一种新型的灰色RBF神经网络建模方法及其应用

Novel modeling method for grey RBF neural network and its application

针对神经网络建模预测时,其建模精度往往受到数据随机性的影响,以及灰色累加生成操作(AGO)具有减小数据随机性,使数据变得有规则的特点,提出了一种新型的建模预测模型———灰色径向基(RBF)神经网络模型。此模型能够减小数据中的随机性,加快网络的建模收敛速度,使神经网络的建模精度得以提高。将此灰色RBF神经网络应用到动调陀螺仪漂移数据建模中,并将其建模验证结果和单纯使用RBF网络的建模结果进行比较,结果证明此方法是可行而有效的。

A novel neural network called grey radial basis function network (GRBFN) is proposed. The reasons why grey theory is introduced into the RBF neural network are based on two facts. First, the modeling performance will be affected by the randomness inherent in the data when neural network approach is used for the model. That is, poor performance results from great randomness, and vice versa. Second, grey accumulated generating operation (AGO), a basis of the grey theory is effective to reduce randomness. Because of these facts, the GRBFN model is presented and expected to have a better modeling precision of random drift in dynamically tuned gyroscope (DTG). The novel grey RBF network is applied to drift modeling of DTG. The numerical results of real drift data from a certain type of DTG prove the effectiveness of the proposed GRBFN model successfully. The RBF neural network modeling approach is also investigated to provide a comparison with the GRBFN model. Under the identical training condition, the GRBFN's training speed is enhanced greatly.