Ran Liu, Huan Jin, Song Liu, Xiao-Jie Liu, Jian-Hai Hao, Jun Zhao, Qing Lv
To ensure the uniformity of the gas fiow in the sintering material layer, improve the sintering efficiency, and reduce the production energy consumption, it is of great significance to predict the permeability index of the original material layer in advance. However, how to achieve accurate prediction in line with the actual production environment has always been a challenge. Based on this, deep learning was combined with finite element numerical simulation, and an integrated pre-diction method with high interpretability and controllability was proposed. This method used the wavelet threshold denoising technology jointly improved based on complete ensemble empirical mode decomposition with adaptive noise (CEEN) to process the original data, so as to improve the data quality. Subsequently, a temporal convolutional network-long short-term memory (TCN-LSTM) model was constructed and trained for permeability prediction. Comparative analysis showed that the proposed model has a higher prediction accuracy than other comparative models, with the coefficient of determination R2 as high as 95%. In the experimental simulation stage, taking a 360 m2 sintering machine of a certain steel plant as the research object, the COMSOL finite element software was used to establish a physical model for process simulation. The results showed that the variation curve of the permeability of the material layer along the depth direction is highly consistent with the measured results, with a relative error of approximately 3.90 and the R2 of 92.38%. In addition, based on the results of finite element numerical simulation, when using the TCN-LSTM model for prediction, the difference between the predicted value and the simulated value is small, with an average relative error of only 4.92%and the R2 of 97.29%, showing a high degree of fitting and matching. Therefore, the method of combining finite element numerical simulation with CEEN-TCN-LSTM can accurately predict the permeability index of the material layer, effectively meeting the dual needs of predicting the permeability in advance and monitoring the change process of the material layer in actual production and providing technical support for the optimization of the sintering process and the production of high-quality sinter.
