A Self-Organizing Neural Network Approach to Data-Based Modelling of Fluid Power Systems Dynamics Using the GMDH Algorithm
Y Xue, J. Watton · Proceedings of the Institution of Mechanical Engineers Part I Journal of Systems and Control Engineering · 1995
The group method of data handling (GMDH) is used to identify the non-linear transient behaviour of fluid power elements based upon input/output observations. It is shown how a variety of multi-layered perceptron-type feedforward neural networks evolve for each application due to the heuristic self-organization of the approach. Single-input-single-output systems are considered and a modified algorithm is proposed and developed for multi-input-single-output systems. The technique is applied to a pressure relief valve and a variety of servo-valve-controlled motor systems.