A Robust Model for the Identification of Bolt Loosening in Steel Structures Based on Modified Convolutional Neural Network

When a bolt becomes free owing to stress, pressure, or bending stresses, this is called spontaneous releasing. A fastened joint may lose all of its preload with even a little amount of spin. That's the number one reason fasteners go loose. Preventing bridge collapse is often possible if missing bolts and nuts are identified before they are used. The focus of this study is on applying deep learning methods to create a new architecture for the early identification of bolt-nut losses in steel structures. The goals are to use deep learning methods to create a methodology for detecting nuts, bolts, and nut slots; to implement that strategy in Python; and to evaluate the effectiveness of that architecture. An 8x8 grid of bridge photos was used as input for CNN-LSTM methods. The CNN's unusually high recognition rate may be explained by the sequential nature of its extraction through convolution and pooling. Nevertheless, statistical analysis of the data indicated that there was no statistically significant distinction in object identification and classification performance among the methods used in the final design. The CNN model is recommended for spotting nuts, bolts, and nut holes on structural steel bridges because to its excellent detection rate, high accuracy, and fast throughput.