Strain-based fault detection of bolted truss structures using machine learning

Abstract

The initial design of baggage-lifting machine structures is primarily based on safety and reliability, but they are often damaged because of unforeseen circumstances and overloads. In this study, a machine learning–based logistic regression method for detecting structural damage to bolted truss structures during field work is proposed. Multiple strain gauges attached to the front of the truss model record the amount of deformation occurring in the member when the vertical load generated at the end of the model is applied. In this process, the scatter or error caused by the sample is analyzed, and the data processing method is presented. Experimental results demonstrate that this method provides a good quantitative basis for fault detection, and it can be effectively applied to partial representative data when handling large datasets.