Group technology and cellular manufacturing with artificial neural networks

Group Technology (GT) / Cellular Manufacturing (CM) is a useful way of increasing the productivity in manufacturing high quality products, improving the flexibility of manufacturing systems and decreasing the costs. Cell Formation (CF) is the key step for GT. CF can identify part families and machine groups. Several kinds of methods can be used in CF. Artifial Neural Networks (ANNs) are very suitable for CF and have been widely applied in CF due to their robust and adaptive nature.In the thesis, a review of different kinds of ANNs from the literature which are used in CF, is presented. An application of Self Organizing Map (SOM) and Competitive Neural Network (CNN) within a new methodology for grouping binary and nonbinary (fuzzy) problem sets simultaneously is made. 15 problem sets gathered from the literature are used as binary problem sets and 6 problem sets gathered from the literature are used as nonbinary problem sets. A performance measure which is created by taking the aritmetic average of five different well-known performance measures from the literature is proposed and used to evaluate and compare the cell solutions. Also, the performance measures in the articles the problem sets are taken from, are used once more to evaluate and compare the cell solutions. SOM and CNN results are compared with the results in the literature. In the last part of the application, different numbers of cells are tested to see whether there is a better cell configuration than the article has found. Matlab 7.5 is used to code the neural networks and find the best groupings.