Integration of a fuzzy neural network and multi-objective genetic algorithm for optimisation of BLU light guide plate injection moulding parameters

Backlight module unit (BLU) is a key component of LCD, and thinning of the BLU light guide plate is the focus of development. Most light guide plates are manufactured using the injection moulding process. The experimental design in this study is the most common research approach applied in the setting of light guide plate parameters. However, the parameter interaction and system non-linearity can easily result in prediction errors of parameters. Meanwhile, since the optimal moulding conditions have not been achieved, the best effects of display illumination and brightness cannot be achieved. Hence, this study first applied the grey relational sorting method to determine the major control factors influencing the BLU light guide plate, and then built the approximate model of a fuzzy neural network, according to experimental planning samples, before using the multi-objective genetic algorithm to determine the optimal manufacturing parameter design. The results suggested that the process capability index C-pk value of the light guide plate X-axis size is improved from 0.43 to 1.56, and has stabilised product quality and enhanced process capability. This study proved that the prediction model of process optimisation, as obtained by using this research method, is effective.