This paper studies the problem of optimal sensor placementfor impact detection and location in composite materials. The studyinvolves a simple impact experiment on a composite box panel. Thetime-varying strain data are measured using piezoceramic sensors. Aneffective impact detection procedure is established using a neuralnetwork approach. The procedure determines the location andamplitude of impacts. A genetic algorithm is used to determine theoptimum sensor positions for a diagnostic system. The main objectof the paper is to studyfail-safedistributions, i.e.thosewhose sub-distributions also have a low probability of detectionerror. The results are validated against an exhaustive search. Thestudy shows that genetic algorithms combined with neural networkscan be effectively used to find near-optimal sensor distributionsfor damage detection. The methods presented are generic and can beused in similar sensor position problems.
展开▼
