On Self-organising Diagnostics in Impact Sensing Networks

Structural health management (SHM) of safety-critical structures requires multiple capabilities: sensing, assessment, diagnostics, prognostics, repair, etc. This paper presents a capability for self-organising diagnosis by a group of autonomous sensing agents in a distributed sensing and processing SHM network. The diagnostics involves acoustic emission waves emitted as a result of a sudden release of energy during impacts and detected by the multi-agent network. Several diagnostic techniques identifying the nature and severity of damage at multiple sites are investigated, and the self-organising maps (Kohonen neural networks) are shown to outperform the standard k-means algorithm in both timeand frequency domains.