Online Inspection of the Size and Flatness of the Aluminum Plate of 0840ac

• 1
  Key takeaways
• 2
  What the aluminum plate of 0840ac has to hold
• 3
  How online inspection actually measures the plate
• 4
  Where inline flatness readings go wrong
• 5
  Turning online readings into process control
• 6
  Symptoms, causes and what to do
• 7
  Inline vs offline inspection for thin aluminum plates

A phone mid-frame plate is thin, anodized on some faces, and drilled in several places. Typical wall thickness lands between 0.8 mm and 1.5 mm, and the outline might run 150 mm × 70 mm. At that ratio the plate behaves less like a block and more like a sheet of paper with holes in it.

Two families of dimensions matter. Size covers the outline, hole positions, hole diameters and edge steps. Flatness covers the whole free face, usually called out as a total value across the entire surface rather than a local bump. A 0.05 mm bow across 150 mm will not show in a caliper reading, but it will show when the display stack is bonded.

Material choice drives the rest. 6061-T6 and 6063 machine cleanly and hold tight form. 5052 bends instead of chipping, which helps forming but hurts flatness control. 7075 gives the best stiffness for a given thickness and the worst residual stress. Heat-treat state matters more than alloy grade in most flatness disputes.

A 5-axis cell can cut the pockets, chamfers and side holes in one setup, which removes the re-clamping error that usually causes size scatter. The plate leaves the machine flat because it was never unclamped mid-process. That is the main reason a 5-axis route beats three separate 3-axis operations on this part.

• 1
  Thin walls0.8–1.5 mm typical, so clamping pressure becomes a process variable.
• 2
  Free-face flatnessHeld as a total value, not a per-unit-area value.
• 3
  One-setup machining5-axis cuts pockets and side holes without re-clamping.

An inline station does not measure the whole part. It measures features it can see fast. Two sensor types do most of the work on aluminum: a telecentric camera for outline and hole position, and a laser profile scanner for height and tilt. Both read in a few hundred milliseconds per field of view, which is what keeps them inside the cycle time.

The camera side is a 2D measurement. It compares edge and hole centroids against a reference CAD overlay. On a clean, matte, anodized surface it repeats within a few microns. On a bright machined face it does not, because specular reflection moves the detected edge. Diffused ring lighting and a fixed part-to-lens distance fix most of that.

The laser side is a 3D measurement. A line scanner sweeps across the plate and returns a height map. Fit a least-squares plane through the points and the maximum deviation is your flatness value. Sampling density matters: 200 points across a 150 mm plate is not enough to catch a short bump. One point per 0.5 mm gives 300 points per line and a usable map.

Measurement uncertainty sets your guard band. If the scanner repeats at ±0.01 mm and the print allows 0.05 mm, you cannot use the full tolerance window for the process. Set the reject limit at the print value and the warn limit at print minus twice the repeatability. Anything inside the warn band gets a second look offline.

• 1
  Telecentric cameraBest for outline, hole position and hole diameter; needs controlled lighting.
• 2
  Laser line scannerBest for flatness and step height; needs one point per 0.5 mm or finer.
• 3
  Guard bandWarn limit = print tolerance minus twice the sensor repeatability.

The first source is not the sensor. It is the fixture. Any clamp, vacuum port or locating pin that pulls the plate down will flatten a bow and hide it. Release the clamps and the bow returns. If the inline station holds the part differently from the assembly fixture, the reading is measuring the fixture, not the part.

The second source is time. Aluminum cut from plate stock carries internal stress. After roughing, that stress redistributes and the plate moves, often for several hours. Measure a plate at minute two and again at hour six and you will get two different flatness numbers. Pick one fixed interval, write it on the inspection plan, and hold it for every part.

The third source is temperature. A 20 °C shift across a 150 mm aluminum plate moves it about 0.05 mm. That is the whole tolerance. Online stations sitting near a door, a spindle or a drying oven will read drift that has nothing to do with the part.

The fourth source is surface finish. A bead-blasted or anodized face scatters laser light and widens the returned spot. A mirror-polished face sends the beam away from the receiver. Both shift the computed height. If the finish changes between lots, re-qualify the scanning recipe before you trust the numbers.

• 1
  ClampingMeasure in the same state the part sees in assembly.
• 2
  Stress reliefFix one measurement interval after machining and keep it.
• 3
  Thermal drift20 °C over 150 mm of aluminum is about 0.05 mm of movement.
• 4
  Finish changeRe-qualify the recipe when the surface treatment changes.

A reading that only accepts or rejects is half a system. The useful part is the trend. Log every measurement with a timestamp and a machine ID. When flatness creeps from 0.02 mm to 0.04 mm over 200 parts, you want to see the slope before the first reject, not after.

Roughing and finishing should be split on thin plates. Take 0.3–0.5 mm in roughing, leave 0.15–0.2 mm for finishing, and let the plate sit between the two operations. That single pause removes more flatness problems than any change to cutting parameters. On some jobs a stress-relief pass at 150–180 °C before finishing does the same job faster.

Tool wear shows up in size before it shows in flatness. A worn Ø3 mm end mill cuts a hole 0.01–0.02 mm small and leaves a taller burr. Watching hole diameter on the camera gives you a free tool-life signal. Change the tool on the trend, not on the alarm.

Gage R&R belongs in the plan from day one. Run ten plates, three operators, three repeats. If the study shows more than 30% of the tolerance coming from the measurement system, the process looks worse than it is and operators start adjusting machines for noise. Repeat the study after any fixture or lighting change.

• 1
  Log every readingTimestamp plus machine ID turns pass or fail into a trend line.
• 2
  Split the operations0.3–0.5 mm roughing, 0.15–0.2 mm finishing, with a pause between.
• 3
  Watch hole diameterIt drifts earlier than flatness and flags tool wear cheaply.
• 4
  Run gage R&RKeep measurement variation under 30% of the tolerance band.

Flatness fails only at the assembly station. Cause is usually fixture-induced flattening during machining, released when the clamps come off. Fix by re-cutting the critical face with lighter clamping, or by adding a stress-relief step before finishing. Check the assembly fixture too; sometimes the press itself is the problem.

Hole positions drift from lot to lot but flatness holds. That points at thermal growth in the spindle or a worn tool, not at the plate. Log spindle temperature against hole position for a week. A 3–5 µm shift that tracks spindle warm-up is a warm-up problem, not a machining problem.

The scanner reads a bow that the CMM does not see. Check the support under the plate. A three-point support on a thin plate lets gravity pull the middle down. The scanner sees that sag; the CMM, sitting on a granite table with full support, does not. Match the support condition between the two systems before arguing about the part.

Readings repeat well but the part is rejected at the customer. That is usually a datum mismatch. The print calls out a datum that the inline fixture does not replicate. Rebuild the fixture so the primary datum matches the drawing, then re-run the correlation study.

• 1
  Fails only in assemblyLook at clamping, not at the cutting program.
• 2
  Size drifts, flatness holdsCheck spindle warm-up and tool wear.
• 3
  Scanner and CMM disagreeMatch the support condition under the plate.
• 4
  Repeats well, still rejectedThe datum in the fixture does not match the drawing.