BP Neural Network Optimization Method Based on Wavelet Basis Function
At present,BP neural network is widely used in deformation prediction,but its convergence speed is slow due to the influ-ence of local extremum.In view of this shortcoming of BP neural network,this paper replaces the activation function of BP neural net-work with wavelet basis function and improves the weight and critical value of BP neural network to form wavelet neural network.Wavelet neural network has excellent time-frequency localization and self-learning ability.After the scaling and translating transfor-mation of wavelet decomposition,we can obtain the series consistent with the properties of the approximation function,which can be used for deformation prediction.At the same time,after introducing two new variables,scaling and translating,wavelet neural network will have more degrees of freedom than wavelet decomposition,and then the accuracy of numerical simulation will be better.The ex-perimental results show that compared with BP neural network,wavelet neural network has higher convergence efficiency and smaller error in deformation prediction,and can achieve better prediction effect.
BP neural networkwavelet basis functionwavelet neural networkdeformation predicationquality analysis
NETL
NSTL
万方数据
