Welding quality prediction model

A solution based on AI potential to predict weld quality. Based on a robotic arc welding cell, data are captured from different sources: welding equipment, external sensor systems and robots. This allows monitoring the welding process variables with the highest impact on joint quality. 

The critical points used to evaluate the quality of welded joints are mainly geometrical parameters. They must comply with the regulations applicable to each type of joint. To achieve this, an inspection system based on laser profilometry is in place, making it possible to obtain profiles of the joint surface and then processing these signals to obtain the geometry results.

These two data sources are synchronised so that a single record from each data source is allocated to each instant. As a result, artificial intelligence models trained with historical data of welds can be built, allowing predictions of the joint geometry before it is verified by an inspection system, or without it.