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（ＡＲＴ）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 A neural network-based process design case inference index model is proposed, which is different from most existing case inference systems. This method uses the neural network (ART) to realize the dynamic classification and indexing of instances. The three-level structure of instance-level classification provides the conditions for implementing instance-based reasoning solution based on symbolic processing to pattern recognition based on neural computing. The adaptive and self-learning ability of neural network will reduce the routine maintenance of the system. The advantages of this method lie in the high-speed and efficient retrieval of instances and the simplification of knowledge acquisition