Please use this identifier to cite or link to this item: `http://hdl.handle.net/11455/2753`
 標題: 類神經網路式最佳參數設計法則Neural Network Based Optimal Design Method 作者: 吳嘉哲CHE, WU CHIA 關鍵字: 共軛梯度法;類神經網路式最佳參數設計法則;直交表 出版社: 機械工程學系 摘要: 在本論文中，我們提出一套以類神經網路結合田口實驗法之最佳參數設計法則。在此一新的最佳參數設計法中，直交表被用來做為建立類神經網路訓練所需輸入、輸出樣本之工具，再由類神經網路訓練來模擬可控因子與誤差因子間之數學模型，最後再利用直接反推法及數值運算法(共軛梯度法)求取參數最佳值。我們先以非線性函數之電腦模擬結果來驗證此方法確能有效求取參數最佳值。然後再以直流無刷線性馬達精密定位控制作為實際應用案例。實際實驗顯示不論是單軸追蹤或雙軸循圓控制，本研究所提出之類神經網路式最佳參數設計法則皆可在不知道系統數學模式下得到最佳之參數值。Neural network based optimal design method, a new methods that combines the neural network and the Taguchi techniques for better improvement of engineering design is proposed in this thesis. In this new parameter design method, orthogonal array, which requires partial experiments is executed first. Each row in the orthogonal array along with its relative responses form a set of training pattern to the neural network. Implicit system model that allows precise prediction and easy optimizing is constructed in terms of a multilayer feedforward neural network. With the neural network based system model, optimal parameters to achieve desired output can be obtained through direct inverse or numerical methods such as steepest descent and conjugate gradient. Computer simulations show that the proposed method can easily find the extreme points of an unknown complex nonlinear problem. In real application, the proposed method is used to tune the optimal PID parameters for precise positioning and circular tracking of a DC brushless linear motor system. Both computer simulation and experimental results illustrate that the proposed algorithm performs much better than the conventional Taguchi approach. URI: http://hdl.handle.net/11455/2753 Appears in Collections: 機械工程學系所

TAIR Related Article