Comparative study on soft measurement methods of flotation foam concentrate grade based on deep neural network models

In recent years, deep learning technology has made great development in the field of intelligent industrial technology and it is widely used in the field of foam flotation. In this paper, deep neural network models were used to establish a grade soft measurement regression model for real-time foam grade measurement, and the study trained the models in combination with transfer learning. Firstly, flotation foam image dataset was established based on a mineral field image database. Secondly, by comparing the application performance of five deep neural network models based on the foam image dataset, it was determined that the three models (GoogLeNet, ResNet50 and ResNet101) were the best, and a hybrid deep neural network model was established based on the feature fusion of the three excellent models. Finally, the simulation experimental results show the accuracy of the hybrid model in the soft measurement regression application of flotation foam concentrate grade.