| Sintering is an important step in the traditional blast furnace ironmaking,the sinter quality directly determines the quality of the subsequent blast furnace ironmaking.In recent years,with the gradual shortage of rich iron ore resources,people began to use lean iron ore which abundant in the natural reserves.Then lean iron ore was sintered to be high quality sinter,which can be treated as a raw material for ironmaking.Sintering is a highly time lag and highly coupled nonlinear complex reaction process.Which can be based on the quality of the current sinter to adjust the sintering process and improve the quality of the subsequent sinter.Chemical detection is the most accurate way to judge the quality of sintered ore,but it takes two to three hours from the sampling to the test results,and the real-time is too bad,which is of little significance to the reference design of the current sintering.Based on the fuzzy control method of production parameters,although the sintering quality level can be predicted by initial sintering parameters,the production parameters tused are not the tracking data of a certain section in the whole sintering process,resulting in poor accuracy of the prediction results.The tail section observation with a high degree of real-time.Experienced workers can quickly determine the sinter quality according to the characteristics of the tail of sinter section observation.But through the sinter section image to judgment the quality of sintered ore has a certain degree of subjectivity,and a long time to observe the tail sinter section will make the workers visual fatigue which may lead to mistakes.Therefore,in order to overcome the shortcomings of the above two methods,in this paper,we use the weighted support vector machine(W-SVM)to estimate the quality of the sintered ore by using the multi-parameter extraction and analysis of the tail image and the process parameters.In addition,this article also established a database to store data,and write client programs to achieve remote access.After a long time in the Maanshan Iron and Steel Co.,Ltd.second iron plant sintering production line to verify the actual operation.This system is not only efficient,intuitive,real-time,but also improved the accuracy of the sinter quality judgement,has a highly practical application value.The main work of this thesis includes the following contents:1.Combine the color image and the infrared thermal image of the tail section.Extraction of red fire layer temperature,thickness,continuity and other data.By employing the regional gray scale ratio filtering algorithm to realize the optimal cross-section image recognition.2.Write OPC program to connect the factory network,to obtain the temperature of the bellows,the temperature of the flue,the conveyor cooling temperature,sintering machine speed and other sintering process parameters.And the sintering history data of the best sinter section image are obtained by inversion.3.Collect test samples and artificial samples as a result of comparison and machine learning data.4.Write the sinter quality online judge system main program.5.The above collected data is used as a training sample to train the weighted support vector machine(W-SVM).The prediction model for regression estimation is obtained and finally used for real-time prediction of FeO content and drum index.And the quality judgment model was used to determine the sintering quality grade.6.Building database,to store the sintering parameters,digital data and image data obtained by applying the main program of the online judge system of sinter quality.7.Write the database remote access client program which can query the sintered data from remote.And it will help to guide the production and summarise experiences.8.Finally,the data of the online judge system of the sinter quality and the corresponding test data are listed,and the results were analyzed. |
PDF Full Text Request