オペレーションプランの自律的獲得に関する研究 : GAに基づく機械学習によるアプローチ

概要

論文の詳細を見る
This paper presents an attempt to automatically and flexibly solve the sequencing problem of machining operations that enable to produce a part under machining constraints . To realize an flexible planning mechanism, two approaches are performed. In these approaches, operation planning problems are modeled as combinatorial optimizaion problems . First, genetic algorithms (GAs) are applied directly. To implement the GAs, a string is encoded as an order of removing operations. Thus, a process that produces a finished part from a raw material by removing the machinable volumes is represented as a sequence of the numbers of machinable volumes. As the second approach, the classifier system (CS) is applied to this problem to obtain the further advantages. In this approach, a classifier rule corresponds to a production rule that instructs the next removal volume for a current machining state. As a result of the learning, a near optimal process plan is represented as a chain of classifiers with higher strength values . Moreover, a construction of synthesized planning systems based on GA-approach and CS-approach is also discussed. On the basis of the proposed approaches, a genetic based operation planning simulator is constructed and some numerical experiments are carried out.
公益社団法人精密工学会の論文
1998-05-05