<title>Smart integrated-optics displacement/force sensor based on speckle pattern detection using neural-net with 0.1-A resolution</title>

Abstract

This paper describes the usage of neural networks in the application of speckle patterns, as seen at the output of an optical waveguide, for sensing displacement/force. The neural network trained for displacement values within a 1 .5J.Lm range is found to generalise with less than JIm error in the input range. It is found that a functional-link net with 1 1 functional links and 3 neurons can be trained for a convergence criterion of le-05 in less than 3000 iterations. The error in the individual targeted output during training was less than JIm. Thus an integrated-optics sensor is now feasible.Most sensors suffer from drifts due to environmental changes or structural changes within the sensor itself. Also, some sensors are affected by electromagnetic interference. A sensor based on speckle pattern detection will also suffer from drifts due to environmental 54/SPIE Vol. 1793 Integrated Optics and Microstructures (1992) 0-81 94-0972-3/93/$4.00 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/27/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx