Prediction of the Cement Grate Cooler Pressure in the Cooling Process Based on a Multi-Model Fusion Neural Network

Abstract

The reasonable control of the grate cooler is the key factor to ensure the heat exchange and cement clinker quality during the clinker cooling process. In this paper, the cement grate cooler pressure of the grate cooler is taken as the research object and a cement grate cooler pressure prediction model is proposed based on the analysis of the current status of the automatic control of the grate cooler. This model uses a multi-model fusion neural network algorithm that combines a BP neural network, a support vector machine and classification and regression trees with a neural network structure. Furthermore, the multimodel fusion quality characteristics are proposed, and the root mean square error and Pearson linear correlation coefficient of the multi-model fusion quality characteristics are used as the evaluation indicators for the prediction results of the multi-model fusion neural network. After the analysis of the cooling process of the cement clinker, we select seven input variables, and then complete the data preprocessing and model parameter selection. Finally, we predict the cement grate cooler pressure using a multi-model fusion neural network, a BP neural network, a support vector machine and classification and regression trees with three training sets to test sets ratios. Through the comparison of the root mean square error and the Pearson linear correlation coefficient evaluation indicators and their change trends, as well as the display and analysis of the final modelling results, it is found that the multi-model fusion neural network algorithm can greatly improve the accuracy of the prediction of the grate pressure, and at the same time it has good practicality for the accurate prediction of the cement grate cooler pressure in the industry.

Research on the Effect of Online Marketing Based on Multimodel Fusion and Artificial Intelligence in the Context of Big Data

“…Nowadays, multitask learning has become one of the most promising directions in machine learning. However, if machine learning wants to train a high-performance learner, it needs a lot of labeled data [18]. As a typical representative of machine learning, each training network in deep learning implies a large number of parameters to be learned.…”

Multi‐model fusion approach for electromagnetic inverse scattering problems

“…Inspired by Ref. [31][32][33][34][35], we propose a new approach for solving EMIS problems, called the Amplitude-Phase scheme (APs). Figure 4 shows the basic configuration of the developed solver for EMIS problems, and the proposed DL scheme is marked by the red rectangle.…”

“…Moreover, it has been successfully employed to solve EMIS problems as well [22][23][24][25][26][27][28][29][30][31]. Recently, researchers' interest has been attracted by multi-model fusion technology, due to its progress in the deep learning algorithm [32][33][34][35][36]. However, the application of multi-model fusion technology in solving EMIS problems has not been found yet.…”

Optimum control of under‐grate pressure of clinker cooler by optimizing the proportional integral derivative controller parameters using honey badger algorithm technique

“…1,2 As a result, maintaining output volume and lowering energy consumption depend greatly on the steady regulation of clinker cooling process. [3][4][5][6][7] The impact of clinker cooling and heat recovery is influenced by the grate cooler pressure, [8][9][10][11] which serves as a crucial parameter and illustrates the equilibrium relationship between hot material and cool air. [12][13][14] The temperature of the clinker was initially measured using thermocouples in grate coolers in order to properly cool the clinker, and the air pressure was then adjusted.…”

“…20 This not only reduced the need for additional sensing components, but also made it easier to handle disturbances like kiln load and speed changes. 4 In the past, PI controllers were employed to manage UGP. 21 However, they were slow to react the significant changes in kiln feed did not successfully react to grate speed adjustments.…”