Stream Owl

Arc welding has been a fundamental and key technology in many industrial manufacturing industries (aerospace, automotive, home appliances, etc.) for years and the development of efficient automated weld quality control methods has been the focus of research (especially in the era of Industry 4.0) to increase the quality of the products and detect defects in the welded joints. Radiography testing is a popular Non-Destructive Testing (NDT) method for weld quality control; however, it is still performed with primitive means, mostly by human visual inspection of X-Ray images by certified radiographers, which makes the whole process inefficient, non-scalable, and costly. In this project, Stream Owl targets Challenge 6 of the EARASHI Open Call on “Collaboration between AI and Human supervisors to solve complex problems”and Mondragon will provide knowledge and expertise thanks to its building block ‘Deep Learning based Industrial Quality Inspection Methodology’ (BB#16).

More specifically, the AIWELd project by Stream Owl aims at integrating Artificial Intelligence (AI) with traditional practices and experts’ knowledge to transform radiography testing (a process that relies on human experience and skills) into an objective decision support system fully integrated in the industrial weld quality control process, which can be used as an empowerment tool by radiographers and improve their daily routine job, increase productivity, decrease stress levels, and improve the quality of the final products. The development of the weld quality assurance method is based on the preparation of a very large dataset of X-Ray weld images, the adoption of latest developments in Deep Learning and Artificial Intelligence, and the testing of the software tool with expert radiographers to ensure that the final product can increase their productivity and ensure that it can be used as an efficient weld quality assurance method.