Balanceamento de rotores flexíveis sem usar massas de teste

Abstract

The present master s degree thesis presents two flexible-rotor balancing techniques, which do not use trial weights. The goal of this approach is to help in situations found in the industrial context, in which other techniques that use trial weights cannot be applied. A few reasons can be mentioned: the time consumed to stop the machine to install the weights are prohibitive; technical reasons make very difficult the work of installing the trial weights. The first technique can be considered as being a modal one and is based on the superposition of the flexible modes of the rigidly supported rotor and the rigid body modes. The characteristics of the bearings are not needed, which is interesting from the practical point of view, because in many cases those characteristics are not available. The other balancing technique is based on pseudo-random optimization methods. In this research work two methods were explored, namely, the genetic algorithms and artificial neural networks. The basic idea is to obtain the flexible rotor unbalance response, which is then mimicked by using a FEM model in which the unbalance masses and their corresponding angular positions are the design variables in the optimization run. This way, an inverse problem is solved and the masses obtained are installed in previously chosen balancing planes at inverted angular positions. Numerical simulations are presented showing the efficiency and limitations of the methodology developed.