Research On Soft-Sensing Model Of Cement Clinker Free Calcium Based On Fine-Grained Feature Extraction

The content of free calcium oxide(f-CaO)in cement clinker is an important index to measure the quality of clinker calcination,and its accurate measurement plays a very important role in the optimization control of industrial process.At present,there is an obvious time lag in the determination method of f-CaO content,which is lack of guidance for real-time cement production control system.In this paper,problems such as multiple working conditions and unbalanced distribution of sample labels in the process of cement firing were discussed.A soft measurement model based on f-CaO content of fuzzy fine-grained cement clinker was established by using data-driven soft measurement technology to realize real-time online detection of cement f-CaO content.The main contents of this paper are as follows:Firstly,the new dry cement production technology and the formation mechanism of clinker f-CaO in the process of cement firing were summarized,the characteristics of cement production process were analyzed,the modeling difficulties of clinker f-CaO were studied and the solutions were proposed.The important factors affecting the content of f-CaO in the cement production process were analyzed,the relevant variables were selected as the input of the model,and the historical data were preprocessed.Secondly,based on fuzzy granular cement clinker soft measurement model,this model by fuzzy membership rules combined with convolution neural network,fuzzy classification of the cement sample data,the fuzzy rules under points further feature extraction and good sample regression prediction,and finally through the study of the solution of the predictions of a subclass blur the final prediction results are obtained.The experimental results show that the soft sensor model based on fuzzy fine granularity has high prediction accuracy and generalization ability,and can be used to realize the real-time online prediction of f-CaO content.Then,based on fuzzy granular cement clinker soft measurement model,the data sample quantity differences between categories,inadequate training subclass containing a small amount of data model,in order to solve these problems,puts forward a kind of virtual sample generation method based on neighborhood remain embedded,compared with general method of sample data enhancement,using this method can learn more details of the features in the data,Improve the quality of samples generated.The generated data and the original data were combined to reconstruct the training data set to train the sub-models corresponding to a small number of samples in the fuzzy fine-grained soft measurement model.The accuracy of the sub-models was verified to be improved,and the performance of the fuzzy fine-grained soft measurement model was optimized.Finally,the field data are used for experiment and analysis,and the comparison experiment is set for comparison.The results show that the proposed soft sensor model has high precision and strong generalization ability.