AI and IoT-based concrete column base cover localization and degradation detection algorithm using deep learning techniques

Abstract

Internet of Things (IoT) and Artificial Intelligence (AI) technologies are currently replacing the traditional methods of handling buildings, infrastructure, and facilities design, control, and maintenance due to their precision and ease of use. This paper proposes a novel automated algorithm for the health monitoring of concrete column base cover degradation based on IoT and the state-of-the-art deep learning framework, Convolutional Neural Network (CNN). This technique is developed for instance detection and localization of the major types of column defects. Three deep machine learning training models; namely, Resnet-50, Googlenet, and Visual Geometry Group (VGG19), with 7 different network configurations and inputs were studied and compared for their classification performance and certainty. Despite that, a few articles consider the certainty of the CNN classification results, this work investigates the certainty and employs the classification error score as a new performance measure. The results of this study demonstrated the effectiveness of the proposed defect detection and localization algorithm as it managed to read all barcodes, localize defective columns, and binary classify the condition of the concrete covers against their surrounding objects. They also showed that the VGG19 network outperformed the other addressed network models and configurations. The VGG19 network yielded a health condition classification accuracy of 100% with an RMSE of 0.33% and a maximum classification error score of 0.87 %.