http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
허정준,김광섭,구자항 대한설비관리학회 1997 대한설비관리학회지 Vol.2 No.2
This research is focused on developing a methodology for detecting the time of model change in the pocess and on designing a procedure for identifying the change based on neuro-modeling. The proposed detection procedure uses a fixed series of one-step prediction errors, which are obtained from a pre-determined neural network model. The model identification procedure is based on neuro-modeling. Kohonen's Self-Organizing Map (SOM) neural network and Radial Basis Function (RBF) neural network are used for modeling of a dynamic manufacturing process which is considered as a quasi-stationary time series. The proposed SOM and RBF neural networks attempt to decompose a picewise stationary series into a set of stationary segments and predict each sub-series thereby. Three different approaches are compared for model identification: the first approach is based on the feedforward network with the back-propagation error rule; the second one is based on the Bayesian discriminant function for minimizing the average probability of error; and the third one is based on the minimum distance measure for mininuzing the Euclidean distance between the input feature vector and feature vectors of the each sub-models. The proposed approaches are verified throught simulations. The performance of the proposed approaches are efficient and the suggested methods can readily employed for real manufacturing applications.