Prism-pattern design of an LCD light guide plate using a neural-network optical model

Abstract

This paper proposes a neural-network optical model for a backlight module of a liquid crystal display (LCD) to expedite the design of the light-scattering prism-pattern of its light guide plate (LGP). First, the prism surface of a light guide plate is divided into several equal regions. Then the neural-network optical model is implemented using a back-propagation neural network to establish the relationship between the distribution density of the prism pattern and the exiting-light luminance of the LGP at each region. The input–output patterns for the neural network training and verification are generated using orthogonal arrays and ASAP simulation. Then a for-loop computational algorithm is executed to search an approximately optimal distribution density of the prism pattern using the neural-network optical model such that high luminance uniformity is achieved. It is demonstrated by the case study of a 13 in. LCD backlight module that luminance uniformity could reach 93.1%. Thus it can be concluded that the developed neural-network optical model effectively expedites the LGP prism-pattern design.