Development of a Predictive Maintenance System for Military Applications Using Visual Language Models

The Grupo de Tratamiento de Imágenes (GTI) is proud to present a new End of Master Project developed by Enol Ayo Sando, titled "Development of a Predictive Maintenance System for Military Applications Using Visual Language Models." This project was recently presented at Escuela Técnica Superior de Ingenieros de Telecomunicación, showcasing an innovative approach to defect detection in steel materials used in the military industry.

Addressing Challenges in Military-Grade Steel Inspection

Steel plays a crucial role in military applications due to its durability and resilience. However, surface defects can significantly affect its performance and reliability. Traditional inspection methods are often time-consuming and expensive, necessitating the development of more efficient alternatives. This research leverages artificial intelligence (AI) and deep learning to streamline defect detection and maintenance processes.

The Role of Visual Language Models in Predictive Maintenance

The project explores the potential of visual language models, specifically CLIPSeg, for detecting and segmenting defects in steel plates. A key advantage of this approach is its Zero-Shot Learning capability, which enables the model to identify anomalies without requiring extensive annotated datasets. This feature significantly reduces costs while improving the efficiency and accuracy of defect detection.

Study Objectives

The research focuses on three main objectives:

• • • Implementation of CLIPSeg for identifying defects in steel materials.
    • Model adaptation and optimization to enhance detection accuracy.
    • Performance evaluation using publicly available defect databases and comparison with state-of-the-art methods.

Advancing Military Predictive Maintenance

This project marks a significant step forward in military predictive maintenance, offering a faster and more precise alternative to traditional inspection methods. By integrating cutting-edge AI techniques, this research contributes to enhancing the longevity and reliability of critical military assets.