ICUBE, advanced machine vision technologies for welding defect identification
In the project «Advanced artificial vision technologies for the identification of welding defects», ICUBE received technical advice from our partner IDONIAL Technology Center. Through the services it provides within the framework of AsDIH, IDONIAL explored the use of 2D and 3D artificial vision technologies in performing non-destructive testing (NDT) on welded joints, specifically those in which the visual component is key, such as visual inspection and penetrant testing.
What problems does it seek to solve or improve?
ICUBE has personnel qualified in performing NDT and, for this reason, they detected the need to improve the uncertainty associated with the inspector’s acuity and skill. Currently, artificial vision systems are known to exist, but they are focused on serial production lines, with acceptable but limited detection quality. The quality of detection can be improved with machine learning techniques or artificial intelligence applied to image processing.
In this specific case, the interest lies in proposing a tool to support the performance of non-serial NDT, even on large or difficult-to-access elements, which increases the comfort, efficiency, and confidence of an inspector already qualified for inspection.
These are the objectives to be achieved:
- Identification of artificial vision solutions suitable for weld inspection, specifically in visual inspection and penetrant testing applications.
- Evaluation of the potential of these solutions in terms of accuracy, efficiency, and cost.
- Proposal of a solution at both the hardware and software levels and validation through proof of concept. Comparison with traditional NDT methods in terms of effectiveness, accuracy, and feasibility of the technology.
What digital technology is used to solve the problem?
They use advanced vision technologies based on machine learning models, trained with a dataset developed at IDONIAL. The company has access to the facilities of the technology centre ‘s Collaborative Robotics Laboratory, which has 2D and 3D artificial vision systems for inspection and quality control.
What results does it produce, and what is its impact?
Within the framework of this project, a method for generating generalised and robust training data has been developed in accordance with the applicable specifications of international regulations for visual inspection and penetrant testing of fusion welds, which has made it possible to validate an application of the trained vision system by satisfactorily detecting all the indicated defects.
The services provided have enabled ICUBE to explore and validate the use of artificial vision technologies to support non-destructive testing, confirming the technical, operational, and economic viability of the system developed.
About the company
ICUBE is a technology-based company whose work philosophy is based on continuous innovation. They aim to achieve the best results in each of the projects they undertake.
This company creates effective, accessible engineering solutions tailored to its customers, understanding their needs and offering them support with quality, innovative solutions for their projects.
