Why Steel and Metal Manufacturers Need AI Visual Inspection
Steel and metal manufacturing produces some of the most challenging surfaces for quality inspection. Hot-rolled and cold-rolled steel coils, structural sections, and machined components all exhibit surface conditions that vary widely based on temperature, mill settings, and material composition. Traditional threshold-based inspection systems that work well on uniform surfaces struggle with the inherent variability of metal surfaces — generating false positives from normal surface patterns while missing subtle but critical defects embedded in that variability.
AI visual inspection handles this challenge by learning the difference between normal surface variation and actual defects from production examples. The model learns what acceptable hot-rolled surface texture looks like — mill scale patterns, minor oxidation, acceptable roll marks — and discriminates these from cracks, laps, seams, inclusions, and other rejectable defects. This is exactly the kind of discrimination that is difficult to programme explicitly but straightforward for a trained neural network.
Surface Defect Detection for Flat Products
For flat steel products — hot-rolled coils, cold-rolled strip, galvanised sheet, and coated products — DeepVision deploys line scan cameras oriented perpendicular to the direction of travel. The system inspects the entire strip width continuously at line speed, flagging defective zones for marking, diversion, or yield accounting downstream.
Detectable defects include surface cracks, laps, seams, slivers, roll marks, scratches, pits, inclusions, scale patterns exceeding tolerance, and coating defects on galvanised or painted products. Each defect type is classified separately, enabling downstream sorting decisions — a zone with cosmetic surface marks may be acceptable for structural applications while being unacceptable for exposed automotive panels.
Inspection for Long Products and Sections
Structural sections — H-beams, channels, angles, and rails — require inspection of multiple faces simultaneously. DeepVision configures multi-camera arrays to cover all exposed surfaces in a single pass. The system inspects web, flange, and edge surfaces in parallel, at production speeds up to 15 metres per second for typical structural mill applications.
For bar and rod products, DeepVision identifies surface cracks, seams, and mechanical damage that reduce fatigue life in the finished component. This is particularly important for automotive and engineering steel grades where surface quality directly determines component performance.
Machined and Formed Component Inspection
Downstream of the primary mill, machined steel and metal components require dimensional and surface quality verification. DeepVision inspects machined surfaces for tool marks, chatter, burrs, and out-of-tolerance surface finish. Formed and stamped metal components are inspected for cracks at bend radii, surface damage, and springback-related dimensional deviations.
For fasteners — bolts, screws, and studs — DeepVision performs 100% inspection of head geometry, thread profile, and surface condition at rates compatible with fastener manufacturing speeds of several hundred parts per minute.
Integration with Steel Manufacturing Systems
DeepVision integrates with existing level 2 and level 3 automation systems in steel mills, outputting defect maps that are linked to coil or bar tracking IDs. Quality records are stored per coil, per batch, or per piece depending on the production model, enabling full traceability for certifiable quality documentation. This is increasingly required for export-grade steel and for steel supplied to automotive and structural fabrication customers with documented quality requirements.
To discuss your steel or metal inspection application, contact Indus Vision for a free technical assessment.