The Line That Welded Here and Not There

An automated hinge line left our floor after two days of running without a fault. Six weeks later the customer sent a photograph of a part with the two welds offset by about a millimetre, and the message that went with it was the one nobody wants to read: it works at your factory.

This is what happened, what it cost, and the one rule that came out of it and is still in force.

latest company news about The Line That Welded Here and Not There 0

The automated hinge line as it left our floor: two welding heads over a walking-beam conveyor, with the control cabinet on the left. It welded correctly here for two days before it was packed.

It Welded Here

The line was built around a hinge the customer sent us in a bag: twenty pieces, clean, burr-free, from a trial run of the stamping tool. We built the fixture around those twenty pieces, ran the line on them, and welded a few hundred hinges without a reject.

That is the normal way to do it, and it is the reason the fault took six weeks to appear. The parts we were given were the best parts that tool would ever produce.

What the Customer Saw

In production the hinge came off a press that had been running for years, and the pitch between the two holes varied by more than a millimetre across a shift, depending on how warm the tool was. Our fixture located on both holes. When the pitch opened up, the part went in tight, and the second welding head fired on the edge of the tab instead of in the middle of it.

Nothing was broken. The line did exactly what it was built to do, on a part that was not the part in the bag.

latest company news about The Line That Welded Here and Not There 1

Hinges on the conveyor, held by the fixture that turned out to be the problem. The locator assumed a datum that the customer's press did not hold.

The Part Was Different, and Nobody Had Said So

The conversation afterwards was short and not entirely comfortable. The customer had sent what they had. We had not asked what the tolerance on that pitch was, or whether the sample came from the production tool, or whether the tool was new. Both sides had assumed the other knew.

The engineering fix took a week: a floating locator on one hole instead of two fixed ones, so the fixture follows the part instead of fighting it. That change is now standard on any part located by two holes that come from different operations.

There is a second lesson here that belongs to the buyer rather than to us. A sample from a trial tool is not a sample of the production part, and asking a machine builder to design against it is asking them to design against something that will not exist. If the tool is not ready, say so at the enquiry. We can still build the fixture, to the nominal dimension, and leave the adjustment in it on purpose.

What We Changed

The rule that came out of it is simple and has cost us a few orders since: for any fixture that locates on more than one feature, we ask for at least thirty parts from the production tool, taken across a shift, and we measure the variation before we design anything. If the customer cannot send them, we say in writing which dimensions we are assuming and what happens if they move.

Thirty is not a magic number. It is enough to see a tool that is drifting, and it is few enough that a customer will actually do it.

What It Cost

Two trips, a rebuilt locator, and a week of the customer's production. We absorbed the fixture work and the second trip; the customer absorbed the week. Neither of us was compensated by the other, and the relationship survived because the diagnosis was honest and arrived quickly.

The line has been running since, and it is one of the references we are happiest to be asked about, precisely because the story does not end at the shipment.

PDKJ continues to provide customized laser welding and resistance welding solutions based on different materials, workpiece structures, welding requirements, and production applications.

The Line That Welded Here And Not There