【摘要】 在对人工神经网络和转子动平衡技术深入研究的基础上，提出了将神经网络技术应用于转子动平衡领域的方法。文中对网络拓扑结构的确定以及学习样本数量、样本质量、学习误差等对网络性能的影响等问题作了深入的分析，得到了有实用价值的结论。将该方法应用于实际的动平衡过程，结果表明，基于神经网络的平衡方法的计算结果优于传统的方法。更多还原

【Abstract】 It is well known that Artificial Neural Networks (ANN) has been widely used in many industry fields including mechanical diagnosis.But till now ANN is not applied in rotor balancing for the following reasons: (1) versatility of data being bad; (2) difficulty in acquiring sufficient data; (3) modal complexity of rotor system. In this paper,a new method is recommended,which can overcome these difficulties and apply ANN to rotor balancing successfully. The method is based on the combination of ANN and rotor balancing. The determination of neural network topological structure is studied firstly,and then some important related elements which affect the performance of ANN are analyzed in details. These elements include the number of learning samples,the quality of learning samples and the learning errors. Conclusions which are of great importance in real application are obtained. With applying to rotor balancing,it is found that this method is better than traditional method such as influence coefficient method.In other words,the method can finish rotor balancing with smaller correcting masses and no degradation of the balancing level. At the same time,the number of rotor starting is not bigger than that of influence coefficient method.更多还原