Predictive Modelling of Thermal Wave Propagation Using Ml Techniques

Corrosion is a major challenge in industrial applications, leading to material degradation, safety risks, and high maintenance costs. Traditional detection techniques, such as ultrasonic and radiographic testing, often require invasive procedures or specialized equipment, making large-scale monitoring difficult. This project aims to develop an automated, non-invasive corrosion detection framework using passive infrared thermography and machine learning-based image processing to enhance detection accuracy and efficiency. The core problem addressed in this study is the difficulty in identifying corrosion early without complex hardware setups or manual inspection. To overcome this, we integrate Contrast Limited Adaptive Histogram Equalization (CLAHE) for contrast enhancement, Gray Level Co-occurrence Matrix (GLCM) for texture-based feature extraction, and K-Means clustering for automated segmentation of corroded regions. These techniques help improve corrosion visibility, accurately segment affected areas, and quantify severity levels based on pixel intensity analysis. After implementing and validating this framework using thermal imaging datasets, our findings show that CLAHE significantly enhances corrosion visibility, K-Means clustering effectively distinguishes corroded versus non-corroded regions, and GLCM analysis reliably quantifies corrosion severity. This approach proves to be a cost-effective, scalable, and efficient solution for corrosion assessment in industrial environments. The study concludes that integrating machine learning with passive thermography can improve corrosion detection accuracy while reducing hardware complexity. Future work will explore real-time corrosion detection using deep learning models and hyperspectral imaging for enhanced defect characterization.