Abstract
The prediction of a microbend sensor response using Artificial Neural Networks (ANNs) has been investigated in this paper. Experiments were conducted with different microbend sensor configurations. By using the one experiment's input and output experimental data among the conducted experiments, the ability of the ANNs in the prediction of sensor response was analyzed. In the training process of the ANN, multi layer perceptron training algorithm such as, Resillient Backpropagation, Levenberg-Marquardt and Fletcher-Reeves Conjugate Gradient algorithms were used. After training process, network was tested and it was seen that, all the algorithms used can predict the sensor response with small errors. Hence, it was concluded that, ANNs can be used to decrease the fault tolerance of fiber optic microbend sensors, to design intelligent and more robust sensors.