Research On Demand Prediction Based Production Plan Problem Of Cold Rolling

Cold rolling production is an important sector of steel production process.Making effective production plan of cold rolling is of great significance for steel enterprises to achieve the continuity of production and reduce production costs.Most previous research works on the planning problems of cold rolling production are only to determine the processing sequence of available orders.However,if the demand information of orders which would be available in the following production periods can be predicted,these orders can be seen as dummy order and be added in the production plan of cold rolling in advance.It can effectively increase the number of available orders and reduce production costs.Taking the cold rolling production process in steel enterprises as background,the demand prediction based production planning problem of cold rolling is investigated.The demand of potential orders is predicted by using the data analysis technology.Decision of the processing sequence of available orders,as well as the dummy orders which would be available in the future,can explore more optimization potential,reduce production changeover cost,as well as guarantee the continuity and load balance of production.The main contents of this thesis are as follows.1)The prediction of order demand for cold rolling production is investigated.Based on the analysis of historical data of order demand for cold rolling production,orders are clustered according to their material group.Then,the demand predictive model is established by multi-kernel learning method to predict the order demand of different types.A differential evolution algorithm is designed to optimize the parameters in the multi-kernel learning prediction model.Improved strategies of mutation and crossover operation are proposed,which can effectively reduce the solving time of the prediction model.The numerical experiments show that the proposed prediction model and parameter optimization method are effective for the order demand prediction.2)A mixed integer programming model is formulated for the demand prediction based production planning problem of cold rolling.The task of the problem is to determine the processing sequence of available orders and dummy orders for cold rolling line.A mixed integer programming model is formulated by considering the technique production requirements,such as unit allocation requirements,processing time requirements and production setup requirements.The objective is to minimize the changeover cost of each unit,minimize the cost for production load deviation,and minimize the cost for demand prediction deviation.Finally,the model is solved by standard software CPLEX.The experimental result shows that the mixed integer programming is correct.3)A branch-and-price algorithm is designed for solving the demand prediction based production planning problem of cold rolling.By analyzing the structural features of the problem,the mixed integer programming model is transformed into a Set-Packing model.Column generation algorithm is adopted to obtain the lower bound for the problem.In the column generation algorithm,the pricing sub-problem is regarded as a variant of travelling salesman problem(TSP)problem,and a dynamic programming algorithm is designed to solve the pricing sub-problem optimally.At the same time,according to the technical characteristics of the problem,an acceleration strategy is proposed to speed up the convergence of the pricing sub-problem.The experimental results on randomly generated testing instances indicate that the proposed algorithm is more effective than directly solving the mixed integer programming model by the CPLEX solver.4)Since optimal solution algorithm cannot solve large-scale instances,an engineering optimization method is proposed for solving the production plan problem of practical scale.Two engineering optimization strategies are proposed:(1)the model dimension reduction strategy based on order clustering is proposed to reduce the number of variables,by clustering multiple orders into one single order.(2)The column generation algorithm-based approximate solution strategy is designed to select reasonable scenarios from all the columns that are generated in the column generation algorithm,by designing a scenario selection heuristic.The experimental results show that the proposed engineering optimization algorithm can obtain near-optimal solutions for problems in a relatively short time.5)Based on the technique requirements of production in steel enterprise and the operating favors of planners,a planning system for cold rolling production is designed based on order demand analysis.Embedded with the proposed model and algorithm,the decision support system can automatically generate the production plan of each cold rolling unit.