Intelligent control of an electro-hydraulic rotary drive system

The main purpose of the work presented was to examine whether intelligent control techniques could solve some of the problems presented by hydraulic systems. From an industrial perspective the interest had been in generating methods which were simple to apply with little a priori knowledge and had a low computational burden yet gave high performance. As a result the intelligent control structures chosen were both of an implicit form. Other structures based around the use of an identified fuzzy or neural model have been considered and are the subject of on-going research. However these methods must be parsimonious since the system dynamics are of high order and some methods suffer from an exponential increase in variable parameters with the number of inputs. In this work it has been shown that a simple feedforward neural network structure can provide excellent control which is highly robust. The neural structure and training method proposed by Newton (1993) is relatively quick to converge and simple to apply, and the results obtained demonstrate that only a small 5 neuron structure can control a complex non-linear system.