Stress Monitoring in Structures Under Variable Thermomechanical Conditions Using Artificial Neural Networks

Structural stress monitoring is an important task often implemented in structural health monitoring systems. Real-time estimation of a structure’s maximum stress during operation is valuable for ongoing safety evaluation and fatigue tracking, enabling scheduling maintenance actions before fatigue failure occurs. This study explores the use of artificial neural networks for on-the-fly estimation of maximum equivalent stress value in a structure under load equipped with strain and temperature sensors. Since real-world structures experience fluctuating temperatures that affect strain and stress distributions, along with varying loading conditions, the stress concentration areas can appear in different locations. Sometimes, there are too many of such locations to put sensors in all of them. For this reason, artificial neural networks are employed to process strain and temperature sensor measurements and predict maximal equivalent stress regardless of its location. The results demonstrate that precise stress estimation is achievable without direct knowledge of the current loading conditions.