Prediction of Sulfur Removal from Iron Concentrate Using Column Flotation Froth Features: Comparison of k-Means Clustering, Regression, Backpropagation Neural Network, and Convolutional Neural Network

Froth feature extraction plays a significant role in the monitoring and control of flotation process. Image-based soft sensors have received great deal interest process due to their low-cost non-intrusive properties. This study proposes data-driven sensor models based on froth images predict key performance indicators The ability multiple linear regression (MLR), backpropagation neural network (BPNN), k-means clustering algorithm, convolutional (CNN) amount sulfur removal from iron ore concentrate column was examined. A total 99 experimental results were used develop predictive models. Extracted features including color, bubble shape size, texture, stability, velocity train traditional models, whereas CNN model directly fed into model. comparison indicated that three-layered feedforward NN (17-10-1 topology) provided better predictions than MLR algorithm. BPNN displayed correlation coefficient 0.97 root mean square error 4.84% between actual data output for both training testing datasets. percentages CNN, BPNN, 10, 11, 15 18%, respectively. can become technical support application intelligent operational variables desulfurize concentrate.