Prediction of humidity levels for maximum optical relative transmittance using modified heterodyne detection and GA-BP neural networks

Abstract

Humidity management is of great significance for optimizing the performance of all-solid-state optical systems, while the intricate interplay between humidity and various optical elements’ relative transmittance (RT) is characterized by a nonlinear relationship, complicating the measurement and calculation of optimal humidity levels necessary for achieving maximum RT. In this paper, we present a modified heterodyne detection method to quantitatively measure RT variations across diverse optical materials under humidity fluctuations. The acquired data are trained into a unified mapping model using a genetic algorithm-back propagation (GA-BP) neural network, enabling precise prediction of humidity levels required for maximum RT. Experimental results demonstrate that the GA-BP method achieve a remarkable correlation coefficient (R²) of 0.998 between predicted and actual values, and with the proposed model, the predicted optimal humidity levels for polarizers, conductive films, and windows are 44 %, 21 %, and 15 %, respectively, yielding RT values exceeding 99 %. These results outperform scenarios without prediction or using traditional BP methods. We believe that this research can offer novel insights for optimizing related optical systems to enhance output quality.