Parameter Estimation of a Rotor-bearing System Using Genetic Algorithm and Simulated Annealing

This work presents a technique that allows the calibration of mathematical models in rotating systems. The method is based on the application of a hybrid meta-heuristic search, which employs genetic algorithm and simulated annealing. The capabilities of the method are evaluated by means of experimental results obtained in the test-rig of Politecnico di Milano. The experimental set up consists of a rotor, supported by four elliptical journal bearings on a flexible foundation. This rotor is excited by unbalance forces. A finite element model, with Timoshenko beams, is used to model the system, considering the gyroscopic effect. A linear hydrodynamic force model is considered for the bearings. So, the estimated parameters of the system are the damping and stiffness coefficients of the bearings and the module and phase of the unbalance applied to the system. The objective function is based on the difference of experimental results and simulated results, considering the parameters to be fitted. This objective function has to be minimized in order to fit these parameters. Once the parameters (stiffness and damping coefficients of each bearing and unbalance module and phase) are estimated, it is possible to calibrate the mathematical model, and then to obtain reliable responses for the physical system studied. The method presented here is also suitable to the rotating machine design area as it presents a relatively simple methodology for the updating and validation of models of machines and structures.