Analysis of LED Array of RGB LEDs by Artificial Neural Network

• Main Authors: Guo-Yang Wu, 吳國揚
• Other Authors: Hsi-Chao Chen
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/12968455321632161681

碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 100 === High power light-emitting diodes (HP-LEDs) always are applied for energy-saving to replace the traditional light sources. Therefore, the high power LED lighting has been regarded in the next generation lighting. In this study, we design the LED array of 2 × 3 with the best artificial neural network (ANN) training model to find the optimization of the neural network calculus. Then, one array data of experimental result was used as the basic to forecast the chromaticity coordinate values of the other 16 arrays. The chromaticity coordinates of the 17 groups LED array were measured by the integrating sphere based on the color temperature of D65. The coordinates of the chromaticity would be simulated and discussed by the optical simulation software and the Artificial Neural Network with optimization algorithms. It was the smallest average error than all with the neural algorithm of Levenberg-Marquardt (L.M.) model. Using the L.M. model to train the model and get the error percentage for the 16 arrays to compare on the basic array. Anyway, the least and highest error rate are 0.2272% and 2.028% for 14th array, respectively. The simulation results with neural network training are 1.027% better than 2.141% that of Trace-Pro simulation. 94.12% arrays could reach the standard with the error rate under 2%, but only 35.29% arrays could reach the standard with the error rate under 1%.