Instead of the establishing the mathematical hydraulic system models from the physical laws as usual with the problems of complex modelling process, low reliability and practicality caused by large uncertainties, a novel modelling method for a highly nonlinear system of hydraulic excavator is presented. Based on the data collected in the excavator’s arms driving experiments, a data based excavator dynamic model using Simplified Refined Instrumental Variable (SRIV) identification and estimation algorithms is established. The validity of the proposed data based model is indirectly demonstrated by the performance of computer simulation and the real machine motion control experiments.