Inspection & Metrology

Application

What is inspection & metrology?

Inspection and metrology cover all methods that ensure the conformity of a component with its specifications — for tolerances, surfaces, completeness or function. The trend for years has been away from final inspection in the QC lab towards in-line inspection directly in the production cell: defects are detected before follow-on costs occur.

Automating this inspection creates three values: complete 100 % inspection instead of sampling, digital traceability of all results, and reduced workload on the already scarce quality inspectors.

Inspection types at a glance

• Visual inspection — surface defects, scratches, dents, discolouration, completeness of subassemblies
• Dimensional measurement — sizes, form tolerances, GD&T-compliant inspection (measurement of edges, holes, angles)
• Presence/position inspection — are all components present? In the correct position? In the correct orientation?
• Functional testing — electrical continuity, leak-tightness, pressure or flow measurement, mechanical mobility
• Material inspection — wall thickness (ultrasound), material types (hyperspectral), heat distribution (infrared), internal defects
• Seam inspection — bond and weld seams for completeness, height, width, porosity

Technologies and sensors

• 2D camera — standard visual inspection, classical image processing. Robust, cost-effective, high cycle rates (<100 ms/inspection).
• 3D vision — structured light, stereo cameras or time-of-flight. For form inspection, seam measurement, bin-picking preparation.
• Line-scan camera — continuous webs (sheet metal coils, textiles, paper) with very high resolution.
• AI-based defect detection — deep-learning models for hard-to-describe defects (scratches, irregularities). Pre-trained or customised with customer data sets.
• Laser triangulation — high-precision profile measurement for edges, heights, flatness.
• Tactile metrology — touch probes for high-accuracy contact measurements (a complement to optical inspection where it hits its limits).
• Thermal camera — thermography for functional and power-loss inspection, in particular in electronics.
• Ultrasound — wall-thickness measurement, defect detection inside the material (automotive, aerospace).

Static or robot-guided — two principles

The inspection station can be built in two fundamentally different configurations:

The NEXOS configuration

• NEXOS.Cube as the inspection cell — light shielding, defined illumination, protection from ambient interference. For cleanroom applications in the stainless-steel variant.
• Robot — cobot or industrial robot depending on the variant (static or robot-guided).
• Vision system — 2D, 3D or AI camera (Cognex, Keyence, Basler, Sick, iFactory and others, market-open). NEXOS integrates the chosen system into the cell control.
• Defined illumination — LED ring light, dark field, back light or structured light depending on the inspection task. Choice of illumination is more critical for success than camera resolution.
• Workpiece fixture — vacuum table, fixture or infeed conveyor depending on the application.
• Reject chute for NOK parts — automatic pusher, flap or robot gripper.
• Optional NEXOS.LinDrive — as a linear motion axis for long parts (automotive, metal construction).
• Integration with NEXOS.DSP — for inspection data archiving, trend analysis, SPC (Statistical Process Control) and QMS integration.

Sequence of a typical cell

1. Position workpiece — via conveyor or robot handling
2. Activate illumination per program (different lighting situations per inspection step)
3. Camera capture — synchronised with illumination and workpiece position
4. Image processing — feature extraction, measurement, classification (rule-based or AI)
5. Comparison with specification, tolerances, learned defect patterns
6. Decision OK / NOK / rework — including decision rationale
7. Routing — OK parts to the good line, NOK parts to reject with defect code
8. Data hand-off to MES / QMS with time stamp, batch ID, image archive of NOK parts

Why it pays off

Market data (2026):

• Market size: Automated Inspection & QC worldwide about 14.8 bn USD (2028 forecast), CAGR around 11.7 %.
• Investment range: simple 2D vision cell with cobot from about 50,000 €, complex 3D inspection station with AI vision and reject chute 100,000-250,000 €.
• Payback: typically 12-24 months, in particular through reduced complaints and eliminated rework.
• Process migration: 80-90 % of dimensional inspections can be moved from the CMM room to the in-line cell — backlogs in the metrology lab disappear.
• Side effects: digital traceability, early warning for process drift, automatic SPC evaluation. For automotive, medical, aerospace often the basic prerequisite for supplier qualification.

Scalability

• Stage 1 — single visual inspection cell with a 2D camera for one specific inspection step.
• Stage 2 — multi-camera inspection of several features on the same part in one cell.
• Stage 3 — robot-guided inspection of complex-shaped parts with many inspection points.
• Stage 4 — AI defect detection for non-rule-based features (surface scratches, colour irregularities, contamination).
• Stage 5 — closed-loop control — inspection data flow back to upstream processes (CNC, welding, injection-moulding cell), which adjust themselves automatically. End of pure defect detection, start of active process correction.

Expansion options

• Multi-sensor fusion — combination of 2D + 3D + thermography in one cell, deciding jointly.
• Cloud QMS integration — via NEXOS.DSP to SAP QM, Siemens Opcenter, IBM Maximo or industry-specific systems.
• Digital twin via NEXOS.XR — virtual programming of new parts, visibility check of camera perspectives, training of AI models with synthetic images.
• Robot-guided CMM-replacement measurement — combination of cobot + laser triangulation for 80-90 % of CMM tasks within cycle time.
• Alerting — automatic information to the operator via app, SMS or email in case of process drift.

Key sizing parameters

• Inspection features — what exactly is to be inspected? Dimension, surface, function, presence?
• Tolerances — what resolution is required? Determines camera and optics.
• Workpiece size and material — affects illumination (reflections, contrast) and cell size.
• Cycle rate — inspections/hour, available cycle time per part.
• Defect signature — are defects rule-based describable or does it need AI with training data?
• Data requirements — batch report? Image archive? Trend analysis? Integration with existing QMS?
• Reject strategy — reject chute, rework diverter, alarm to the operator?

Common pitfalls

• Illumination is underestimated — more camera resolution does not replace bad illumination. Investing in illumination is often the ROI lever.
• Calibration and maintenance — vision systems need regular calibration. The maintenance concept must be considered from the start.
• False positives / false negatives — AI systems without clean training data have a high error rate. Data collection and curation are project success factors.
• Change management — quality inspectors are not replaced but freed up for higher-value tasks (defect analysis, process optimisation). Communicate early.