Method for predicting chemical components of sintered ore on basis of deep belief network

Abstract

The invention discloses a method for predicting chemical components of sintered ore on the basis of a deep belief network. By the method, according to chemical components of a sintered mixed material,the chemical components of the sintered ore are predicted by a prediction method on the basis of a DBN (deep belief network) algorithm; the method comprises the following specific steps: firstly, acquiring historical data during actual production of a sintering plant, eliminating abnormal data and normalizing; then determining input and output parameters which affect the quality of the sintered ore, and verifying the rationality of the input parameters by a gray correlation analysis method; then establishing a prediction model for predicting the chemical components of the sintered ore on thebasis of the DBN, training by using the historical data to optimize the prediction model; finally, predicting the chemical components of the sintered ore by using the prediction model, and reversely normalizing results to obtain predicted values of the chemical components of the sintered ore. Compared with the prior art, the DBN-based prediction model provided by the invention has the advantages as follows: approximation of a complex nonlinear function can be more accurately achieved, the prediction accuracy of the chemical components of the sintered ore is improved, and application and popularization values in actual production are achieved.

Description

technical field [0001] The invention belongs to the technical field of iron and steel smelting, and relates to a method for predicting the chemical composition of sintered ore, in particular to a method for predicting the chemical composition of sintered ore based on a deep belief network. Background technique [0002] Sinter is the main raw material for blast furnace ironmaking, and the chemical composition of sinter is an important indicator for evaluating the quality of sinter. Batching is the first process of sinter production, which has a great influence on the chemical composition of sinter. Due to the wide source, variety and complex composition of sinter chemical raw materials, the sintering process has the characteristics of long time lag, strong coupling and nonlinearity, making it difficult to accurately control the chemical composition of sinter. It is of great significance to improve the quality of sinter by accurately predicting the chemical composition of sin...

AI Technical Summary

Problems solved by technology

The neural network, gray theory and support vector machine mentioned in the above literature belong to the shallow learning algorithm. When the shallow learning algorithm is given a limited number of samples, it is difficult to effectively represent the nonlinear complex function, and the generalization ability is limited. restrictions, which in turn affect the prediction results of chemical composition of sinter

Images