| Large rotor vibrations caused by rotor mass imbalance distributions are a major source of maintenance problems in high-speed rotating machinery. Among various balancing techniques, the influence coefficient (IC) technique is an effective one that has been widely utilized in industry. Within the IC balancing framework, the least squares (LS) method and the min-max method are two important balancing methods which have different optimization objectives. The objective of this dissertation is the development of new approaches to balancing. When there are significant measurement errors or other uncertainties in the data collected, the conventional balancing methods, the LS method and the min-max method often do not produce a good result. In cases such as these, often many balance runs are required.; A new optimum stochastic balancing method is proposed for effective balancing of rotor systems with uncertainties of known distribution. Optimum stochastic balancing, as a robust balancing scheme, aims to minimize the maximum mean square value of the residual vibration by modeling the uncertainties. Several simulations using industrial rotors show that, with uncertainties, the optimum stochastic balancing method outperforms other existing balancing methods. For rotor systems with data uncertainties, the rotor vibrations are supposed to be balanced under acceptable levels even in the presence of uncertainties and therefore the number of balancing runs can be reduced. Furthermore, as the objective of optimum stochastic balancing is to minimize the residual vibration for general cases and it is more applicable for industrial use.; Also experiments were performed on a flexible three mass test rig for tests both on the theory and the balancing software. The influence coefficients for this test rig change over time. It provided us a very good simulator of real industrial rotors to test the new balancing methods. A few test runs were applied to obtain the distribution of the influence coefficients. The experimental results confirm that, under uncertainties with known distributions, the optimum stochastic balancing indeed outperforms existing balancing methods. |
