On risk-based active learning for structural health monitoring

A primary motivation for the development and implementation of structural health monitoring systems, is prospect gaining ability to make informed decisions regarding operation maintenance structures infrastructure. Unfortunately, descriptive labels measured data corresponding health-state information structure interest are seldom available prior a system. This issue limits applicability traditional supervised unsupervised approaches machine learning in statistical classifiers decision-supporting SHM systems. The current paper presents risk-based formulation active learning, which querying class-label guided by expected value said each incipient point. When applied monitoring, class can be mapped onto inspection order determine its state. In paper, process explained visualised via representative numerical example subsequently Z24 Bridge benchmark. results case studies indicate that decision-maker's performance improved classifier, such decision itself taken into account.