An Evaluation of Machine Learning Algorithms in an Experimental Structural Health Monitoring System Incorporating LoRa IoT Connectivity

Abstract

The Internet of Things (IoT) offers dynamic mechanisms and methodologies for a broad number of practical applications by virtue of its integrated powerful advantages encompassing immense reliability, and superb robustness. This will expedite and simplify the deployment of IoT devices in civil structures and building settings for structural health monitoring (SHM) applications, including early warning systems (EWS). A SHM system extracts and provides information about variations in an individual component or in the complete structure. In this paper, an experimental SHM system incorporating the LoRa wireless IoT connectivity is presented, which includes assortment of sensor devices to acquire measurements of commonly monitored physical variables in typical SHM systems. The acquired concurrent measurements performed on the structure are aggregated at a designated cloud server, where they are analyzed, to predict the health status of the monitored structure. To that end, we conduct a comparative performance analysis of a collection of machine learning (ML) classification algorithms to evaluate their detection capacities in determining faults or variations in a monitored structure state. Numerical analysis results demonstrated that in our SHM system, the presence of a fault could be effectively predicted by means of a subset of the collection of considered ML classification algorithms. Computed evaluation metrics to characterize performance of our SHM system served to identify an optimum ML algorithm based on attained results, in addition to a training methodology for employment in SHM systems, to accurately detect the presence of a fault or damage in the monitored structure.