Control Where the Defect Is Made

In-process control exists to catch a deviation while the affected material is still identifiable and small. That single sentence determines where controls belong: at the steps where a defect is created and where the cost of continuing to process bad material is high. A control placed only at final inspection is not in-process control; it is sorting with extra steps.

Zipper production has a short list of steps that meet both conditions. Each of them deserves a written control, and most of them deserve more than one characteristic.

The Critical Points

A practical set of control points for a zipper plant, with the characteristic that matters most at each:

  1. Tape formation and dyeing: width, weave density, shade against the approved standard, and lot identification. Shade errors found later cannot be corrected.
  2. Molding: tooth pitch, tooth profile, body dimensions and colour. Mold wear drifts slowly, so trend data matters more than pass or fail.
  3. Stamping: thickness, burr height, width and logo position. Tool wear produces a gradual burr increase that reaches the limit before it is noticed.
  4. Plating: coating thickness, adhesion, colour and salt spray performance. Bath chemistry drifts, so the control is partly chemical and partly visual.
  5. Assembly: slider function, puller attachment strength, open and close force, top and bottom stop security.
  6. Cutting and end treatment: length within tolerance, squareness of cut, correct end stop.
  7. Packing: count per bag, label data, lot code and carton marking.

Not every characteristic needs the same attention. Separate them into critical, major and minor, and set the control method accordingly. A critical characteristic justifies a check at every shift and a recorded value; a minor one may need only a periodic visual check.

What a Control Plan Must State

A control plan that names a characteristic without the rest of the information is a wish. The minimum content for each line is:

  • The characteristic and its specification, with units and tolerance.
  • The measurement method and the specific gauge or fixture.
  • The sample size and the frequency, expressed in time or in pieces.
  • The person responsible, by role rather than by name.
  • The recording location, so the data can be retrieved later.
  • The reaction plan, stating exactly what happens when the result is out of specification.

The reaction plan is the part most often missing and the part that decides whether the control works. "Notify supervisor" is not a reaction plan. A usable one says: stop the machine, quarantine output back to the last good check, mark the boundary, notify the setter and the quality engineer, and record the quantity affected.

Setting the Frequency

Frequency is a compromise between detection lag and inspection load. Two approaches are defensible. The first is statistical: check often enough that the number of units produced between checks is small relative to the quantity you are willing to scrap if a shift occurs. If a process produces six hundred units per hour and the acceptable exposure is two hundred units, the check interval cannot exceed twenty minutes.

The second is economic: increase frequency while the expected cost of the defects produced between checks exceeds the cost of the check itself. In symbols, check interval in units should be less than the check cost divided by the product of defect rate and cost per defect. Both approaches point the same way: fast, high-volume steps need frequent checks, and slow steps need fewer.

The frequency should also change with evidence. After a drift or a failure, run tightened checks for a defined number of production hours before returning to normal. Write that rule down, because otherwise the tightened period ends when the pressure of the schedule arrives.

First Piece, Patrol, Last Piece

Three checkpoints frame each production run and cost very little:

  • First piece approval after a setup, a mold change, a material lot change or a restart after a stoppage. Nothing should run until the first piece is measured and released.
  • Patrol checks at the set interval, with results recorded at the machine, not reconstructed at the end of the shift.
  • Last piece check at the end of the run, which confirms that the process was still in specification when it stopped.

First piece approval is the highest-leverage of the three, because setup error is a common cause of long runs of identical defects. It is also the easiest to skip when the line is behind schedule, which is why it needs a signature rather than a habit.

Capability, Not Just Conformance

A process can pass every check and still be incapable, because conformance looks only at whether units fall inside the limits. Capability indices compare the spread of the process with the tolerance. The capability index Cp is the tolerance width divided by six standard deviations; Cpk is the distance from the process mean to the nearer specification limit divided by three standard deviations.

Cpk is the number to act on. A Cpk near one means the process will produce defects whenever it drifts slightly. Values between one and one and a third are marginal for a mature process, and a target above one and a third is a common and reasonable expectation for a characteristic that matters. A low Cpk cannot be fixed by inspecting more; it requires a process change.

Practice and the Reason for It

YCC records molding and stamping dimensions as variables rather than pass or fail, so that tool wear appears as a trend before it appears as a defect. The data is reviewed at the start of each shift, and a tool change is triggered by the trend, not by a rejected carton. The saving is not in the inspection; it is in the scrap that never happens.

What Makes In-Process Control Fail

Controls fail in predictable ways: the recording form is filled in at the end of the shift from memory; the gauge is out of calibration and nobody noticed; the reaction plan exists but requires a decision that the operator is not authorised to make; the check is performed on the easiest unit rather than a random one. Each of these is a design problem, not a discipline problem. Fixing the form, the gauge schedule and the authority level is more effective than reminding people to be careful.