A High Precision Dental Surface Form Analysis Method

• 1
  Key takeaways
• 2
  What the dental surface form analysis method actually measures
• 3
  Setting up the workpiece and coordinate frame
• 4
  Separating removal from elastic deflection
• 5
  Predicting the final form and feeding it back to the tool
• 6
  Step by step: running the analysis
• 7
  When to use which form check

A dental surface form analysis method starts from one question: after shaving, what does the tooth flank look like at the micron level? The dental surface here is the working flank of the gear tooth, the curved face that carries load. Shaving removes a thin layer from that face, roughly 10–50 μm of stock on a hardened or pre-hardened gear blank. The shape left behind sets contact pattern, noise, and life.

Most shops measure the finished flank and stop there. That tells you the result, not the cause. This method measures the flank twice, before and after the cutting pass, and compares the two surfaces point by point. The difference is the removed layer. Fit that layer against the tool path and you can predict the next part without cutting it.

The input is a dense point cloud, not a few trace lines. We typically capture 2–5 million points per tooth flank with a structured-light or laser scanner at 5–10 μm point spacing. Fewer points smooth out the very waviness you are trying to see. On a 4,000 mm gear segment the same logic holds, but the scanner has to be repositioned and stitched with a known artifact.

One boundary matters up front. This is a comparative method. It needs a reference surface and a repeatable scan setup. If the part moves between scans, or the fixture clamps the blank differently, the difference map is garbage. Lock the part once, scan in place, cut in place where possible.

• 1
  Flank, not whole bodyScan one or two teeth at high density rather than the full gear at low density.
• 2
  Two scans minimumPre-shave and post-shave, same fixture, same coordinate frame.
• 3
  Point spacing 5–10 μmCoarser spacing hides the waviness that drives noise.

Alignment decides whether the data means anything. Bore the gear on a mandrel with a fit of H6/h5, then indicate the bore to 5 μm TIR or better. Any runout you leave in the setup shows up as a false helix error in the form map. On a Ø400 mm rotary table, dial the face and the bore before the first scan.

Choose one datum: the bore axis for cylindrical gears, or the mounting face for a segment. Build the coordinate frame once and reuse it for every scan. Do not re-fit the frame to each scan separately. Re-fitting each scan hides real movement and creates a clean-looking but wrong difference map.

Temperature is the quiet error. A 100 mm steel part grows about 1.2 μm per 1 °C. If the shop swings 4 °C between the pre-scan and the post-scan, you have added roughly 5 μm of fake form change. Let the part soak 30–60 minutes next to the machine, and log the room temperature with the scan file.

For small module gears below 1 mm, the flank is short and the scan is quick, so soak time dominates. For coarse module gears above 6 mm, the scan itself takes longer and the risk shifts to scanner drift. Re-scan a reference ball every 20 minutes and check the deviation.

• 1
  Bore fit H6/h5Indicate bore to 5 μm TIR before scanning.
• 2
  One datum per partBore axis or mounting face, fixed for all scans.
• 3
  Soak 30–60 minutesMatch part temperature to the machine before scanning.

The difference between pre- and post-shave scans is not pure removal. Under cutting force the flank springs back, so the material that leaves the part is removal minus the elastic recovery. On thin-rim or thin-web gears this term is large. A rim under 3 mm thick can deflect 5–15 μm during a 200 N cut, and most of it comes back after the pass.

Estimate deflection with a simple beam or FE model of the rim, or measure it directly with a strain gauge on the web during a trial cut. Then subtract it from the difference map. If you skip this step, you will blame the tool for form error that the fixture caused.

Tool wear is the second term. As the shaving cutter wears, the removed layer shifts. Track the difference map across a batch of 20–50 parts. A steady drift means wear. A sudden step means a setup change, a chipped tooth, or a workpiece that moved. Group the maps by cutter index so the pattern is visible.

Installation error of the cutter also enters here. A cutter mounted with 10 μm of runout cuts unevenly around the gear and leaves a once-per-revolution form signature. That signature is a diagnostic, not noise. Read its amplitude and phase to trace it back to the arbor or the mounting face.

• 1
  Deflection termThin rims deflect 5–15 μm under a 200 N cut.
• 2
  Wear trackingCompare maps across 20–50 parts to separate wear from setup shifts.
• 3
  Runout signatureA once-per-revolution error points to cutter mounting.

Once removal and deflection are separated, you can project the final flank. Take the post-shave surface and add the expected removal for the next pass, using the same cutter path. That gives a predicted form before the part is cut. Compare the prediction with the tolerance band on the drawing.

Feed the result back into tool geometry. If the predicted flank is convex at the tip and concave at the root, adjust the cutter profile or the infeed angle rather than the machine. This is the practical payoff of the analysis: fewer trial cuts and a shorter setup loop. We use the predicted form to set the first-article plan, then confirm on the CMM.

For finishing tools and chamfering tools the same map tells you where the tool will rub. A predicted contact band that runs to the tooth edge means the chamfer tool will cut into the flank. Move the tool path or change the profile before you cut metal.

Keep the loop short. Predict, cut one part, measure, compare prediction against reality. If the error is under 5 μm, release the batch. If it is larger, find which term moved: deflection, wear, or setup. Do not adjust two terms at once.

• 1
  Predict from the mapProject the next pass from the post-shave surface plus expected removal.
• 2
  Adjust tool, not machineUse the predicted form to set cutter profile and infeed angle.
• 3
  One change at a timeIf error exceeds 5 μm, fix deflection, wear, or setup separately.