應用類神經網路預測LED元件之螢光粉比例

A Study of Artificial Neural Network on the Phosphor Ratio Prediction of LED Components

鄭博仁(Po-Jen Cheng)

指導教授 ： 黃博滄

中原大學/電機資訊學院/工業與系統工程研究所/碩士(2014年)

https://doi.org/10.6840/cycu201400503

摘要

LED製造流程一般分為上游磊晶、中游晶粒、下游封裝，而LED封裝廠為了取得符合客戶規格之CIE色座標，所以在點膠製程需要進行螢光粉試比例作業，來決定螢光粉比例值。本研究收集點膠製程相關資料，並將相關資料彙整統計分析，運用倒傳遞類神經網路建置PLCC LED封裝廠的核心技術之「螢光粉比例預測系統」，讓工作人員除了方便掌握各項資訊外，並運用這些歷史資料預先模擬適當的比例處方，不僅可以縮短螢光粉試比例所花費的時間也可以提升產出週期、節省人力、時間及相關資源成本，並可減少作業員因為反覆抄寫錯誤而導致產品良率降低。由實驗結果顯示，螢光粉試比例次數可以大幅度由每批平均4.5次降為1次，平均每批可節省78%試比例所花費的時間。而未來期望能藉由本研究建置之「試比例預測系統」，更進一步擴充到所有的LED封裝類產品，由此改善來提升產業之競爭力，進而使LED照明燈具可以加速平民化，落實節能減碳及無汙染照明之效益，降低溫室效應對生態之影響。

關鍵字

LED、LED封裝、PLCC LED、類神經網路、倒傳遞類神經網路

並列摘要

The production flow of LED can generally be separated to upstream epitaxy, midstream chip, and downstream packaging. In order to produce the right CIE coordinates for the needs of customers, LED packaging manufacturers need to test different phosphor to find out the best ratio of during the dispensing process.
This study collects data from dispensing production process. A statistical analysis is applied to distinguish the important factors that would influence the ratio of phosphor. Finally, the BPN is deployed to construct the PRP system for LED components. This system not only helps the workers to easily handle the information, the capability of using historic data to simulate the most appropriate ratio also allows shortening the times spent on the tests. It also improves production cycle, man power, saves time and other related resources, and to avoid mistakes from human type error data which lead to low yield.
The testing result of the system indicates that the average number of tests has been done from 4.5 to 1, saving 78% amount of time per lot compared to current situation. Through this study of PRP system for LED components, one expects to further apply the result to other packaging LEDs to enhance the competency of the industry and speed up the process of affordable LED lighting fixtures to everyone, that truly helps reducing our carton footprint, realize the benefits of zero-pollutant lighting, and bring down the greenhouse effect that has done to our eco planet.

並列關鍵字

LED package ； Super PCNeuron ； Back-Propagation Network ； Artificial Neural Network ； PLCC LED ； LED

參考文獻

35. 黃瑩發(2011)，適用於肢體障礙者之電腦操作輔具，國立成功大學。

32. 張英隆(2012)，兩岸LED照明產業與應用市場分析，淡江大學。

36. 楊宗勳、許梓恂、陳建成、陳正岳(2005)，白光LED的色彩特性調制，光學工程，(90)，76-82。

46. 羅梅君、安振基(2012)，照明用白光LED定量螢光光譜測試與白光色彩結構在LED封裝工業上的應用，中華印刷科技年報，396-404。

47. 台灣LED產業現況與發展建議-國家政策研究基金會(2013)。