实例推理中基于ART算法的动态记忆模型的研究

提出了一种基于神经网络的工艺设计实例推理索引模型，与现存大多数实例推理系统不同，该方法用神经网络实现实例的动态分类和索引。实例层次分类的３层结构，为实现基于符号处理的实例推理求解模式向基于神经计算的模式识别求解模式映射提供了条件。神经网络的自适应，自学习能力将减少系统的日常维护工作，该方法的优点在于实例的高速，有效检索，知识获取的简化等。

Based on Art for Case Based Rasoning

A new approach to the neural indexing models for case-based process planning of turning parts is presented. Unlike most of the existing CBR system, this approach makes use of neural networks(ART) to realize the dynamic classification and to index the case. Three levels for hierarchical classification of cases are presented so as to image the case-based reasoning based on symbol processing with the pattern recognition based on neural calculation. This approach makes the neural networks all self-organized and therefore may reduce periodic human maintenance of the system. Advantages of this approach are the more effective retrieval of the case, the easier knowledge acquisition and the more robust retriever based on neural networks.