Determination of manufacturing throughput for mounting machine by using artificial neural network / Nornita Abdul Rahman

This thesis presents the application of artificial neural network to determine the manufacturing throughput for a semi-conductor machine. Two types of neural networks have been used, i.e. Back propagation and radial basis function network. Both models developed have three layers i.e. input layer, hidden-layer and output layer. To determine the manufacturing throughput, the system behavior was studied based on the machine downtime report. For both networks, the same sets of data have been used in training and testing process as the data were taken from a monthly downtime report of production from a semiconductor company for a year. Tests were carried out and the results were compared on the basis of learning rate, momentum and number of hidden nodes. From these results, it was shown that ANN can be used for determining manufacturing throughput. The radial Basis Function network was more accurate compared to the back-propagation network.