| Summary: | 碩士 === 元智大學 === 機械工程學系 === 92 === This thesis is on optical efficiency optimization for a direct type backlight of the liquid crystal display. The goal is to get the greatest uniformity in a direct type backlight, while the brightness is maintained in a satisfactory level. Both uniformity and brightness are implicit functions that have to be evaluated by optical simulation software Speos. This research will adjust the geometric dimensions, which are discrete design variables, to get the best optical efficiency.
The Sequential Neural Network Approximation Method (the SNA method) is used in this research. In the SNA method, two back-propagation neural network are trained to simulate the rough maps of the feasible domain and the objective function of this optimization problem using a few representative training data. A search algorithm then searches for the “optimal point” in the feasible domain and the objective function simulated by the neural network. This new design point is simulated by the optical simulation software to check its true objective values and whether it is feasible. This new information is then added to the training set and the neural network is trained again. Then we search for the “optimal point” in this new approximated feasible domain again. This process continues in an iterative manner until the approximate model insists the same “optimal point” in consecutive iterations.
In this thesis, a two-variable example is used to illustrate the process of SNA. A four-variable optical efficiency optimization problem for a direct type backlight of the liquid crystal display is then solved using the SNA method. In both examples, the number of optical simulations required is greatly reduced.
|
|---|
