A novel optical fiber system for non-touch detecting roll wear based on neural network and software imitation

Abstract

A non-touch on-line optical fiber measurement system for roll wears is designed, based on the principle of reflective displacement fiber-optic sensor. By using a novel probe of three optical fibers in equal transverse space, the effects of fluctuations in the light source, reflective changing of target surface and the intensity losses in the fiber lines are automatically compensated. Meantime, an optical fiber sensor model of correcting static error based on BP artificial neural network (ANN) is set up. Also by using interpolation method and value filtering to process the signals, effectively reduce the influence of random noise and the vibration of the roller bearing. So enhance the accuracy and resolution remarkably. A simulation-software program is compiled with Microsoft Visual Basic 6.0 based on this detective principle. By using it, the detecting signal curve is automatically drawn. And the date of I/O signal for this system is gained in good time.