How to Keep Your Pocket Dimensions Stable
Most people hit this problem when they're running long batches on a machine that doesn't quite hold its thermal equilibrium, or when they're using a cutting tool that's seen more jobs than it should have. The pocket starts at 20mm on the first part, then slowly trends toward 20.08mm by part fifty, then jumps to 21mm by part one hundred and twenty. You're not losing tolerance from tool wear alone. There's something else working against you. I spent about three weeks dealing with a batch of aluminum housings where the bottom pocket was consistently 0.12mm undersize compared to the top pocket, and my first instinct was to blame the tool length offset. It wasn't. The real issue was that my coolant flow wasn't reaching the bottom of the pocket cavity properly, so the material near the spindle nose ran hotter than the material at the tool tip. The thermal expansion was uneven across the pocket depth, and the control system had no idea it was happening.What Pocket Drift Actually Is
Pocket Drift describes the gradual dimensional shift of a machined pocket feature over the course of a production run. It's not tool wear in the traditional sense. Tool wear is predictable and usually consistent. Pocket Drift is a combination of thermal effects, clamping relaxation, chip recutting, and sometimes even controller drift in older machines. The pocket moves, and it's often harder to detect until you've already made a dozen bad parts. The reason it's called Pocket Drift specifically comes from the geometry. Pockets are closed features with walls on all sides. When thermal conditions change or clamping pressure shifts, those walls move inward or outward unevenly. A slot might float a bit, but a pocket locks that movement in and magnifies it because there's nowhere for the stress to relieve itself.Identifying It Before It Ruins Your Batch
The first sign is usually the gauge check at the start of the run. You cut five parts, measure them all, and they look fine. You run twenty more, then pull another gauge check, and suddenly you're out of spec on two features. The pockets got bigger or smaller depending on the material and the direction of thermal expansion. The most reliable way to catch this early is to place an in-process probe routine between parts thirty and forty. I don't mean probing the finished pocket. I mean probing the raw stock pocket before the final pass, or probing a reference edge on the part that you know shouldn't move. When I ran that housekeeping job, I set up a probing cycle on a corner of the fixture plate itself, not on the part. That way I could see if the entire work envelope was shifting due to machine heat, separate from the pocket moving on its own. The fixture plate probe showed a 0.03mm drift over two hours, which turned out to be half the total pocket variation I was seeing. The other half came from the part itself relaxing after the semi-finishing pass removed too much material too quickly.The Workaround That Actually Works
Here's what I do now instead of chasing offsets after the fact:I split the pocket operation into three passes instead of two. The first pass removes about sixty percent of the remaining material with a conservative stepover of thirty percent. The second pass takes another thirty-five percent. The final pass is a light finishing cut at ten percent depth of cut. This keeps heat generation distributed and prevents the clamping forces from fighting against a part that's still stressed from roughing. I also run a dry pass on the first part of every new tool. No coolant. Just air. This lets the tool and the machine head reach a stable temperature before any real material removal happens. The first part from a cold tool usually measures differently from the second part because the tool itself expanded slightly during that initial pass. Skipping this step costs me about eight minutes per tool change, but it saves me from rework that takes forty-five minutes or more. For the probe routine, I set it to call every fifteen parts during a run longer than two hundred parts. The probe measures the pocket floor and one side wall, compares the values to the target, and applies a micro-offset if needed. The offset is never more than 0.02mm per call. If it's asking for more than that, the routine stops the machine and flags it for manual inspection instead of auto-compensating. Auto-compensating on large drifts hides the real problem and usually makes things worse on the next feature.
