MULTI-IA

development of a system for material detection, multiple defect identification, and anomaly prediction

MULTI-IA: Development of a system for material and multiple defect detection and anomaly prediction based on computer vision, artificial intelligence, and the Internet of Things (IoT).

Challenges

Analyse different types of defects in various materials in real time without requiring multiple inspection stations. In other words, the system will be scalable and customisable for quality control of any component manufactured on a production line.

The application of AI will enable correlation between production process variables and possible anomalies, allowing predictive capabilities.

Solution

Design and development of a new real-time system based on computer vision and deep learning, capable of multi-defect and multi-material detection and anomaly prediction in a single station.

The system will include image analysis techniques able to detect when a defect occurs in a manufactured component and predict the appearance of anomalies according to the production process variables.

Se desarrollará un procedimiento de etiquetado semisupervisado que permitirá la generación de conocimiento de tal manera que el sistema tenga capacidad de aprender de forma incremental.

A semi-supervised labelling procedure will be developed to generate knowledge, enabling the system to learn incrementally.

AI models will be deployed at the edge—on hardware physically located on the production line—to minimise response time and ensure processing speeds suitable for strict industrial requirements.

Medium- and long-term storage (local, cloud, or hybrid) of collected data will enable subsequent analysis.

The solution will be validated in two different use cases to confirm satisfactory results.