Based On The Application Of Intelligent Algorithms In Nerual Networks For The Prediction Of Coke Quality

The coking process is a complex production process which contains thecharacteristics of multi-parameter, time-varying, nonlinearity and uncertainty etc. Aseries of physical and chemical changes occur in this process and therefore it is difficultto use a conventional, determined of the mathematical model to predict the quality ofthe coke. And the neural network is the complex network of systems which areinterconnected by a large number of neurons, it can achieve or approximate the mappingbetween the desired input and output because of having a strong ability to simulatenonlinear systems. This nonlinear mapping capability makes it approximate a nonlinearof function at arbitrary precision. The neural network provide a new method for solvingsuch problems based on the above advantages.First this article introduces the development status of coking, such as preheatingtamping coking technology, coal coking technology, coal humidity and large-scale ofcoke oven, and so on. The analyzes to the research status of coke quality prediction,such as coke ash content and sulfur content prediction, coke cold strength prediction,coke thermal state strength prediction.Then，introduced the process of coke quality prediction model based on multipleregression analysis. With the volatile (Vd af), the glial layer thickness (Y), the ratio ofcoke oven height and width (H/D) and（H/D）2as independent variables, Withcrushing strengthM40and abrasion strengthM10for coke, as an important index ofcoke quality prediction, to build the model and formula ofM40andM10And themodel formula of the regression relations for the significance test. The results ofsimulation show that there is a good linear relationship to the coke strength modelformula, the system model has a better prediction to the coke quality predictioneffectively, and can meet the requirements of coke predict effectively, which provide animportant basis for the improvement of coking technology.Secondly, the crushing strength, abrasive resistance, reactivity index and strengthafter reaction for the quality of coke, base on the analysis of genetic algorithm and BPneural network algorithm, genetic algorithm optimization to the BP neural network is established on the basis of the coke quality prediction model. Base on the analysis ofgenetic algorithm and RBF neural network algorithm, genetic algorithm optimization ofthe RBF neural network is established on the basis of the coke quality prediction model.The results of simulation show that the RBF neural network bases on genetic algorithmto optimize performance of coke quality prediction model is superior to the geneticalgorithm optimization of BP neural network prediction model, and superior to the cokequality prediction model base on multiple regression analysis, and easy to implement, atthe same time, the network has good learning ability and strong applicability, predictionaccuracy and prediction ratio should be higher.Finally, make a summary to the article.