A semantically constrained neural network for manufacturing diagnosis
本文在缺陷-元原因-根本原因表示法基础上，开发了一种受语义约束的前馈神经网络用于制造过程诊断，将元原因概念与隐藏节点关联，并分析了约束学习行为。
In an earlier work (Ransing et al . 1995), we represented the causal relationship in a defect-metacause-rootcause form. This representation was perceived to be of considerable importance to the research community as well as industry, as it is applicable to any form of manufacturing process. Based on this representation we proposed 'A Semantically Constrained Bayesian Network' for the diagnostic problems (Lewis and Ransing 1997). In this paper, we develop another popular Artificial Intelligence tool, 'Feedforward Neural Network', for such diagnostic problems. The network is constrained to defect-metacause-rootcause topology and it has been shown that metacause concepts can be successfully associated with the hidden nodes. The errors are calculated at both the output layer and the hidden layer. Although the learning process is based on the back-propagation algorithm with a momentum term, the weight changes would occur at a link connecting a node only if at least one of the nodes connected to it in the preceding layer has non-zero activation. The theoretical analysis of such constrained learning is given and it is shown that the network behaviour is acceptable for the diagnostic problems considered.