Investigation on structural health monitoring for industrial IOT applications using deep learning approaches

Abstract

Structural Health Monitoring (SHM) is a mechanism of developing newlinea damage identification approach for civil, aerospace, and mechanical newlineengineering infrastructure. If the machinery or structure gets damaged, it does newlinenot mean that it loses its functionality, but it is decided that the system is not newlinein optimal condition, and suppose if the damage in the structure increases then newlineit may collapse. Accordingly, Structural Health Monitoring (SHM) is the newlineprocess employed to find the damage by periodically collecting the data newlinethrough sensors such that it allows to detect the damage of the system and to newlinemodel the health status of the structure. However, monitoring applications newlineenclose different disciplines from aerospace to diagnostics of machines as newlinewell as mechanical systems. This research developed three different newlinecontributions with the machine learning and deep learning methods in the IoT newlineparadigm to find the health status of the structure. The proposed system has newlinebeen applied for monitoring small wind turbines and civil infrastructures. newlineThe system has been developed using Optimized Artificial Neural newlineNetwork (OANN) for the proactive maintenance of small wind turbines as newlineone of the contributions that would help to prolong the lifetime of the wind newlineturbines. The monitoring system for civil structures has been designed with newlinethe Bat Ant Lion Optimization based Generative Adversarial Network newline(BALO based GAN) approach. A routing protocol is integrated with the newlineMonarch-EarthWorm Algorithm (Monarch-EWA) for selecting the secure newlineand optimal path in network routing. A Deep Neuro-Fuzzy Network (DNFN) newlineis constructed for measuring the health status of the structure and its behavior newlineat the time. The proposed model has been observed to have better accuracy, newlinesensitivity, specificity, and throughput compared with the existing systems. newline