- 作者: 王培仁; 吳宗明
- 作者服務機構: 清華大學動力機械工程學系
- 中文摘要: 本文採用建立系統鑑別之模式方式,針對塑膠押出系統之溫度控制系統參數,進行自我調整參數之最佳化研究。對於押出機料管溫度之動態乃以神經網路學習理論進行探討,期能進行控制參數之最佳化調整;神經網路學習法則乃採逆傳遞法則原理,並進行神經元個數及網路層數之篩選及比較。經由實驗驗證之神經網路模式則疊代於傳統之比例一積分一微分控制法則中以求取控制響應之最佳化;研究結果顯示,無論理論或經驗式之最佳化實驗結果皆可節省調整時間及改善控制響應。
- 英文摘要: Extrusion has been one of the most important processes for making plastics products for years Inthis paper, a self-tuning temperature control system, based upon a system identification scheme for tuningcontrol parameters, is investigated with regard to the plastics extrusion process For tuning of controlparameters, neural networks has been employed in modeling the temperature dynamics of the barrel onextrusion in order to further optimize the process control. Concerning the learning of the networks, aback-propagation (BP) algorithm has been chosen for updating of the weightings of the neural networks.Initially, comparisons on the selection of different multi-layer networks are illustrated with experimentalverifications. After careful study, an on-line identification scheme based upon the model has been testedon conventional PID type control method. The control gains for the PID algorithm have been on-lineself-tuned by using an iterative numerical scheme for performance optimization Both theoretical andempirical formulations have been employed for optimization of the control gains Experimental resultsbased upon this self-tuning method show significant improvement in the optimization of control perfor-mance.
- 中文關鍵字: temperature control; self-tuning; neural networks
- 英文關鍵字: --