查詢結果分析
來源資料
相關文獻
- Optimizing a Replication Process in CD-R/DVD±R Manufacturing Using a Hybrid Procedure
- Optimizing Printed Circuit Board Design by Neural Network, Desirability Functions and Particle Swarm Optimization
- 手寫數字辨識模式之建立--結合遺傳演算法與類神經網路
- Form Segmentation and Component Classification for Clinic Document Image Analysis
- 倒傳遞類神經網路在波浪時序列預報之應用
- 達成中央空調舒適度與省能控制之研究
- 應用於洪水演算類神經倒傳遞網路法最適參數推估
- 類神經網路應用於房地產估價之研究
- Experimental Design for the Treatment of a Dyeing/Finishing Wastewater by FBR-Fenton Method
- 運用類神經網路建構臺灣地區農會信用部金融預警系統
頁籤選單縮合
題名 | Optimizing a Replication Process in CD-R/DVD±R Manufacturing Using a Hybrid Procedure=以混合式演算程序最佳化CD-R/DVD±R製程之複製加工 |
---|---|
作者 | 徐志明; 陳凱瀛; Hsu, Chih-ming; Chen, Kai-ying; |
期刊 | 品質學報 |
出版日期 | 20051200 |
卷期 | 12:4 民94.12 |
頁次 | 頁299-312 |
分類號 | 494.55 |
語文 | eng |
關鍵詞 | 複製加工; 倒傳遞類神經網路; 望想函數; 粒子群最佳化; 多重回應值參數設計; Replication; Backpropagation neural network; Desirability function; Particle swarm optimization; Multi-response parameter design; |
中文摘要 | 在CD-R/DVD±R製程中,將電鑄階段所完成之壓模固定於射出成型機,並擠壓熔解之塑膠原料以成型生產大量碟片的過程稱為複製加工。這些碟片的物理特性對於光學可錄式媒體成品之品質性能具有顯著之影響,而複製加工中有許多製程相關參數會同時影響到這些塑膠碟片之品質特性。傳統上,工程師利用試誤法決定這些參數之最佳設定值,導致碟片品質性能不佳以致蒙受很可觀之損失。為了解決在製造碟片的複製加工中所面臨之複雜的多重回應值參數設計問題,本研究提出一個整合倒傳遞類神經網路、望想函數、粒子群最佳化的混合式演算程序。以倒傳遞類神經網路建構輸入控制因子與輸出回應值間的函數關係,然後利用粒子群最佳化程序尋找最佳參數設定組合,其中每個粒子的適合度則以望想函數衡量之。本演算程序實際運用於台灣某家光學可寫入式媒體製造商之複製加工中以驗證其可行性及有效性,並獲致相當好之結果。 |
英文摘要 | In manufacturing CD-R/DVD±R, a replication process is designed to produce plastic discs by stamping the molten plastic on an injection molding machine fitted with a stamper created by an electroforming process. The physical characteristics of discs make a great impact on the performance quality of the finished optical recordable media. Furthermore, various process related parameters involved in replication simultaneously influence these quality characteristics of plastic discs. Traditionally, engineers determined the optimal parameter settings through trial and error, resulting in considerable losses owing to unsatisfactory disc performance. For solving a complicated multi-response parameter design problem in a replication process of manufacturing discs, a hybrid procedure based on backpropagation (BP) neural networks, desirability functions and particle swarm optimization (PSO) is derived and proposed in this study. The BP neural network is employed to model the functional relationship between the input control factors and output responses. Moreover, a PSO process is applied to optimize the control factor settings through the well-trained BP model, where the fitness of each particle is assessed using desirability functions. The proposed procedure is applied at a Taiwanese optical recordable media manufacturer, and the implementation results show its feasibility and effectiveness. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。