| Description |
x, 79 leaves : illustrations ; 28 cm |
| Summary |
"This paper discusses a demodulation technique for a Dual Extrinsic Fabry-Perot Interferometric (DEFPI) sensor using an artificial neural network (ANN). The DEFPI consists of two Extrinsic Fabry-Perot Interferometric sensors that are co-located so that the same strain affects both sensors equally. One issue with a single Extrinsic Fabry-Perot Interferometric (EFPI) strain sensor is that the strain range is very limited due to the ambiguous points arising from the non-linear periodic response with respect to strain. The DEFPI sensor system removes most of the ambiguous points. The three-layered ANN proposed consists of 35 input neurons, 15 middle neurons, and one output neuron trained using the gradient decent with momentum training algorithm. The training and testing used theoretical sensor signals from DEFPI sensor sets. The neural network demodulated absolute strain for three sets of co-located DEFPI sensors over an absolute strain range of 1500 microstrain"--Abstract, leaf iii. |
|
