| Summary: | Control chart pattern recognition has become an active area of research since late 1980s. Much progress has been made, in which there are trends to heighten the performance of artificial neural network (ANN)-based control chart pattern recognition schemes through feature-based and wavelet-denoise input representation techniques, and through modular and integrated recognizer designs. There is also a trend to enhance it’s capability for monitoring and diagnosing multivariate process shifts. However, there is a lack of literature providing a critical review on the issues associated to such advances. The purpose of this paper is to highlight research direction, as well as to present a summary of some updated issues in the development of ANN-based control chart pattern recognition schemes as being addressed by the frontiers in this area. The issues highlighted in this paper are highly related to input data and process patterns, input representation, recognizer design and training, and multivariate process monitoring and diagnosis. Such issues could be useful for new researchers as a starting point to facilitate further improvement in this area.
|
|---|