Stochastic method for in-situ damage analysis

TL;DR

This paper introduces a stochastic process-based method for in-situ structural health monitoring that effectively detects damage by analyzing changes in deterministic dynamics under realistic noisy conditions.

Contribution

The paper presents a novel approach extending stochastic equation reconstruction to real-world noisy excitations for damage detection in structures.

Findings

Damage-related slope changes are more sensitive indicators than eigenfrequency shifts.

The method successfully distinguishes damaged from undamaged structures under turbulent wind noise.

Experimental validation confirms the effectiveness of the stochastic approach in real-world conditions.

Abstract

Based on the physics of stochastic processes we present a new approach for structural health monitoring. We show that the new method allows for an in-situ analysis of the elastic features of a mechanical structure even for realistic excitations with correlated noise as it appears in real-world situations. In particular an experimental set-up of undamaged and damaged beam structures was exposed to a noisy excitation under turbulent wind conditions. The method of reconstructing stochastic equations from measured data has been extended to realistic noisy excitations like those given here. In our analysis the deterministic part is separated from the stochastic dynamics of the system and we show that the slope of the deterministic part, which is linked to mechanical features of the material, changes sensitively with increasing damage. The results are more significant than corresponding changes in eigenfrequencies, as commonly used for structural health monitoring.