Research On Parameters Determination Of Blending Orthogonal Test And ANN Forecast

Much progress has been made on second-added coal technical compared to traditional blending pattern. The quality of ball-making's directly tested by vertical agglutination speed, the material level permeability as well as agglutination material thickness and so on, and this technical played an important role in the agglutination process.Firstly, in view of present ball-making situation of this steel company, the blending orthogonal tests were designed on the small pelletizer in order to observe the blending movement law and measure ball-maing rate, blending value by similar principle. The prime task was to study how the rotational speed influences the blending movement law through installing one monitor in the pelletizer, and then discussed how the water content, the packing rate and the depth of powdered coal increase influence the ball-making rate and the blending value.The analysis of test result has obtained the water content is the main factor influencing on the ball-making rate and the blending value, the rotational speed, the packing rate and the powdered coal increase depth is the secondary factor; Synthesizing two measuring test results, the research indicated:The best ball-making and blending is water content 7.1%, rotational speed 17r/min, the packing rate 11%, the depth of powdered coal increase is 5m.Secondly, BP artificial neural network chosen in this article was also elaborated in detail this article, including network architecture and learning algorithm. Taken the orthogonal test data as the sample, the BP network model by programming with Matlab7.0 was established, and then carried on the training, the forecast to the network.Finally, based on the VC++ formidable integrated development environment, the mix programming between Matlab and VC++ was made in use of application program interface, by transfering Matlab the engine storehouse function and the neural network toolbox's. The programming realizes the surface with the name of blending ball-making rate and the blending value forecast based on the neural network. Parameters can be input or selected in the tabulation frame, and demonstrated after operation.Through the research technique introduced in this paper, result about ball-making rate and blending value can be forcast preliminarily before the experiment carried, which saved the manpower and resources, raised the efficiency. Therefore, all of this really guided much the production of steel company.