Artificial Intelligence and 3D Scanning: The Future of Weld Inspection?

How Do 3D Scanning and Artificial Intelligence Work?

The modernisation of industrial inspection relies on three distinct stages that must be considered separately. Physical data acquisition records the exact shape of the component using an optical sensor. Mathematical analysis then processes these measurements using specialised software. Finally, the ultimate decision rests exclusively with the person responsible for quality assurance.

3D Scanning Converts Surface Geometry into Measurements

3D scanning works by sweeping a laser beam across a metal surface to calculate distances without any physical contact. This optical measurement generates a point cloud, a collection of three-dimensional coordinates describing the exact geometry of the weld bead.

A 3D laser profilometer can record data with a very fine measurement spacing of 0.07 mm between points. This scanning process reconstructs an accurate surface profile using optical triangulation. It enables the assessment of local geometry at the weld toe, including the transition radius and weld angle. These two parameters directly influence the fatigue resistance of mechanical structures.

Raw laser measurement data often contain digital noise caused by metallic reflections or vibrations. It is therefore essential to apply digital filters, such as Gaussian or low-pass filters, to smooth the profile without distorting the actual geometry. This accurate acquisition of the real surface profile complements engineering processes such as numerical welding simulation, which predicts joint deformation before manufacturing.

AI Identifies Suspicious Patterns and Signals

Artificial intelligence comes into play immediately after digitisation to automatically identify manufacturing anomalies. It relies on an algorithm, a sequence of automated computer instructions that performs complex calculations.

To operate, the system must first be provided with training data. These data consist of reference images or signals used to train the artificial neural network. The neural network compares measured 3D profiles, radiographic images or ultrasonic signals with the examples it has learned during training. The system immediately alerts the inspector to areas showing geometric deviations or suspicious signals.

This software provides only a preliminary screening alert. In the absence of a standard specifically certifying AI for this application, the decision to accept or reject a component remains the responsibility of a human inspector qualified in accordance with ISO 9712.

Understanding these operating principles makes it easier to assess the actual capabilities of these technologies when dealing with defects encountered in the workshop.

What Defects Can These Technologies Detect?

To ensure effective quality control, the person responsible must select the inspection method best suited to the location of the defect. A surface imperfection cannot be detected using the same tools as a cavity hidden deep inside the metal. Furthermore, the presence of a measured imperfection does not automatically mean that the component must be rejected.

Surface Dimensions and Visible Imperfections Measured in 3D

3D scanning directly measures the visible geometric features on the external surface of welded joints. The system automatically calculates weld bead width, height and convexity, as well as undercut depth and plate misalignment.

These continuous optical measurements can replace conventional manual inspections performed using mechanical measuring instruments such as the Cam Gauge welding gauge or the Hi-Lo Gauge. Manual gauges provide isolated measurements at specific points. In contrast, a 3D scanner can inspect the entire length of a welded joint without leaving areas unchecked.

Visual inspection procedures are governed by ISO 17637. The tolerance limits and quality levels for measured weld imperfections are specified in ISO 5817.

Internal Defects Analysed Using Appropriate NDT Methods

To inspect the interior of a metal component, inspectors must use NDT (Non-Destructive Testing), which encompasses methods used to assess the integrity of a component without damaging or destroying it. ISO 17635 establishes the general rules for selecting the appropriate non-destructive testing method according to the material and type of welded joint.

An optical 3D scanner cannot detect anomalies located inside the material. It cannot identify internal porosity, slag inclusions, internal cracks or lack of fusion. To analyse these hidden defects, artificial intelligence must be combined with inspection technologies capable of examining the interior of the material, such as digital X-ray radiography or ultrasonic testing. This combination of technologies helps assess welding defects and acceptance criteria required by applicable standards and regulations.