Multiplexed fiber optic strain sensor system

ABSTRACT This paper presents a multimode fiber optic sensor system suitable for large smart structures. Bymany observations of speckle patterns from different multimode fibers, it seems that each speckle pattern has its own signature and uniquely defining its fiber. This prompted us to superimpose (multiplexed) the speckle patterns generated by two different fibers, and detect the resultant pattern.A two-stage feature extraction algorithm is then applied on the image, which reduces the dimensionof the pattern vector. Next, a backpropagation neural network with single hidden layer is trained inmapping the feature vectors to the stress applied on each fibers. Series of experiments are conducted on two fiber sensors glued on a square shape plate under different point loads. The sensors successfully estimated that which fiber was under stress with light or heavy weight.Keywords: Strain sensors, multimode fibers, speckle pattern, neural networks, smart structures. 1. INTRODUCTION Observation of the output intensity pattern of a multimode optical fiber under physical disturbances