Identification and propagation of parametric uncertainty of a hydrostatic drive train model

This paper shows an approach to identify the uncertainty of the parameters in a lumped parameter model from experimental data. This approach is applied to a hydrostatic drivetrain setup in which the behavior of both the hydraulic pump and motor vary as the temperature of the oil changes. The model contains a large amount of parameters, describing the dependency between the speed, pressure and stroke of pump and hydromotor and their efficiency. To decide which parameters to retain for the uncertainty identification, an identifiability study is performed followed by a study of the residuals. The parameter uncertainty identification procedure estimates the parameter standard deviations using a maximum likelihood procedure between the calculated distribution of the model output and the observed residuals. The resulting parameter uncertainty is evaluated in order to propose a design improvement to lower system uncertainty. Also the potential use of the uncertainty model for robust and stochastic control is discussed.