應用類神經網路於工件辨識

Machined Part Recognition Using Artificial Neural Network Approach

姚景凱(Ching-Kai Yau)

指導教授 ： 鄭春生

元智大學/工程學院/工業工程與管理學系/碩士(1995年)

摘要

由於電腦輔助設計(CAD)輸出工件資料中，缺乏電腦輔助製程規劃(
CAPP)系統做製程決 策所需之加工資訊，以至於現有CAPP系統均無法直接
讀取CAD資料，尚需人為介入處理二者 資料的轉換工作。有關前述CAD/
CAPP整合問題的研究，主要朝向兩大方向發展：(1)特徵為 基設計(2)特
徵辨識。

本研究採用類神經網路來分辨工件具有的切削特徵。在輸入節點方面，本
研究採用特 特徵辨識中的語法型態辨識法作為單一特徵平面的辨識法則
，利用單一特徵平面作為類神 經網路輸入節點的編碼依據。經由類神經
網路連結權數的計算，依序辨認出工件內部所包 含的三維特徵與重疊特
徵。

本研究使用Borland C++ 3.1 做為系統軟體的構建工具，並以ACIS實體模
型系統構建 工件模型，以其邊界表示法輸出之工件資料為系統的輸入，
本研究並以垂直交疊、平行交 疊及混合交疊的模型測試，驗證所提演算
法的實用性及可行性。

關鍵字

類神經網路 ； 電腦輔助設計 ； 電腦輔助製造 ； 電腦輔助製程規劃 ； 特徵辨認

並列摘要

The machined part recognition problem has been recognized as
one of the major bottleneck in developing a fully automated
process planning system. Most previous works employed a
template approach which recognizes a feature from a set of
faces which match the template features. This approach is
inherently limited because machined parts cannot be
exhaustively described by the limited number of templates.

This research extended the process requirement concept proposed
to recognize machined part. Based on this concept, a
machined feature is present by a set of faces which can be
machined by a cutter moving along a set of cutter paths.
Therefore, one machined part can dispel into several feature
planars, combine these planars which give some information for
3D feature detection. An input format has been developed
which includes the face descriptions and face-face
relationships. An algorithm for recognition using neural
network techniques has been developed. The net architecture is
described, and a few examples are presented which highlight the
strengths and weaknesses of the recognition algorithm.

並列關鍵字

artificial neural network ； computer aided design ； computer aided manufacturing ； computer aided process planning ； feature recognition