Coal Preparation » General
This report presents the progress and findings of a project which focused on addressing the challenges in inspecting and assessing the condition of corrosion protection coatings in coal mining infrastructure, particularly coal handling and preparation plants. Traditional inspection methods, which involve qualified inspectors assessing the condition of protective coatings using a rating system, are labor-intensive, subjective, and expensive, especially for large and complex structures. As protective coatings are essential for steel structures to protect them from environmental damage, there is a need for a more reliable, cost effective, and objective evaluation method.
Recent advancements in machine learning and machine vision have led to studies exploring automated approaches for detecting corrosion and coating defects, offering real-time and low cost alternatives. However, quantifying and distinguishing the severity levels of corrosion and coating defects remains challenging due to the complex relationship between defect generation mechanisms and their external expressions. To address these challenges, this project aimed to develop new computer vision and deep learning techniques by creating a hybrid framework that combines convolutional neural networks (CNN), transfer learning, and decision level image fusion. This framework can effectively identify the severity of corrosion and coating defects in CHPPs.
The performance of this system has been tested using real world images and demonstrated promising results. Additionally, a pixel-wise corrosion segmentation model using CNN and DeepLabV3 configuration was developed. This model was tested on available corrosion image datasets and showed potential for accurately quantifying corrosion, and assessing the structural capacity of CHPP. As a result, the proposed methodology aims to automatically evaluate surface conditions, quantify affected areas, enable data-based decision making, increase the reliability of corrosion condition assessments, and reduce the time between inspections and repairs through faster assessment responses.