Tsung-Hung Lin

Applying ANN/GA algorithm to optimize the high fill-factor microlens array fabrication using UV proximity printing process

An artificial neural network and genetic algorithm were used to achieve a high quality microlens array fabrication using a UV proximity printing process in this study. The UV proximity printing process can precisely control the geometric profile of the microlens array in the fabrication process without thermal reflow. The major objective in using the robust design is to reduce the variations in the focal length of the microlens array, allowing improved focus and enhanced illumination brightness. The artificial neural network was used first to characterize the nonlinear relationship between the manufacturing parameters and the properties derived from experimental data. The manufacturing parameters which affect microlens array uniformity include: (1) geometrical ratio, (2) printing gap, (3) spin coating revolution speed, (4) exposure time and (5) developing time. It is very important to control these parameters to decrease the sensitivity to noise. The L18 orthogonal array was used as the learning data for the artificial neural network to construct a system model that could predict the focal length for arbitrary setting of parameters. Then, the genetic algorithm was applied to obtain the robust setting of parameters. The results showed that the microlens array quality could be significantly improved in comparison with the original design.

Fabrication of a microlens array electroformed mold with low roughness and high hardness

A systematic approach to achieve a LRHH (low roughness and high hardness) electroforming process is developed in this study. Because electroformed molds with low roughness and high hardness are required for microlens array fabrication using the LIGA-like process, the invented Ni?Co alloy-pressurized electroforming process is used to fabricate a metallic micro-mold for microlens array molding. The electrolyte parameters such as Co content, current density, brightener content, pH value and temperature will be examined with respect to roughness and hardness. An artificial neural network (ANN) is used to construct a system model that accurately predicts the responses for arbitrary parameter settings. A genetic algorithm (GA) is applied to minimize the surface roughness and improve the micro-mold hardness. The results show that the micro-mold surface roughness and hardness could be significantly improved using the ANN/GA approach. A LRHH electroforming process is carried out using parameter design to improve the surface morphology and increase the service life of the micro-mold during the forming process.