中文摘要：

在水电机组状态检修系统中，轴心轨迹是判断机组状态的一个重要特征。该文提出边缘检测和矩特征提取相结合的方法，利用粒子群寻优算法来获取与待识别样本最接近的已知样本，应用改进的BP神经网络进行识别，将轴心轴迹的不变性矩作为神经网络的特征参数，对几种典型的轴心轨迹进行了辨识。某水电站机组试验表明该方法识别速度快、精度高，具有较高的实用价值。

英文摘要：

In the maintenance system of hydropower units, shaft centerline orbit is an important feature for the diagnosis of the unit condition. An algorithm combining edge detection with moment feature extraction was presented. Particle swarm optimization （PSO） algorithm was used to obtain the nearest known sample with samples to be identified. This paper applied improved BP neural network for identification. It took the invariant moments of shaft centerline orbit as the characteristic parameters of neural networks. Some typical orbits were also identified. Experiments in some hydropowcr station show that this identification method is fast and precise, with a highly practical value.