SIMULATION OF FLUID-POWER CIRCUITS USING ARTIFICIAL NETWORK MODELS PART-1 - SELECTION OF COMPONENT MODELS

Consideration is given to the transient analysis of fluid power circui ts based solely upon a variety of artificial networks rather than the conventional mathematical modelling approach. Networks, trained using data from practical tests, are developed for an axial piston pump and filter, a proportional pressure relief valve, a lumped volume and a su bsystem containing a proportional flow control valve driving an axial piston motor with its integral load assembly. A variety of networks ar e considered ranging from the linear-regression type, BARMAX, the self -organizing group method of data handling (GMDH) using nonlinear regre ssion theory to multilayer perceptron (MLP) networks by the back propa gation algorithm. Part 2 (1) then shows how the network models may be linked with the aid of additional steady state network models to predi ct the behaviour of a motor speed control system.