Prediction Model Of Sinter Quality Based On Big Data And Machine Learning

Sinter is the main raw material of blast furnace.The sinter data are characterized by nonlinearity and hysteresis.A model of sinter ore quality prediction based on the big data of the whole process of sinter and with machine learning as the technical architecture is proposed to provide accurate and effective quality information for the sinter site.Integrate data in SQL Server and Oracle databases to increase data quality in data pretreatment technology.The time span is constructed for 2 years,including 74parameters,and the sinter the whole-process data sample set of 13175 sets of data.The delay and ingredient stability rules of the return data were fused to fusion TCN and DF,building the combined forecast model of number of ore returned.The results show that the elimination of the lag rending forecast can interpret the Explained var score(Evar)to increase by about 20%.Classification forecast ore Evar can reach more than84%.In the error of 9%,the forecast hit rate is more than 90%.According to the sinter end position,the drum index is divided into non-normal burning and normal burning,and the prediction model is established by the Catboost algorithm.Compared to data sets,the determination coefficient(R~2)of abnormal burning and normal burning was increased by 2.43%and 5.03%,respectively.In the error of0.25%,the forecast hit rate is more than 95%.The data decomposition technology and the characteristic construct method are used,and a single-variable input model is fused,and a multivariate input model is used to present a comprehensive forecast model of FeO content.The results show that the comprehensive forecast model is significantly reduced and close to zero,the hit rate within the real value prediction section can reach more than 90%.The number of ore returned,drum index,and FeO content are predicted as the quality parameters of sinter.Clustering of the three quality indicators by using the K-means algorithm,the quality of sinter is defined as excellent,good,general and poor.It solves the problem of incomplete information of sintering ore single quality index,it is useful for staff to obtain a clear picture of the quality status of sintered ore.Figure 50;Table 23;Reference 117...