Seat-rail assembly checks by variant
Seat-rail assembly inspection should establish the required variant before evaluating its parts. Block the wrong variant.
Industry // Metal components
Request an assessmentMetal parts inspection: machined, stamped, cast and welded components
Camera checks for machined, stamped, cast and assembled metal parts, including error-proofing and traceability on automotive component lines.
Technical review: Frank GuoReviewed
// Short answer
Metal parts inspection with machine vision checks parts for missing holes, slots and other features, key dimensions, visible scratches and dents, missing components in an assembly, and part codes. On metal, the result depends mostly on glare, oil films and how the part is held: a fixture that presents every part the same way, and lighting chosen for the surface finish, matter as much as the camera. Internal cracks, porosity and hardness are not visible.
// Where it fits
Inspection points from machining to packing
Parts may travel on a conveyor, in a fixture or by robot; the checks are the same.
[Schematic]Process flow · not to scale
Swipe sideways to see the whole drawing →
[1] Features & dimensions: holes, slots and key sizes after machining or stamping.
[2] Surface: scratches, dents and marks on a clean surface.
[3] Components: fasteners, clips and seals present for the variant.
[4] Code: the part mark present, readable and correct.
[5] Count: the right number of parts per box or tray.
// Defect risks
What goes wrong at each step
Visible faults by step, and the application that checks for them.
| Production step | Typical visible faults | Inspection |
|---|---|---|
| Machining | Missing holes, slots or chamfers; features out of position; the wrong variant from a program mix-up | Dimensional measurement Presence & counting |
| Stamping and forming | Missing or partly punched holes, burrs, split edges, bent tabs | Dimensional measurement Surface defect detection |
| Handling and washing | Scratches, dents, handling marks, residue and stains | Surface defect detection |
| Assembly | Missing fasteners, clips, washers, O-rings or bushes; the wrong variant fitted | Presence & counting |
| Part marking | Missing, unreadable or wrong laser-marked or dot-peened codes | OCR/OCV & code inspection |
// Inspections
Checks that apply to metal components
Feature, dimension and code checks can often share one fixtured station.
Dimensional measurement
Lengths, diameters, hole positions and gaps, with the part held at a known position.
At: after machining or stamping
Presence & counting
Holes, features and fitted components present for the variant being run.
At: machining and assembly
Surface defect detection
Scratches, dents and marks, with lighting chosen for the surface finish.
At: after washing
OCR/OCV & code inspection
Part codes and text marked on the part, read and checked against the job.
At: after marking
// Application map
Metal component inspection scenarios by part
Typical parts from South African machine shops, press shops, foundries and automotive component suppliers, what a camera checks on each, and the application page with the detail. The method in each row is a starting point, confirmed on your own parts. Internal flaws, hardness and thread gauging need other methods.
| Product / line position | What is inspected | Typical fault caught | Inspection method | Related application |
|---|---|---|---|---|
| Turned and ground shaftsafter the lathe or grinder | Diameters, lengths, grooves and chamfers in silhouette | Missing groove or chamfer, wrong length, the wrong variant | Backlight with a telecentric lens, the shaft held in V-blocks | Dimensional measurement |
| Stamped bracketsat the press exit or in a nest | Holes and slots present and in position; tabs formed | Missing or partly punched hole, slug left in a hole, unformed tab | Backlight under the part in a nest | Dimensional measurement |
| Bolts, screws and nutson a sorting feeder before packing | Head shape, length, visible thread and nylon lock insert | Mixed lengths, missing thread, missing nylon ring | Backlit profile as parts pass on a rail or glass disc; top view for nuts | Dimensional measurement |
| Aluminium and iron castingsafter fettling and shot blasting | Visible surface, cored holes and fettled edges | Cold shuts, surface blowholes, blocked cored holes, fins left on. Internal porosity needs X-ray | Low-angle or dome light; a height-profile sensor for relief | Surface defect detection |
| Welded assembliesafter the welding cell | Weld beads present, in position and of the expected visible length; weld nuts fitted | Missing or misplaced weld, missing projection-weld nut, heavy spatter. Penetration is not visible | Diffuse light, or a laser profile sensor for bead shape | Presence & counting |
| Laser and dot-peen marked partsafter the marking station | Data Matrix code readable and matching the job | Unreadable, missing or duplicated serial number | Low-angle or dome light with a reader set up for DPM | DPM code reading |
| Ball and roller bearingsafter assembly and sealing | Balls or rollers present, seal or shield fitted, snap ring | Missing ball, missing or damaged seal | Camera facing the bearing with a ring light | Presence & counting |
| Coil springsafter coiling and end grinding | Free length, ground ends and coil profile | Wrong length, unground end, uneven pitch | Backlit silhouette | Dimensional measurement |
| Brake padsafter bonding and grinding | Friction material edges, backing plate, shim and wear-indicator clip | Chipped friction edge, missing shim or clip | Camera from above with low-angle light | Surface defect detection |
| Powder-coated, zinc-plated and galvanised partsafter the curing oven or plating line | Coating coverage on visible faces and edges | Runs, drips, bare patches, craters and scratches | Colour camera with diffuse or dome light; large panels in several views | Surface defect detection |
| Machined housings and manifoldsafter the machining centre | Holes, ports and O-ring grooves present for the variant; tapped holes seen at the entry | Missing hole or groove, untapped hole. Thread quality needs a gauge | Ring light on each face, or a robot presenting faces in turn | Presence & counting |
| Fastener and fitting kitsat kitting, before sealing | Every compartment filled with the right part | Empty compartment, wrong part in a compartment | Overhead camera with diffuse light over the open tray | Presence & counting |
// Integration
Conditions specific to metal parts
On metal, lighting and presentation usually decide the result before the software does.
Glare
Reflective surfaces
Polished metal reflects like a mirror; hot spots hide or imitate defects. Dome or low-angle lighting, sometimes with polarising filters, is chosen for the finish.
We askSurface finishes on the line
Oil films
Coolant, oil and mist
Cutting fluid and oil change how a surface reflects and can look like stains; mist and chips foul optics. Checks sit where the surface is consistent.
We askSurface condition at the check
Fixturing
Part presentation
Measurement needs the part at a known position and distance. A fixture or nest often decides whether a measurement repeats.
We askHow parts are held and loaded
Tolerance
Tolerance against optics
Resolution is shared across the field of view, so tolerance, part size and optics are weighed together; precise edges may need telecentric optics.
We askDrawing with the critical tolerances
Variants
Similar parts, different jobs
Variants can differ by one hole or chamfer. Each needs a recipe, and the station can confirm the part matches the job being run.
We askVariant list and job selection
Part marks
Direct part marks (DPM)
Laser and dot-peen codes have less contrast than print and vary with the surface; they need suitable lighting and a reader set up for DPM.
We askMarking method and surface
// Automotive suppliers
Error-proofing and traceability evidence for automotive component suppliers
Where vision fits in the checks that vehicle manufacturers expect from their suppliers.
Error-proofing (poka-yoke) at an assembly or machining cell uses a camera to confirm, before the next step, that the right variant is loaded, the right way round, with every fastener, weld nut, washer, clip or O-ring present. The result is interlocked with the cell PLC, so a wrong part stops the cell instead of travelling on. It is often the first automotive inspection a component supplier automates, because a missed feature found at the customer is far more expensive than one found at the cell.
Traceability evidence. Vehicle and equipment manufacturers audit their suppliers on whether each part can be traced to its batch, machine and test result. A DPM code reading station reads the serial marked on the part and stores it with the inspection results and image, so the evidence exists per part rather than per shift sheet.
Metal surface defect detection on machined, ground or plated faces depends mostly on lighting: low-angle light for scratches, dome light for shiny curved parts, and fixtures that present every part the same way. Casting porosity and cracks below the surface stay with X-ray or ultrasonic testing.
// Limits
What a camera cannot tell you about a metal part
We agree these boundaries before a project starts.
Needs another method
- Internal cracks and porosity. Sub-surface cracks, casting porosity and weld penetration need methods such as X-ray, ultrasonic or eddy-current testing.
- Hardness and material. Heat treatment, hardness and alloy are not visible.
- Thread quality. Vision may show that a hole is threaded, not that the thread passes a gauge.
- Hidden features. Features inside bores or on faces the camera cannot see need another view or method.
- Full geometry to CMM accuracy. A line station checks selected dimensions on every part; full 3D geometry and first-article inspection stay on the CMM.
Works reliably when
- Parts are fixtured. Same position and distance for every part.
- Lighting suits the finish. Tested on each finish that runs on the line.
- Tolerance fits the optics. Checked on samples before a measurement is agreed.
// Samples & acceptance
What to send for a sample test
We test your own parts before proposing a station; quantities and shipping are agreed first. The same samples later define acceptance.
Sample preparation guideGood parts
Parts across the normal range of finish, and each variant that runs.
Defect parts
Missing features, out-of-size parts, scratches and missing components, each marked.
Drawings
Drawings with the dimensions and tolerances that decide acceptance, and how you measure them today.
Acceptance
Measurements compared with your reference method, plus false rejects and missed defects on marked samples; levels agreed per project.
// Manual inspection
Why manual inspection struggles on these lines
Conditions on metal component lines that make checking by eye unreliable over a full shift.
- Glare. On machined and polished surfaces, scratches and dents appear and disappear as the angle of view changes.
- Oil, coolant and chips. Films and residue look like stains, so an inspector has to judge each mark rather than simply see it.
- Similar variants. Parts that differ by one hole or chamfer are easy to mix on a busy cell, and hard to tell apart at a glance.
- Measurement by sampling. Gauges are used on samples, so drift between two checks is not seen until the next one.
- Shift fatigue. Repetitive handling of heavy or sharp parts over a long shift makes the rare fault easier to miss.
These are reasons to automate the checks that are defined and visible, and to keep people for judgement and for fixing causes. The guide to manual or automated camera inspection compares the two.
// Line solutions
Line solutions for metal components
Complete stations for lines in this industry, each with the HIKROBOT hardware, installation and an indicative installed budget for a typical application in rand, excluding VAT. All line solutions →
| Line solution | Faults it catches | HIKROBOT configuration | Indicative budget (excl. VAT) |
|---|---|---|---|
| Rebar counting | Short or over-counted bar, tube and pipe bundles; hand counts with no record | 20 MP camera + VisionMaster deep-learning count, operator screen | R 430 000 – R 1 190 000 |
| Assembly verification | Missing screws and clips, O-rings, wrong variant, connector not latched | SC vision sensor, or smart cameras + ID5000 with PLC interlock | R 100 000 – R 960 000 |
| Tyre barcode reading | Tyres with no read, unreadable labels, wrong code for the lane | Two to six HIKROBOT ID5000 readers in a frame, MES check, divert | R 260 000 – R 1 390 000 |
| Fastener inspection | Wrong length or diameter, thread missing, mixed parts, profile burrs | Telecentric camera + backlight, VisionMaster, air-jet reject | R 270 000 – R 930 000 |
| Glue bead inspection | Gaps, thin or wide sections and off-path beads; low bead height in 3D | 2D camera + ring light, or DP 3D profiler on an axis or robot | R 230 000 – R 1 590 000 |
| Anode plate inspection | Copper anode flatness, lugs, fins, blisters; carbon anode cracks, chips, stub holes | DP-series 3D profiler + 20 MP camera with deep learning, cooled | R 1 190 000 – R 2 760 000 |
Basis of the budget column: an indicative installed budget for a typical application, in rand excluding VAT, as at 2 October 2026, the price date used on the line-solution pages. Projects involve custom requirements, so the firm price follows the sample test. Each row links to the line-solution page that shows the cost breakdown.
Separate visible workmanship from dimensional conformity
Sassda describes the Southern African stainless-steel value chain, including local manufacturing and fabrication. These processes combine surface appearance, assembly and measurement questions; they should not share one undifferentiated pass signal. (Southern Africa Stainless Steel Development Association) A machined surface may carry acceptable tooling marks while a visually clean component can still be outside its drawing tolerance or have an internal material fault.
| Operation | Vision task and datum | Failure consequence |
|---|---|---|
| After machining and cleaning | Visible burr, missing hole and profile from a repeatable datum | Assembly interference or additional sorting and rework |
| After tapping or fastening | Visible hole entrance or fastener presence | A missing fastener can escape; internal thread quality still needs its own method |
| After coating or finishing | Appearance with defined lighting and surface class | Cosmetic rejects or customer sorting; adhesion remains a separate test |
| Before packing | Variant, orientation, quantity and marked identity | Mixed parts or traceability break at the customer |
For a turning or milling cell, state the condition of the part at inspection. Coolant film changes edge contrast, and loose chips can look like burrs. A cleaning step is part of the measurement method if calibration and acceptance were performed on clean parts. Test permitted residual contamination explicitly rather than assuming every production part reaches the camera in the same condition.
A hypothetical decision sequence is to locate the datum, confirm the correct variant, measure the exposed profile and check required components. Failed location produces an incomplete inspection; the station must not substitute a dimension measured from the fixture. Save the feature overlay and recipe identity so a disputed dimension can be reconstructed.
NIST’s gauge-study framework separates variability from bias. Compare the same physical features with an appropriate reference method and include reloading in the study. (NIST e-Handbook, 2.4 Gauge R&R studies) The South African Measurement Units and Measurement Standards Act supplies the national measurement context, while JCGM 100 explains uncertainty evaluation; neither supplies the customer’s drawing tolerance. (South Africa: Measurement Units and Measurement Standards Act 18 of 2006) (JCGM 100:2008, Guide to the expression of uncertainty in measurement) If the optical result is too close to a tolerance boundary for a defensible decision, route the part for the agreed reference measurement instead of silently widening the tolerance. Account for that review burden in the quotation and acceptance plan.
// FAQ
Metal component inspection questions
Can machine vision detect scratches on metal parts?
Often, when lighting makes the scratch stand out from the normal machining pattern; low-angle and dome lighting are the usual starting points. Fine scratches on oily or textured surfaces are harder, so we test your parts and agree which marks count.
How accurate is camera measurement of metal parts?
There is no single figure. It depends on field of view, resolution, lens, lighting, edge quality and presentation. We compare your drawing tolerance with what the optics resolve, and check repeatability on your parts, before agreeing a measurement.
How are missing components detected in an assembly?
The station checks each position where a fastener, clip or O-ring should be, against the recipe for that variant. Hidden components need another view.
Can the station read laser-marked or dot-peened codes?
Often, with lighting and a reader set up for direct part marks. Reading confirms the code is present and decodes; it is not the same as grading mark quality to a standard. Marks on curved, rough or oily surfaces are tested on samples first.
Can machine vision sort mixed bolts, screws and nuts?
Often, when the difference is visible: length, head shape, thread present or a missing nylon insert. Parts are usually fed one at a time past the camera on a rail or glass disc and blown off when rejected. Thread gauging, hardness and plating thickness need other checks.
Can a camera check welds on fabricated parts?
It can check what is visible: that each weld is present, in the right place and of the expected visible length, and that weld nuts are fitted. A laser profile sensor adds bead shape. Penetration, internal porosity and cracks below the surface need X-ray, ultrasonic or other test methods.
Sources
- Southern Africa Stainless Steel Development Association (local stainless steel value chain; read 2026-10-27)
- NIST e-Handbook, 2.4 Gauge R&R studies (measurement study design; read 2026-10-27)
- South Africa: Measurement Units and Measurement Standards Act 18 of 2006 (national measurement framework, not an inspection tolerance; read 2026-10-27)
- JCGM 100:2008, Guide to the expression of uncertainty in measurement (sections 4–6: uncertainty evaluation and combination; read 2026-10-27)
// Production checks
Machining cells: choose a station by the operation
Machined housings, flanges and precision components need a more specific route than a generic metal defect list. Use the machining guide to map drawing characteristics, missed operations, burrs and variant identity to a feasible inspection point.
| Decision or event | Check or action | Acceptance boundary |
|---|---|---|
| Dimensions | Fixtured visible characteristic measurement | Keep hidden or full 3D geometry on its applicable reference method. |
| Missed machining | Hole, chamfer and visible feature presence | Do not infer internal thread quality. |
| Edge defects | Visible residual burrs after deburring | Oil, chips and occlusion require validation. |
| Part identity | Visible variant or DPM matched to the order | Appearance cannot establish alloy or heat treatment. |
Start with drawings, realistic sample variation and a defined material hold/reject route.
// Metal components
Request an assessment for your parts
Tell us the parts, finishes, variants and what you need to check. We will say which checks apply and what a sample test needs.