A novel integrated soft sensing model for online cement clinker quality monitoring based on fuzzy fine-grained classification

Abstract

The content of free calcium oxide (f-CaO) in cement clinker is an important index for cement quality. For the problems of multiple working conditions and unbalanced distribution of sample labels in cement clinker production, a soft sensing method of cement clinker f-CaO content based on fuzzy fine-grained classification (FF) is proposed. First, a divide-and-conquer strategy is used to divide the samples into high, medium, and low subsets according to cement clinker f-CaO and extract the fine-grained features under diverse types of multiple production conditions. Second, fuzzy classification based on the membership function is used in the FF model to solve the uncertainty of the sample categories. To ensure the rationality of the classification, the fuzzy membership rule is combined with a convolutional neural network to implement the fuzzy classification method. Finally, different feature extraction methods are proposed to be selected according to the data size of various categories of samples. After experimental validation, the evaluation metrics of $\mathrm{RMSE}$ decreased by 4.5% and $R^2$increased by 17.8% for the direct classification model compared to the single model. The $\mathrm{RMSE}$ of the fuzzy classification model over the direct classification model was again reduced by 2.34% and $R^2$was again improved by 7.24%, showing the effectiveness of the proposed soft measurement model.