Deep Learning-Based Durability Prediction System For Composite Materials In Harsh Climates

Abstract

This work introduces a new artificial intelligence model for estimating how long composite materials resist damage during difficult weather conditions. A CNN-based network built using PyTorch allowed the model to detect both key visual elements from composite images and changes over time taken from the sensor data. Using this technique, material degradation can be accurately predicted, allowing greater accuracy and fewer errors when measured against traditional CNN and LSTM models. With attention mechanisms, the model can point out overlooked environmental conditions that might affect durability of the equipment. A broad selection of tests on a complex dataset illustrates the effectiveness and practical benefits of the system for supporting composite maintenance and lifespan management under severe weather conditions. Even though the system involves more calculations, the hybrid model combines both strong and flexible features. The next phase of work will try to broaden the range of data and allow the model to deploy swiftly on devices with minimal resources. The developed methods offer a reliable base for predicting durability of materials in challenging environments through AI.