Investigating Experimental Repeatability and Feature Consistency in Vibration-Based SHM

Abstract

Structural health monitoring (SHM) has been an active research area for the last three decades, and has accumulated a number of critical advances over that period, as can be seen in the literature. However, SHM is still facing challenges because of the paucity of damage-state data, operational and environmental fluctuations, repeatability issues, and changes in boundary conditions. These issues present as inconsistencies in the captured features and can have a huge impact on the practical implementation, but more critically on the generalisation of the technology. Population-based SHM has been designed to address some of these concerns by modelling and transferring missing information using data collected from groups of similar structures. In this work, an experimental campaign is discussed, in which vibration data were collected over a series of tests on a set of four healthy, full-scale composite helicopter blades. During the tests, variability was introduced by adjusting boundary conditions between each testing repetition. It is well known that changes of boundary conditions, even from careful repositioning of the structure, can alter selected feature’s properties, changing dynamic responses from normal condition and thus raising false alarms which degrade the effectiveness of SHM. In addition, nominally-identical structures may have slight differences in geometry and/or material properties. These variations can present as changes in the dynamic characteristics of the structure, which can be very problematic for SHM based on machine learning. This paper demonstrates the applicability of SHM when such deviations occur. In this work, a normal condition for the set of helicopter blades is established and tested via a point-wise outlier analysis approach and by defining a general model for the blades, called a population form, using Gaussian process regression.