Blind Riveting Automation: The Witness

Why robotic riveting systems “fail” — and what they’re actually telling you

When I walk up to a robotic cell that isn’t running, the language is usually pretty consistent.

I’m told, "The tool isn’t breaking the rivet correctly".

There’s frustration behind it. Pressure. The line has to run, and this system was supposed to already be the solution. By the time I get there, the robot is usually offline, and people are working around it because something has to move.

I don’t argue when I hear that, but I do understand what it means.

Most people believe the tool has control over when the rivet breaks. So when it doesn’t, they assume something is missing—some setting, some adjustment, something they haven’t been told yet.

And when they’ve already been told it’s not the robot or the tool, but they don’t know where else to look, that’s when the stress starts to build.

When I actually get to see a robot run, it doesn’t force anything. If it feels resistance when it starts to place the rivet into the substrate, it backs off and moves to the next position. That’s what it’s programmed to do.

Most of the time, though, I don’t even get to see that. By the time I arrive, the robot has already been taken offline, and I’m watching videos of what it used to do during runoff or pre-delivery instead of seeing the system in motion. The one thing that was consistently showing the problem is no longer part of the process.

Unless it’s a programming fault, the robot isn’t the problem. It isn’t doing anything different than a person holding the same tool would do. It just executes. It doesn’t adjust, it doesn’t compensate, and it doesn’t cooperate.

The robot is seeing everything.

What used to be adjusted for, pushed through, or compensated by hand now shows up the same way, every time. The robot isn’t introducing anything new. It’s witnessing what was already there.

I was in a plant once where a robotic system had been taken offline. The expectation was that the tool on the robot arm wasn’t performing correctly. There was intense scrutiny around it—not just because the cell wasn’t running, but because that system was supposed to be finished. It was supposed to be working.

Instead, it was sitting there, and nearby, four operators were doing the job by hand.

That was the first signal. Because if the robot isn’t running, but people are now doing the same process manually, you have a direct comparison. Same application. Same parts.

Same outcome. Just different hands.

When I looked at the spent mandrels, they told the story. They weren’t breaking cleanly—they were ripped. Stretched, like taffy.

I asked one question: were these the same rivets the robot was using?

Because if they were, it didn’t matter who was holding the tool. The result was going to be the same.

I was asked if a different tool would solve the problem.

I said no.

Not because I didn’t want to help, but because that wasn’t the solution.

There was pressure to fix it quickly. There were other projects tied to that system, and the expectation was that this part should already be resolved. That’s not unusual. Most people in that position are trying to keep multiple things moving at once.

But this wasn’t a tool issue.

Once that was understood, everything shifted. The conversation changed, and the focus moved to the right place. More importantly, the tension started to ease.

When people can finally see what’s happening, they don’t have to guess anymore.

Maintenance knows what to go fix. Engineering understands what needs to be addressed.

Management stops feeling like they were misled.

What felt complicated becomes clear.

Nothing about blind riveting is inherently difficult, but if you don’t know what you’re looking at, it’s easy to chase the wrong thing. It’s easy to replace tools, adjust settings, or keep asking maintenance to overcome it.

It’s even easy to blame the robot.

But the robot didn’t cause the problem. It just showed it to you, consistently, every time.

If you’re looking at automation—or trying to understand why it isn’t performing the way you expected—start with the process.

Watch it. Study it. Run it before you try to automate it.

Because once the robot is in place, it’s not going to adapt to the system. It’s going to reflect it.

It’s just the witness.

Until next time... M