.06 mm, the distance to the upper limit is 0.25 mm and to the lower limit 0.35 mm. Cpk is the smaller of those divided by three sigma, which is 0.25 ÷ 0.18, or 1.39.

A Cpk of 1.33 is the working floor for release. Below 1.0, the tool will produce rejects during a normal run and the answer is a tool correction, not a wider tolerance. Between 1.0 and 1.33, release with 100% inspection of the critical feature until the correction is made.

The study is only meaningful if the parts come from the production tool on the production press. Parts from a tryout press, or from a tool that has been hand-adjusted for the sample, measure the tool room's skill rather than the process.

How many parts does a shipment need?

Counting inspection is an attribute check: pass or fail, no numbers. ISO 2859-1 sets sample sizes against lot size, inspection level and acceptance quality limit. At AQL 1.0 on general inspection level II, a lot of 1,201 to 3,200 pieces needs 125 samples, accepts 3 defectives and rejects at 4. A lot of 501 to 1,200 needs 80 samples at 2 and 3.

Three cautions apply: sampling plans assume a well-mixed lot and random selection that a pallet of stamped parts rarely is, AQL describes the quality level the plan will accept most of the time rather than a guarantee about the parts you receive, and a sample of 125 cannot find a defect that occurs once in 5,000 pieces. A safety-related characteristic therefore belongs on 100% inspection or a poka-yoke, not on a sampling plan.

For the two or three critical dimensions, variables data from 5 parts per shift plotted on a chart tells you more than 125 go/no-go checks, because it shows direction before it shows failure.

What should the checking fixture hold?

A checking fixture is worth building when three or more characteristics are measured by hand and their relationship matters. It removes the reading error that causes two inspectors to disagree about the same part, and it measures hole position the way the assembly does.

Build it to the drawing's datums, not to the part. The primary datum locates on three points, the secondary on two and the tertiary on one; a fixture that clamps on a formed face will spring the part straight and pass a part that will not fit.

Pin clearance decides what the fixture can claim. A go pin at the hole's low limit needs 0.02–0.05 mm of clearance to enter without force, which means a fixture is a functional check rather than a precision measurement. If the drawing calls for positional tolerance of ±0.2 mm, that resolution is fine. If it calls for ±0.05 mm, the fixture will not resolve it and the part belongs on a CMM.

Verify the fixture on a known good part and a known bad part at build. Keep the verification record with the fixture. A fixture that has never been challenged is a fixture that may be passing everything.

What happens when the chart goes out of control?

Stop the press, quarantine the output back to the last known good check, and count the parts at risk. If the interval is 5,000 strokes and the die is single-out, that is up to 5,000 parts between the last good check and the detection. A two-out tool doubles the number.

Then separate tool wear, material, and press or setup, because they lead to different actions:

• Tool wear. Values drift in one direction, slowly. Regrind, reset the interval, restart.
• Material. A step change at a coil change, with burr and hole size moving together. Quarantine by lot, not by time.
• Press or setup. A step change with no coil change, often after a die-height adjustment or a shut-height error. Check the press before touching the tool.

Rework of a stamped part is limited. A hole can sometimes be opened or a flange adjusted, but a distorted flat face or a thinned wall cannot be brought back. Deciding this quickly matters more than the report that follows.

How do you catch the defects a gauge cannot measure?

Cracks at a draw radius, wall thinning and orange peel, slug marks and double hits, and galling do not show up on a dimensional report:

• Cracks at a draw radius. Visible under 500 lux or better at the radius, or by dye penetrant. Section the first article and look at the wall.
• Thinning and orange peel. A section at 25%, 50% and 75% of draw depth shows whether the wall is being stretched beyond 15%.
• Slug marks and double hits. A double hit leaves a step, a bright band or a distorted hole; it also damages the die, so it is a tool event as much as a part event.
• Galling and pick-up. Material welded to the die face leaves score marks that repeat at the same position on every part. It shows up first on stainless and aluminium.

Photograph the reference defect for each of these and keep the photos with the approved sample. A written description of a slug mark is useless at 06:00 on a Sunday shift; a photograph with a scale is not.

What Ganesh runs for stamped parts

Ganesh stamps on mechanical presses from 20 T to 300 T, with a 250 T punching press for blanks, louvres and perforations, and an in-house tool room that builds press tools, progressive dies and the checking fixtures used at the press. Because the fixture and the tool come from the same drawing revision, the datum scheme on the inspection sheet matches the one the tool was built to.

Custom parts run from 0.8 mm to 40 mm in mild steel, CRCA, stainless, aluminium, brass and copper, with ±0.5 mm typical tolerance and tighter features by agreement. First article inspection and PPAP documentation are available, and the plants hold ISO 9001:2015 and ISO 14001:2015. The stamping route supplies programmes for Mahindra, Mahindra Powerol, Minda and Godrej & Boyce.

The sheet metal stamping capability page lists the press range and the tool room scope, and the stamping and press tool hub collects the related design and tooling material. If the wear side of this problem is the one you are chasing, the progressive die maintenance guide covers sharpening intervals and die life in detail.

What to put on the drawing and the inspection record

• Datum scheme, with three datums and a note on which feature is primary.
• Critical characteristics marked, ideally with the control method stated.
• General tolerance class, for example ISO 2768-1 class m, so the sheet is not full of ±0.5 mm notes.
• Burr direction and maximum burr height.
• Material grade and the thickness tolerance band, not just nominal thickness.
• Coating specification and the surface that must remain unpainted for earthing or welding.
• Any characteristic that must be measured after coating rather than before.

On the record itself, capture the drawing revision, coil lot, die number, press, stroke counter reading since regrind, gauge identity and calibration date, the values, and the inspector. Six fields turn a measurement into evidence.

When this inspection plan is wrong

A plan built on dimensions does not travel well to a welded assembly, where the datum strategy has to be set after welding because distortion moves the frame. Coated parts must be measured after coating for anything that the film thickness affects, and before coating for features that the film will fill.

Very high volume changes the economics. Above roughly 500,000 pieces a year on one part, in-die sensing and 100% vision inspection usually cost less than the inspector time and the containment risk of a sampling plan, and the tool should be specified for it from the start.

We will also say where a plan cannot be met. A hole of 1 mm in 4 mm steel, a tolerance of ±0.05 mm on a formed feature, or a flatness of 0.1 mm across a 1 m panel are not ordinary press results; they need a separate operation or a different process, and it is cheaper to hear that at quotation than after the tool is built.

The stamping tolerance guide for brackets covers what a press can hold by feature. Use the RFQ checklist to make sure the revision, the annual volume and the acceptance method reach us in the first message, then send the drawing and the inspection plan for quotation.