An artificial neural network for applications in automated industrial systems

The goal of the system described is to learn successfully the operational routines of a gantry crane. The intention is to explore the possibilities of devising a neural-network-based system which is capable of transporting the load as quickly as possible while keeping the angle of swing under control. The effects of variations of the learning parameters on the learning process of a multilayered perceptron are explored. Some observations which have lead to the empirical optimization of the learning process are presented. The empirical technique proposed is founded upon a self-adaptive type of algorithm. The proposed approach essentially amounts to an added capability of the network to capitalize on its experience while still accomplishing the task of learning.<<ETX>>