Evaluation of the quality of welds through two main approaches: 1) Image analysis, where artificial vision combined with deep learning is used to classify the welds according to their suitability, as well as identify and segment defects such as pores or lack of fusion. 2) Data processing collected through a laser profilometer, where the penetration of a weld bead is predicted based on its visible face or part, thus avoiding piece cuts or non-destructive techniques for its evaluation.
Deployment and Application
Computer vision-based inspection systems
Computer vision-based inspection system
Profilometer for joint tracking and geometry inspection.
Profilometer for joint tracking and geometry inspection.
Models for the classification of suitable/unsuitable welds based on optical images. This service is complemented by generative artificial intelligence algorithms that allow the creation of synthetic samples to expand the datasets.
Time series analysis to develop predictive models to anticipate errors or estimation algorithms to infer the hidden part of welds.