3D Scanning Helps Improve Inspection Efficiency of Cast Aluminum Auto Parts

A die-cast or sand-cast aluminum housing has three surfaces that matter: the machined sealing face, the machined bores, and the as-cast geometry in between. A touch probe handles the first two well. The as-cast surface is where it slows down. Ribs, bosses and draft walls are freeform, and a probe path that follows them is long to program and easy to get wrong on the next revision.

Structured-light or laser scanning captures that freeform surface in one pass. A typical automotive housing scans in 3 to 10 minutes depending on size and how many setups are needed to see the back side. The scanned mesh is then aligned to the CAD model by best-fit on the datum features, not on the whole surface. Aligning on the whole surface hides parting-line shift. Aligning on datums shows it.

The output you want is not a pretty colored map. It is a dimensional report with the callouts your drawing already has: flatness on the sealing face, position on the bolt holes, wall thickness at the thin sections. If the scan software cannot produce those numbers with a stated alignment, the scan is a picture, not an inspection.

• 1
  Good fitFreeform as-cast geometry, thin walls, first-article layout, parts too large for a CMM envelope.
• 2
  Poor fitDeep bores and undercuts the scanner line-of-sight cannot reach without many setups.
• 3
  What it replacesManual height-gauge checks and go/no-go fixtures on non-critical cast features.
• 4
  What it does not replaceCMM verification of tight bores, threads and datum features used for PPAP.

Two inspectors can scan the same casting and report different flatness numbers. The difference is almost always alignment. If the software best-fits to the entire mesh, a bowed casting will be split in half and each half will look acceptable. If it aligns to the three datum planes on the drawing, the bow shows up where it belongs.

For automotive castings we align to the datum reference frame on the print, then report profile of a surface against the CAD model. Position on hole patterns is measured as a cylinder fit to the scanned hole wall, not as a single point. That distinction matters on cast holes where draft makes the top and bottom diameters different.

Scan noise is another variable. A matte, blasted aluminum surface scans cleanly. A shiny as-cast skin or a part still wet with coolant returns noisy points, and noise inflates flatness and profile results. Cleaning and, when needed, a light dusting coat cost a few minutes and save a re-scan.

The gain is not that scanning is faster than a probe on every feature. On a 200 mm housing we might spend 8 minutes scanning and 25 minutes on the CMM for the same part. The gain is that the scan runs while the CMM is doing a job it is actually good at, and that the scan catches a trend before the part reaches the CMM at all.

On a die-cast program, the first 30 shots after a die repair are the risky ones. Scanning them one by one and overlaying the meshes shows die wear as a drift in the parting line, not as a single out-of-tolerance part. That early signal is worth more than the inspection time it costs.

For in-process checks on a machining cell, a scanner on a fixture can verify the machined face and hole pattern in a couple of minutes between cycles. It will not certify a bore to ±0.005 mm. It will tell you whether the fixture moved.

• 1
  Batch first-articleScan five parts from one shot and overlay them to separate die variation from measurement noise.
• 2
  Die repair follow-upScan the first parts after repair and compare against the pre-repair baseline mesh.
• 3
  Fixture driftQuick scan of a known datum before and after a fixture change.
• 4
  Supplier comparisonSame scan recipe run on incoming parts to compare cavity-to-cavity spread.

A scan report is only useful if it changes something. On a cast-then-machined part, the common finding is stock variation on the face that gets milled. If the casting shifts 0.3 mm, the first machining setup either cuts air or cuts too deep. Scanning the incoming casting and adjusting the setup offset keeps the machined face in the same place relative to the datums.

We run this loop on aluminum castings in ADC12 and A356, plus 6061 and 7075 when the part is machined from billet instead of cast. The scan tells us where the stock is. The 5-axis setup removes it. The CMM confirms the result. Three tools, three different questions.

If you are evaluating a new casting supplier, ask for the scan alignment method and the datum scheme before you ask for the color map. A map with no stated alignment is not evidence of anything.