摘要
讨论了基于神经网络的球半径测量方法。当球面点的坐标分量之一已知，而另外两个分量固定不变但未知时，通过建立球半径与辅助变量之间的关系，可达到测量球半径的目的。一种基于递推预报误差（ＲＰＥ）的算法被用来训练神经网络，该算法优于常规的反传（ＢＰ）学习算法。试验表明该方法精度高、应用方便，具有一定的理论意义与现实意义。
狪t is essential to obtain the radius of a spherical part for precision test and measurement. Usually the radius can be calculated by solving a series of spherical surface equations when all the coordinates (xi, yi, zi) (i=1,2,3,4) of four spherical points are available. This paper has discussed the measurement of the radius of a sphere based on neural networks. Through constructing the function between the radius of a sphere and its auxiliary variables, the radius can be obtained if one of the coordinates of each spherical point is measurable and the other two are constant and unknown. A recursive prediction error based learning algorithm which is more effective than the conventional back propagation algorithm is applied for training the neural networks. The experiment results show that the measurement method can be applied easily with high precision.
出处
《计量学报》
CSCD
北大核心
1998年第3期170-174,共5页
Acta Metrologica Sinica
关键词
球
半径
测量
神经网络
Sphere
Radius
Measurement
Neural network
