A POLYNOMIAL NETWORK MODELING APPROACH TO A CLASS OF LARGE-SCALE HYDRAULIC SYSTEMS

This paper investigates the application of a polynomial abductive netw ork modeling technique to a class of hydraulic actuation systems used in heavy-duty mobile machines. Previous techniques to model and simula te heavy-duty hydraulic functions include using a combination of analy tical, numerical and steady state solutions in a partitioned hierarchi cal manner [N. Sepehri, P. D. Lawrence and F. Sassani, Proc. IASTED In t. Conf. on Modeling and Simulation, pp. 170-174(1991)] or, a transmis sion line modeling approach to facilitate parallel processing [J. D. B urton, K. A. Edge and C. R. Burrows, ASME J. Dynamic Systems, Measurem ent and Control 116, 137-145 (1994)]. Using the results from the appli cation of these methods to an excavator-based machine, databases are d eveloped which are then used to construct polynomial abductive network models in order to replace the conventional iterative derivations for fluid flow distribution among the actuators. The networks are generat ed using a mechanism based on a hybrid learning logic called ''abducto ry induction'' [G. J. Montgomery and K. C. Drake, Proc. SPIE(-)-Int. S ec. Optical Engng, pp.56-64 (1990)]. The performance of the polynomial abductive network model is compared to the previously established mod el. It is shown that the computation time is reduced considerably, yet the simulation results show a sufficient degree of predictive accurac y. Additionally, the new modeling scheme eliminates the need for expli cit derivations of the main valve orifice areas, and therefore facilit ates the use of experimental observations in developing simulation mod els.