Balanceamento global de eixos através do uso de modelos numéricos e técnicas de otimização

Abstract

Rotating machines are used in numerous applications such as in oil, paper, power generation industries, among others. The usage of these machines, that are more and more compact, in harsh environments, in critical positions and especially in very high rotations demand flawless performance. In the manufacture and assembly it is impossible to obtain defect-free systems and the most common of these defects is the residual unbalance, which can cause loss of efficiency and/or premature damage to the machine. There are several methods to perform the balancing of rotors. The most known ones use test masses and requires the machine to be stopped and started several times. To avoid the various cycles of starting and stopping the rotor, required in conventional techniques, in this work it is implemented a comprehensive methodology for identifying the balancing from known responses, obtained experimentally, and from responses of a numerical model of the rotor. This methodology is based on a reverse procedure of identification, where the numerical model of the system and the measured response are known. A finite elements code was used for the dynamic analysis of the numerical model. With this, using the modal parameters, one can find the response to different excitations. To find the unbalance of the physical model, repetitive calculations of the error between the numerical and experimental responses of the system are performed, varying the magnitude and position of the unbalance mass. This process is carried out by an optimization routine which was added to the numerical rotor model, allowing the insertion of the experimentally obtained data, and set the identification procedure. The existing finite element code to build the numerical model was programmed on the Fortran platform, to which was added the balancing routine. To test the code and the proposed methodology, an experimental-numerical example was presented. Studies of convergence and local minima related to the work are discussed.