The purpose of this paper is to study techniques of pattern recognition in the strain field as structural health monitoring tools. The changes in the strain field may be very intense at the tip of a crack but smooth out very quickly. So trying to get information about damage occurrence from strain measurements is a difficult task, as the detected strain changes may be very small and masked by temperature drifting, load changes or any other environmental factor.
It drives to the need to include a large sensor array into the structure, which is not difficult when using optical fiber sensors. Experiments were done on a simple cantilever beam, instrumented with 32 sensors and submitted to loads and progressive damage conditions. The same approach was applied to a more complex structure, the wing of an unmanned air vehicle (UAV) made in composite materials.
Algorithms based on principal component analysis (PCA), damage indices and damage thresholds were used and shown to be simple and robust enough for this task.
The data treatment was done in a fully automated approach; an algorithm to compare and extract information from the multiple strain measurements was developed for this task.
