Troubleshooting Processes Experiencing Drift or Shift

Source: Getty Images

Reader’s Question: Our shop has been having issues chasing tool offsets. We hover the nominal and can keep most parts in spec but we chase it constantly throughout the day and week. Can you give us some guidance how to nail this down for good?

Most troubleshooting fails for a simple reason. Rather than conducting true root cause analysis, people start assigning causes before they properly classify behavior. They see a bad number, wave it off as tool wear or temperature and reach for the offset table, rather than taking a moment to reflect if that bad feature is a part of bigger trend that was inevitably going to happen or was a sudden event that is about to come and go or arrive and stay.

Featured Content

Turn Your Operational Gaps into Efficiency Gains READ MORE

Apply AI and GPU Simulation for More Efficient CAM Programming READ MORE

10 Minutes to Influence U.S. Manufacturing Policy READ MORE

To put this practically, you need to identify if the process is experiencing drift or if it’s a sudden shift. Drift usually comes from circumstances that change gradually. Shift usually comes from circumstances that change suddenly. For example, if a bore size moves slowly from in tolerance to out of tolerance across 50 parts, that’s drift. If you run 49 good parts ±0.0005" and the 50^th is suddenly out by +0.002", that’s a shift. Both events are not ideal, but without knowing the trend, the offset table becomes a band aid, not a cure, and you’ll be re-learning the same lesson on the next batch.

The second step is assigning causes that match the shape. Was the shop heating up, did the tool chip, did you change operators? It tells you what kind of containment action will teach you something. With drift, you control trends. With shift, you remove single-point events by standardizing inputs, especially human inputs.

Drift

Drift is a slow movement of a size or position over time. It’s the kind of change you can usually relate back to work offset, a tool offset or a tool condition. Drift has dependencies. Thermal is a common one. Tool wear is another. Coolant condition, machine heat soak and human patterns can all create drift. The key is that drift has a shape. If you capture it in order, you can often predict it.

Drift problems get solved by isolating dependencies and capping the slope. If you suspect temperature, measure it and correlate it. If you suspect tool wear, track parts-per-edge and correlate it. If you suspect coolant, stop guessing and check concentration and temperature. Then contain one variable at a time. Cap tool life at a conservative value and see if the drift shrinks. Control the HVAC as much as you can and see if the trend changes. If it doesn’t, you just learned one more thing that doesn’t matter for this process.

Shift

Shift is different. Shift is a sudden movement. The process is running good, then one part is suddenly bad with no obvious lead up. Sometimes it happens once and everything returns to normal, like a bird’s nest that clears or built-up edge that finally lets go. Sometimes it happens and stays put because something seated differently or something slipped like a loose bolt on a fixture. Shift can also sit on top of drift, breaking the trend or magnifying it.

Shift problems need a different mindset. A shift is not a trend to be modeled. It’s an event to be trapped. When a process runs good and then one part is suddenly off, look for a discontinuity. A tool chipped. A chip got under a locating surface. A clamp seated differently. An operator changed the way they torque the fixture. The common thread is not gradual change. It’s a moment where reality jumped.

Mitigation

I was in a shop in Missouri in the middle of summer. At the start of the day the shop temperature was about 75°F. By the middle of the day it had climbed to 95°F. The part had a 4" bore with a tolerance of plus or minus a thousandth. The bore was drifting as the day heated up. Everyone knew why. Nobody needed a lecture about thermal growth. What they needed was a decision they could live with: at what temperature do we set the boring bar so we can cut acceptable parts morning, noon and night?

That is the difference between knowing a cause and running a process. “It’s thermal: is not a plan; it’s a category. The plan is a study that tells you where to center the process so normal environmental change doesn’t push you out of tolerance. A lot of shops chase perfection at one moment of the day. They want the bore dead on at 8:00 a.m., and then they spend the afternoon correcting the process back toward that morning snapshot.

In Missouri, the path was straightforward. Record the sequence: part number, time, measured size, temperature, tool state. When you lay it out that way, drift shows up as a line. Once you can see the line, you can make choices that aren’t guesses. One of those choices is that you might not aim for nominal at any one point. If your tolerance band is plus or minus a thousandth, the smartest target may be a position in the band that stays safe across the day. That can offend our sense of perfection, but it’s how you keep parts in tolerance from morning to evening.

There’s one more trap worth calling out. If anybody can tune offsets whenever they feel like it, you can keep a job alive, but you can’t learn from it. You're erasing evidence while you diagnose the problem. Put boundaries on the knobs, record changes and make adjustments traceable. Otherwise, you’ll fix symptoms and keep the root cause.

Back in Missouri, they didn’t fix heat; they learned how heat moved their bore, chose a setpoint strategy that stayed inside tolerance across the day and stopped pretending that one perfect morning setup was going to survive a summer afternoon. That’s what grown-up troubleshooting looks like. Start with the shape, then match the response. Drift or shift. Once you get that right, you stop chasing numbers and start building processes that behave, even when the day doesn’t.

Do you have a machining question? Ask the expert. John Miller leans on more than a decade of industry experience to answer machining questions from MMS readers. Submit your question online at mmsonline.com/MillersEdge.