臺灣博碩士論文加值系統

本文之目的在於建立一以氧化鋁油石為研拋工具之模具研拋製程規劃
系統, 此研拋製程規劃系統之輸入為放電加工後之SKD61 模具最初表面粗
糙度及經研拋後之模具最終表面粗度, 輸出則為研拋過程中所需採用之研
拋道次及各道次加工所相應之加工參數組合. 首先考慮影響模具研拋
製程之重要參數, 藉由實驗分析找出各種粒度的油石適合加工之表面粗糙
度範圍及研拋加工參數的最佳區域, 並以此建立模糊邏輯控制規則. 再導
入兼具類 神經網路之學習能力與模糊邏輯之推論功能的類神經模糊決策
系統, 利用類神經網路之自動調節能力可迅速且正確的建立各研拋參數在
模糊界面之歸屬函數, 並由模糊推論機構研判出原始, 最終表面粗糙度及
其相應之研拋參數組合. 最後則對研拋製程作一合理的加工製程規劃,
以使模具之最初表面粗糙度可經由一或多個道次的加工而逐步達到所要求
之最終表面粗操度, 以所需加工時間最短之原則下, 對各組之道次組合作
比較, 自動選取出最佳加工道次安排. 循此架構建立起業界普遍使用之模
具材料SKD61 的油石研磨拋光製程歸劃系統.

The purpose of this thesisis to develop a process planning
systemwith Al2O3 abrasive stones for die and mold polishing
system to replacelabor intensive polishing process.The inputs of
this system are the EDMedsurface roughness(initial surface
roughness)of die &mold and the polishedsurface roughness(final
surface roughness)of die &mold.The outputs of thissystem are the
polishing step set and the reasonable polishing parameter sets
corresponding to each polishing step. The important
variables affecting the polishing process are taken intoaccount
in the experiment to find the suitable polishing range of
surfaceroughness for the abrasive stone with various grain size.
Moreover, the experiment results are used as a reference for
designing the fuzzy logic rules. Then the neuro-fuzzy technique
combining the learning ability of neural network and the
interring ability of fuzzy logic system is used hereto construct
the exact membership function of the fuzzy interface and
decidethe reasonable polishing parameter sets of each polishing
step. Finally, process planning is analyzed for the polishing
process in order that the initial surface roughness of die &
mold can be polished to final surface roughness of die & mold
through one or several consecutive polishingsteps. By comparing
the total polishing time the various polishing step set takes,
the optimal polishing step set will be selected automatically
basedon the shortest total polishing time. Thus, the process
planning system forthe pair of SKD61 and AL2O3 is established.