Statistical process control (SPC) is a method for improving the quality of products. Control charting plays a most important role in SPC. SPC control charts arc used for monitoring and detecting unnatural process behaviour. Unnatural patterns in control charts indicate unnatural causes for variations. Control chart pattern recognition is therefore important in SPC. Past research shows that although certain types of charts, such as the CUSUM chart, might have powerful detection ability, they lack robustness and do not function automatically. In recent years, neural network techniques have been applied to automatic pattern recognition. Spiking Neural Networks (SNNs) belong to the third generation of artificial neural networks, with spiking neurons as processing elements. In SNNs, time is an important feature for information representation and processing. This thesis proposes the application of SNN techniques to control chart pattern recognition. It is designed to present an analysis of the existing learning algorithms of SNN for pattern recognition and to explain how and why spiking neurons have more computational power in comparison to the previous generation of neural networks. This thesis focuses on the architecture and the learning procedure of the network. Four new learning algorithms arc presented with their specific architecture: Spiking Learning Vector Quantisation (S-LVQ), Enhanced-Spiking Learning Vector Quantisation (NS-LVQ), S-LVQ with Bees and NS-LVQ with Bees. The latter two algorithms employ a new intelligent swarm-based optimisation called the Bees Algorithm to optimise the LVQ pattern recognition networks. Overall, the aim of the research is to develop a simple architecture for the proposed network as well as to develop a network that is efficient for application to control chart pattern recognition. Experiments show that the proposed architecture and the learning procedure give high pattern recognition accuracies.
展开▼
