Zum Hauptinhalt springen
- Metal and steel industry

Surface inspection in the Steel and metal processing - AI revolutionizes quality control

The use of state-of-the-art AI algorithms in steel and metal processing ensures the highest quality at every stage of production: from the raw material to the forming and machining of components, right through to final assembly.

From defect to competitive advantage - with AI, quality assurance becomes a driver of efficiency and growth.

In the steel and metal processing industry, every detail counts: from material quality and dimensional accuracy to weld seams and surfaces. The highest product quality is essential, but manual inspections and traditional methods such as anomaly detection quickly reach their limits with increasing product diversity, tight tolerances and high deadline pressure.

The AI-supported quality assurance offers companies decisive advantages along the entire value chain. Instead of merely detecting anomalies, the deep learning-based technology is able to precisely recognize different error patterns and thus consistently eliminate pseudo-defects. At the same time, automatic logging and storage ensures that all defect types are documented and traceable at all times. 36ZERO Vision's scalability and flexibility are particularly impressive in steel and metal processing, where high product diversity and changing requirements are part of everyday life. The solution is extremely cost-efficient, as it is hardware-agnostic and can be easily integrated into existing systems, thus reducing investment and maintenance costs. It also contributes to stable processes, less waste and reliable compliance with quality standards - clear added value in terms of sustainability and compliance with standards.

The possible applications are diverse: in raw material processing, AI enables the inspection of sheet metal, profiles, pipes and wires. In forming technology, it is used for surface inspection of cast and forged parts, while in joining technology, weld seams and soldered joints are reliably inspected. In machining, the technology also ensures reliable burr and chamfer inspection after turning, milling or grinding work. Finally, it supports both in-line and end-of-line inspections in assembly and final production. This makes quality assurance not only more efficient, but also a real competitive advantage.

The cloud platform is used to develop AI models individually for each application. It is initially trained with known error patterns and can be continuously improved through interaction with employees. It adapts flexibly to new product variants, applications and production conditions and standardizes quality decisions worldwide.

2D images from industrial cameras are analyzed autonomously, on-premise and in real time with the trained models, defect patterns are detected with the highest precision and workpieces are objectively evaluated. Surfaces, shape and dimensional accuracy can be checked without delay.

In addition to visual inspection, companies also rely on ultrasonic testing, X-ray testing, magnetic particle testing, penetrant testing, hardness testing and metallography, depending on requirements.


Typical types of errors in 
Steel and metal processing industry:

1st material


  • Defective surfaces (mill scale, cracks, chatter marks)
  • Inclusions (e.g. slag, oxides, sulphides)
  • Structural defects (e.g. coarse-grained structure, hardening cracks)

2. manufacturing


  • Shape and dimensional errors (insufficient dimensional accuracy, distortion, inaccuracies)
  • Surface defects (scoring, scratches, burr formation, built-up edges during turning/milling)
  • Heat treatment defects (e.g. overheating, decarburization, hardening cracks, insufficient hardness)
  • Welding defects (pore formation, binding defects, penetration notches)

3. handling errors


  • Transport damage, pressure marks, incorrect storage (e.g. corrosion)
  • Mixing up parts
  • Indentations, dents and chipping on edges and corners due to knocks, impacts or improper grippers

Challenges for the Quality management

Complexity of the manufacturing processes

Casting, forging, welding, heat treatment, machining

Maximum precision vs. cost and deadline pressure

Balance between productivity and quality

Sustainability and resource efficiency

Monitoring and improvement of energy-intensive processes

Certifications and standards

Compliance with strict norms and standards (ISO, IATF, EN, ASME), as well as extensive documentation requirements and audits

Product service life and operational safety

Defects in the material or in production can lead to safety risks

Digitalization & Industry 4.0

Shortage of skilled workers in the metal industry makes it difficult to maintain consistently high quality
sweat spatter
The weld seam is uneven, which indicates inadequate guidance or fluctuating heat input.
course
A small gas bubble is visible in the weld seam, caused by trapped air or impurities.
bubble_small
There is welding spatter around the seam due to excessive current or insufficient protection against spatter formation.

Added value through AI-supported quality assurance with Error pattern recognition


  • Cost reduction (fewer rejects, less rework, fewer complaints)
  • More efficient processes (more stable processes, better utilization of machines)
  • Quality improvement (longer service life, higher customer satisfaction)
  • Competitive advantages (better reputation, stronger market position)
  • Legal security (compliance with standards & certifications, e.g. ISO, TÜV, EN)

AI - applications in the steel and metalworking industry

  • 1 // Raw material processing:
    Inspection of sheet metal, profiles, tubes, bars and wires in forges, steelworks and rolling mills
  • 2 // Forming technology:
    Surface inspection of cast steel, die casting, investment casting, continuous casting, forged and pressed materials
  • 3 // Joining technology:
    Weld seam inspection and inspection of soldered joints
  • 4 // Machining & machining:
    Chamfer inspection, burr inspection after turning, milling, drilling and grinding work
  • 5 // Assembly & finishing:
    In-line inspection and end-of-line testing

The most important Questions
at a glance:


What surface defects and structural defects can AI-based visual quality inspection detect in metal and steel processing?

36ZERO Vision recognizes and classifies the following defect spectrum in metal and steel processing: surface defects such as scratches, score marks, rust, temper colors, rolling defects, and oxidation damage on sheets, strips, coils, and profiles; weld defects such as pores, inclusions, seam interruptions, spatter, and lack of fusion; casting and forging defects such as shrinkage cavities, cold shuts, shrinkage cracks, shape deviations, and inclusions; as well as dimensional deviations and positioning errors in sheet metal processing, forming, and assembly. The deep learning algorithm works particularly reliably on mirror-like, highly reflective metal surfaces, as it distinguishes reflections and lighting variations from the pixel structure of real defect features.

How robust is AI-based visual quality inspection under the demanding environmental conditions in metal and steel production?

AI-based visual quality inspection is inherently robust under the challenging environmental conditions of the metal and steel industry – dust, heat, vibrations, highly varying lighting, and changing material batches. 36ZERO Vision learns error patterns directly from real production data, making it adaptive to environmental fluctuations. The edge client runs entirely on the edge without internet dependency, reliably, even in network-critical and geographically isolated industrial environments in steel and metal processing.

How quickly does AI-based visual quality inspection pay for itself in the metal processing industry, and what are the main cost drivers?

36ZERO Vision often pays for itself in the metal processing industry within 5 months. The main savings potential lies in scrap reduction through early inline error detection before expensive downstream processing steps, the elimination of false positives, the reduction of personnel costs for manual visual inspection, as well as the avoidance of customer complaints and returns. Since existing camera infrastructure can be used via retrofit, high upfront hardware investments are also eliminated, further shortening the payback period.

Can AI-based visual quality inspection meet the high throughput requirements in steel strip and continuous sheet metal production in real-time?

The 36ZERO Vision Edge client achieves inference speeds of under 20 milliseconds per image evaluation in steel and sheet metal production. As the platform is completely hardware-agnostic, it can be combined with camera systems suitable for high throughput – from high-speed line cameras for continuous steel strip inspection to area cameras for sheet metal part inspection in stamping and punching operations. The specific achievable inspection speed in meters per minute or parts per hour depends on camera resolution, lighting concept, minimum defect size, and handling system. These parameters are jointly specified in the technical proof of concept.

Digital transformation through AI in the Metal and steel industry