Reliable Quantification of the Performance of SHM Systems

Abstract

Structural Health Monitoring has great potential to replace in-service NDT, giving the opportunity to increase intervals between maintenance shutdowns and to have early warning of the need for component replacement. However, it is essential that the performance of the SHM system is thoroughly evaluated to give confidence that it will detect the required defects at an acceptable level of false calls. This is more difficult than the evaluation of NDT methods as the SHM system will operate across a wide range of environmental operating conditions (EOCs). Although damage can be applied to test structures in a laboratory environment to evaluate the performance, it is often prohibitively expensive to comprehensively investigate the sensitivity of damage detection on different damage types, sizes, and locations under different EOC. In this paper, a cost-effective methodology is presented to evaluate the performance and estimate the sensitivity of an SHM system by synthesizing undamaged records collected under varying EOCs and simulated artificial damage signals. We produce the receiver operating characteristics (ROC) of damage detection methods on different synthetic datasets to predict the performance in practical scenarios.