適應性網路模糊推論系統於深引伸成形極限最佳化料片之研究(II)
學年 97
學期 1
出版(發表)日期 2009-01-01
作品名稱 適應性網路模糊推論系統於深引伸成形極限最佳化料片之研究(II)
作品名稱(其他語言) A Study of Optimum Blank of the Forming Limit in the Deep Drawing Process by Using Adaptive Network Fuzzy Inference System(II)
著者 李經綸; 葉豐輝
單位 淡江大學機械與機電工程學系
描述 計畫編號:NSC98-2221-E032-006 研究期間:200908~201007 研究經費:515,000
委託單位 行政院國家科學委員會
摘要 本研究計畫係結合動顯函有限元素分析與適應性網路模糊推論系統,以前向模式預測深引伸成形後之杯高,並發展一套適應性網路模糊推論系統之逆向模式,透過此逆向模式來預測深引伸成形達固定杯高之最佳化料片輪廓外形。透過適應性網路模糊推論系統之複合式學習演算法,可有效的建立知識規則庫與最佳化歸屬函數。在逆向模式之預成形設計中,則以節點號碼(N)、初始料片節點座標(XY),及變形後杯高值(H)為輸入資料,逆解得到最佳化料片輪廓外形之節點座標。然而,於一般成形過程所建構之知識規則庫是被允許的,但當達到成形極限,甚至超過成形極限時之訓練參數,即不適用於建構知識規則庫,否則在逆向模式中可能導致不合理之結果,故需反覆進行訓練,以建構一最佳化知識規則庫。本計畫於第一年度已設計一組圓杯深引伸成形模具,並建立其成形極限之異向性材料的知識規則庫。第二年度擬再設計一組非軸對稱之方杯深引伸成形模具,將模具幾何非線性之影響因素加入,以修正先前建立好之知識規則庫,建構出一泛用型深引伸成形極限分析之知識規則庫。最後將動顯函有限元素分析及ANFIS數值模擬所得之沖頭負荷與衝程關係、杯高分佈、應力與應變分佈,及成形極限與實驗結果作比較,以驗證本計畫所建構知識規則庫之實用性。 This project is to combine the dynamic-explicit FEM and Adaptive Network Fuzzy Inference System (ANFIS) to predict the height of the cup after the deep drawing, by the forward model. Meanwhile, an ANFIS inverse model will be developed to predict the profile of the optimum blank, when a desired target height of the cup is reached, in the deep drawing process. From the hybrid-learning algorithm, it can efficiently construct the rule database and optimal distribution of the membership function. Whereas, using the node number, the node coordinates of the initial blank, and the deformed height of the cup to be the input parameters, the node coordinates of the profile in the optimum blank can be solved inversely in the inverse model. However, the rule database established from the common drawing process is allowed. While the training parameters reach or even exceed the forming limit, it is not appropriate to be used to establish the rule database. Otherwise, it will get an unreasonable result in the inverse model. Hence, it should be trained repeatedly to establish the optimum rule database. In this project, we designed a set of tool for the cylindrical cup deep drawing, and established the rule database for the anisotropic material on the forming limit in the first year. In the second year, we will design a set of square cup deep drawing tool, combine with the factor of geometric nonlinear effect, to modify the rule database established before, and then extend the rule database in the general deep drawing process. Finally, we will compare the results of the dynamic-explicit FEM and ANFIS numerical simulation in the relation of punch load and punch stroke, the distribution of the cup height, the distribution of the stress and the strain, and the forming limit with the experimental results. The comparison results will be used to verify the practical applications of the knowledge rule database established in this project.
關鍵字 動顯函有限元素; 適應性網路模糊推論系統; 深引伸; 最佳化; 成形極限; Dynamic-explicit FEM; ANFIS; Deep drawing; Optimum; Forming limit
語言
相關連結

機構典藏連結 ( http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/47063 )

機構典藏連結