Skip to main content

Zero defect manufacturing - ZeroDefectWeld

Automated defect detection in welds with computer vision and machine learning.


Business Category
Technical Category
Computer vision

StairwAI First Open Call Challenge Challenge 5: Zero defect manufacturing


What is the challenge that is being addressed?

The project targets the "Zero defect manufacturing" challenge and aspires to develop an automated Deep Learning (DL) framework for Non-Destructive Testing (NDT) for the detection of weld defects from X-Ray images based on computer vision. Defect detection in welded joints is still performed with primitive means, mostly by human visual inspection by expert radiographers, which is an inefficient, non-scalable, non-standardised and costly method. Multi-class detection and classification is most wanted in the weld industry; manufactures not only need to know that there is a defect, but they need to know what type of defect for root cause analysis.


What is the AI solution the project plans to implement?

We propose an automated defect detection method based on advanced Computer Vision approaches (e.g., contour tracking, morphological operators, edge detection, histogram equalisation). Also, our proposed solution replaces the old-fashioned and inappropriate bounding boxes of the annotation tools which introduce too much of a noise margin in the X-Ray images. We propose the development of a multi-class classification method based on Deep Learning which can detect the type and size of a defect in welded joint.

The introduction of efficient Artificial Intelligence (AI) methods in the welding process for automated quality assurance enables improvements to the quality and safety of welds. In addition, it contributes to cost-savings by reducing the recalls due to quality issues by sub-optimal welding and reduces production downtime caused by non-automated weld monitoring.

In this proposal, it is envisaged that the integration of efficient Artificial Intelligence (AI) methods, and Deep Learning in particular, in the classification process of defect detection in welded joints will lead to improvements in performance and robustness of our product. This development will greatly affect the development process for the consideration of multiple types of defects that may appear in welds (more than 15 types of weld defects exist, see 

Moreover, this project will stimulate the generation and use of weld databases which are currently limited. It is expected that the project will boost the interest of the community in AI methods and the development of machine learning tools and methodologies for multi-class classification in welded joints.

StairwAI use case


Who will help implement this solution?



This project is funded and supported through the first Open Call of the StairwAI project.

The selected company will receive a maximum of €60.000, that includes:

  1. fixed lump-sum: up to € 26 000
  2. vouchers: up to € 10 000 to access HW Resources and up to € 24 000 for AI experts (mentors and system integrators).
    • Support from Technical and Business Experts and access to specialized tools as a part of a Support Program.
    • levelling up the business in terms of AI maturity.
    • Improving key processes in the company, which will increase competitiveness.


Learn more about StairwAI's Open Calls:

StairwAI logo