An approach to the structural health management (SHM) of future aerospace vehicles is presented. Such systems will need to operate robustly and intelligently in very adverse environments, and be capable of self-monitoring (and ultimately, self-repair). Networks of embedded sensors, active elements, and intelligence have been selected to form a prototypical “smart skin” for the aerospace structure, and a methodology based on multi-agent networks developed for the system to implement aspects of SHM by processes of self-organisation. Problems are broken down with the aid of a “response matrix” into one of three different scenarios: critical, sub-critical, and minor damage. From these scenarios, three components are selected, these being: (a) the formation of “impact boundaries” around damage sites, (b) self-assembling “impact networks”, and (c) shape replication. A genetic algorithm exploiting phase transitions in systems dynamics has been developed to evolve localised algorithms for impact boundary formation, addressing component (a). An ant colony optimisation (ACO) algorithm, extended by way of an adaptive dead reckoning scheme (ADRS) and which incorporates a “pause” heuristic, has been developed to address (b). Both impact boundary formation and ACO-ADRS algorithms have been successfully implemented on a “concept demonstrator”, while shape replication algorithms addressing component (c) have been successfully simulated.
Feature extraction and soft computing methods for aerospace structure defect classification