| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 88 === This thesis studies the superiority of lighting quality evaluated by using artificial neural network technique. For evaluating the whole lighting quality by artificial neural network systematically, the following four factors are included simultaneously, thus average illumination, uniformity, unit power density consumed by luminaires and direct glare. We evaluate lighting quality affected by different designs of luminaires and their layout. And an teaching environment, such as classroom, is used to simulate and compare the overall lighting quality.
First, the research is simulating luminaires layout and luminous parameters, which installed using different arrangements, to obtain the relative parameters to be evaluated by different commercial softwares, Lumen-Micro and Glare-Index. Then the data are graded, according to National Standards, inputting to artificial neural network. The score inferred through learning and recalling process becomes the major index for evaluating lighting quality. Finally, the result is compared with related researches in order to certify the accuracy and adequately of applying artificial neural network to evaluate lighting quality.
Experiment shows the inferred error RMS is about 2% pproximately, and its results meet related reports. It is evidently that one could evaluate the light quality systematically on the four parameters, average illumination, uniformity, direct glare and energy consumed by using artificial neural network technique. The most valuable contribution of this research is to evaluate lighting quality effectively and quantitatively among the different lighting designs. It is aimed to promote the developments and researches in lighting engineering.
|
|---|
