Non-linear Grey-Box Models Applied to DC Motor Identification

Abstract

Grey-box modeling integrates both qualitative (expert-based) and quantitative (measurement data) knowledge. A grey-box model combines a formal structure of the phenomenon with a data-driven, generic model. The model is fitted to data until a complete and precise representation is achieved. In this paper, a grey-model based plant identification is carried out to estimate the parameters of a Pittman GM9413-3 DC motor, about which little of information is available. It is the main source of locomotion of the SCORBOT-ER V plus robot manipulator. With the parametric identification of the DC motor, it is possible to approximate the dynamics of the manipulator. The mathematical model is partially obtained by fitting the values of the internal resistance and inductance of the stator, using a linear regression of the data obtained from the DC test and rotor blocked test, respectively, and assuming that no losses are present due to electromagnetic conversion. The data were acquired using an FPGA-based data acquisition system tailored for the application. Results show that the model is precise, with a fitness higher than 84% and a final prediction error of less than 1%.