Study On The Data Analytics Based Material Utilization In The Steel Industry

Taking steel production process as background,the production output variation analysis of multiple processes and the data analytics based open-order slab allocation problem are investigated in this thesis.This research has a great significance in reducing the generation of inventory materials,increasing the open-order slab utilization,and increasing the profit of steel enterprise.In this thesis,combining with data analytics and optimization techniques,how to improve the material utilization in the steel industry from the aspects of reducing inventory material generation and optimizing inventory material allocation are investigated.By analyzing the historical data of production output,a Markov chain model of output variation is built to predict the variation of the production output in different stages,so as to guide the accurate feeding and reduce the generation of inventory material.By forecasting the order demand,the forecasted orders as well as the customer orders are used to be matched with open-order slabs.An integer programming model is formulated with the consideration of order demand.Then an improved discrete differential evolution algorithm is proposed to solve the problem.The model and algorithm can significantly expand the optimization space and improve the open-order slab utilization.The main research contents of this thesis are as follows.(1)Study on the production output variation analysis of multiple processes in the steel industry.The production output variations of five typical processes including iron-making,steel-making,continuous casting,hot rolling and acid rolling in a large domestic steel enterprise are selected as prediction object.Through the analysis of history data,the output variation is divided into different variation states.A Markov chain model of output variation between two adjacent processes is built.The computation results indicate that the Markov chain model can accurately forecast the output variation.In addition,the accuracy of the Markov chain model across multiple processes is also tested.(2)Study on the prediction problem of the order demand.Taking the historical data of hot rolling orders in a large domestic steel enterprise as the data source,the orders are classified by analyzing the historical order demand,and time series models are built for different types of order,where the parameters are optimized by the distribution estimation algorithm,so as to forecasting the demand required by different types of order.Finally,the rationality and accuracy of the model are tested by numerical experiments.(3)Study on the open-order slab matching problem with the consideration of the forecasted demand orders.In this thesis,the forecasted demand orders as well as the customer orders are used to be matched with the open-order slabs.The scenario tree is introduced to reduce the risk of inaccurate orders prediction.Then,a stochastic slab matching model is built.An improved discrete differential evolution algorithm is designed to solve the problem effectively.Random disturbance and neighborhood search are designed to speed up the convergence rate of the algorithm.Finally the validity of the proposed algorithm and the improved strategy is proved by a numerical experiments,and the scenario tree can improve the robustness of the model is proved as well.(4)A decision support system for open-order slab matching considering the forecasted demand orders is designed and developed based on the above research,the function of orders demand prediction and the automatic allocation for the open-order slabs are realized,improving the efficiency of the planning staff.