排程在生產過程中扮演極為重要之角色。然而﹐隨著生產系統日益複雜﹐
大部份的排程問題已證明為NP-Complete。傳統的作業研究工具已無法有

效率地解決上述問題﹐因此本研究針對動態排程問題提出一套結合傳統分

析工具與現代人工智慧方法具有學習能力之智慧型排程系統。 SAOSS 採

用以特徵值為導向之彈性排程策略。亦即在不同的系統狀態下使用當時最

適切之派工法則 (Dispatching rule) 。基於以上特性，第一章首先闡

述有關排程之基本概念;製造過程中所發生之排程問題 SAOSS 必需具有學

習動態排程知識、組織相關資訊的能力。因此，SAOSS採用一套連續型二

分法算則 (A continuous ID3 Algorithm)，構建一由 ADALINES 所組成

之類神精網路系統。網路節點間之加重權數則隱含了動態排程所需之法則

。由此方法產生之排程法則十分精簡， SAOSS 的運作效率因而大為提昇

。以及本文中使用之研究方法。第二章則回顧以往排程之方法。第三章則

介紹智慧型排程系統之基本架構。第四章則陳述此智慧型排程系統的發展

步驟。第五章則應用一彈性製造單元作評估。第六章結論與未來發展方向

。經由實証，顯示 SAOSS 能處理繁雜之動態排程問題。此外，由 CID3

算則所產生之生產排程法則亦較其它方法簡潔。

Scheduling, as part of production planning and control, plays

an inportant role in the entire manufacturing process. Most

sche duling problems have been proven to be NP-complete which

degrade s the performance of a conventional OR techniques.

Hence, a new approach which can deal with sophisticated,

especially dynamic s scheduling problems is strongly desired.

In this study, a system attributes oriented knowledge based s

cheduling system (SAOSS) with inductive learning capability is

i ntroduced. SAOSS takes a multi-algorithm paradigm so that it

is able to tackle a variety of scheduling problems. In other

words, different strategies are are inferred by SAOSS with

respect to t he scheduling conditions which makes SAOSS more

intelligent, fle xible, and suitable than others in tackling

complicated, dynamic scheduling problems. The embedded

knowledge acquisition mechanism enables SAOSS to acquire

scheduling heuristics from both expertise's experience a nd

experiments through inductive learning. An efficient inductiv e

learning method, a continuous ID3 (CID3) algorithm, induces de

cision rules for scheduling by converting corresponding

decision trees into hidden layers of a self-generated neural

network. Con nection weights between hidden units imply the

scheduling heuris tics which are interpreted into scheduling

rules later. An FMC scheduling problem is given for

illustration and justification.