一种新矩在水电机组轴心轨迹识别中的应用

结合不变矩和神经网络提出一种水电机组轴心轨迹自动识别方案，证明了离散点状态下曲线不变矩具有平移不变性和旋转不变性，但不满足比例因子缩放不变性;构造一组同时满足平移不变性、旋转不变性和比例因子缩放不变性的新曲线不变矩;通过引入动量项的BP神经网络对轴心轨迹自动识别。并给出了仿真和实际水电机组轴心轨迹识别诊断结果;对初步定性诊断结果进行定量分析，计算轴心轨迹形状特征参数，为故障定位分析提供参考．

Application of a new moment invariant to identification of axis orbit in hydro-generator unit

A method to automatically identify axis orbit of hydro-generator unit on the basis of moment invariant and BP neural network was proposed. The influences of scale factor on curve moment invariant features in discrete state were analyzed and an improved curve moment with scale, translation and rotating invariant was introduced. The simulation and real diagnosis examples were given to prove that the construction of the performance of the new moment invariant. Momentum item was introduced to improve BP neural network learning speed. It acquired parameters of the axis orbit shape to help operation personnel to diagnose by optimization method.