A sintering end-point prediction system integrating data technology and process knowledge was proposed.First of all,the mass historical data of actual sintering production was collected,cleaned and integrated,a big data warehouse for sintering was established.On this basis,the relation-ships between the sintering end point and the sinter quality indexes were analyzed.Then,the important characteristic variables related to the sintering end point were selected by combining data mining with process knowledge.The prediction model of sintering end point was established by using Gradient Boosted Decision Trees(GBDT)algorithm,the parameters involved in the algorithm were optimized by means of grid search and cross validation,and the corresponding expert rules were established in the outer layer of the prediction model.The overall forecast hit rate of the forecasting system is over 88%.Compared with the previous prediction models,the prediction accuracy and generalization ability of the prediction system are improved significantly,which plays an important role in guiding the actual pro-duction of sintering.
sintering burn through point predictionparameter screeningGBDT algorithmexpert rules
NETL
NSTL
万方数据
