Prediction of Cement Strength Based on Hybrid Neural Networks

Cement grinding is the last link in the cement factory, and its working conditions are complex, resulting in the optimization and adjustment of grinding parameters becoming complicated and difficult, while the prediction of the strength of the finished cement product has a crucial position in helping cement grinding to be better controlled. In order to better predict cement strength, this paper constructs a hybrid deep learning model combining convolutional neural network (CNN) and bidirectional gated recurrent unit (BiGRU) with an attention mechanism to solve the above problem. The model parameters are updated using stochastic gradient descent method and loss function during training. Finally, the validity, accuracy, and superiority of the CNN-BiGRU-Attention model were verified by using the data for testing and comparing it with other models using standard error and coefficient of determination as metrics.