FeO content prediction method based on sintering machine tail section video

Abstract

The invention discloses a FeO content prediction method based on a sintering machine tail section video, and relates to the field of steel production. The method mainly comprises the following steps:determining a falling moment, removing an abnormal audio sequence, dividing an image sequence and preprocessing, carrying out audio preprocessing, extracting image features, extracting audio features,constructing a video quality database, and constructing, training and testing a quality prediction model. Compared with the prior art, the method has the advantages that the severe conditions of thesintering machine tail are considered, the characteristics of the video information are represented more comprehensively from the aspects of images and audios, and quantitative prediction of the FeO content can be achieved.

Description

technical field [0001] The invention relates to the field of iron and steel production, in particular to a method for predicting FeO content based on a sintering machine tail section video. Background technique [0002] In the iron and steel industry, the quality of sinter has a very important impact on the performance of the final blast furnace smelting, and the performance of blast furnace smelting, such as utilization coefficient, fuel ratio, etc., will further affect the final economic and technical indicators. The quality of sinter is related to the moisture content of the mixture, the uniformity of distribution, the position of the sintering end point and the FeO content level in the sinter. The high quality of the sinter is the guarantee for the realization of high-quality and high-yield steel. Therefore, the quality of sinter plays a vital role in the whole iron-making process. Among them, the level of FeO content has a direct impact on the reducibility of subsequent...

AI Technical Summary

Problems solved by technology

[0006] (1) In the method of predicting based on visible light images, the fixed time interval cannot effectively extract key frames under real-time conditions. In addition, directly removing the values ​​​​of the G channel and B channel for red light processing loses the correlation Information

[0007] (2) In the method based on visible light and infrared images, because the extracted feature network is relatively simple, and the network structure is simple, the FeO content cannot be quantitatively predicted

At the same time, due to the time lag between sampling and testing, the key frame and the detected FeO content cannot correspond well

Images