Response Data Repair Method of Gate Structure Based on Improved Convolutional Neural Network
Aiming at the problem that the missing data affects the safe operation of structures in the process of struc-tural health monitoring,a data completion method based on improved convolutional neural network is proposed.By using the matrix completion method,the multi-channel data is repaired simultaneously.Firstly,the missing signal is obtained by Fourier transform,and the real and imaginary parts of the matrix are used as the input of training,and the inner prod-uct is made with L1-CNN neural network.The base coefficient is used as the basis of the convolution operation,and the original signal and repair signal are compared and updated,and L1 regularization is applied to update the convolution ker-nel,and the repaired signal is obtained.Finally,the acceleration sensor signal of spillway hydraulic gate structure is taken as an example to verify the rationality and effectiveness of the proposed method.The results show that the proposed data completion method can repair the signal with a missing rate of less than 30%without distortion.The signal with 40%missing rate was repaired,and some areas were found to be distorted.Through comparison and analysis with other meth-ods,the data completed by this method has the highest fitting degree with the original data,and the forecast trend is close to the complete data.
structural health monitoringdeep learningdata restorationneural networksstructural response
万方数据
维普
