Raw Material Demand Prediction And Process Analytics On Order-Planning In Iron & Steel Industry

Adequate raw materials in iron and steel industry are the prerequisite that ensures orders’normal continuous production. Nowadays, the unbalanced supply plan and the actual demand of raw materials result in unreasonable stock of raw materials in many iron and steel enterprises. Exact prediction of the factual demand of raw material could optimize inventory management. The factual demand, however, is difficult to achieve due to the uncertainty and fluctuation of production. Steel production is a multi-stage process. Order plan is throughout each steel production process. any accident in the production process may lead to order tardiness. Order process refers to the completion of the order on each production stage, including the excess/owed amount and the product specifications of the orders. The prediction of order process is to predict the completion of the future orders in this phase based on the history data of this phase, and then give guidance to the production plan and guarantee the lead time of the orders. Therefore, study the prediction method for the demand of raw materials and the order process is significance to guarantee the continuation of production and the rationality of inventory structure.This thesis studies the raw material demand predicting problem and the order process predicting problem associated with order plan. Least squares support vector regression (LSSVM) and improved particle swarm optimization are used to analyze history data and establish optimal prediction model. Based on the proposed model and method, a decision support system of hot rolling order process prediction is developed and further implemented in practical production. The main work of the thesis is composed of three aspects listed as follows:(1) Based on the background of the raw materials in blast furnace production, the raw material demand prediction problem was studied. For this problem, using current inventory, actual demand, prices and other influential factors of raw materials as input, a dynamic demand predict model of raw material was built by LSSVM, and an improved particle swarm optimization with guiding set strategy was designed to optimize parameters of the model. The experimental results based on actual data show that the proposed algorithm can give a good solution in raw material demand predicting, thus provide the basis for the decision of raw materials procurement planning and reduce the purchasing cost.(2) Based on the background of order production in hot rolling process, the order process prediction problem was studied. For this problem, an order process prediction model was built by LSSVM, and an improved particle swarm optimization with dynamic adjustment learning factor strategy was designed to optimize parameters of the model. Experimental results show that the proposed algorithm can predict the order process effectively.(3) Based on the prediction models and methods proposed, taking an iron and steel enterprise as background, the decision support system of hot rolling order process was developed. The system is real-time monitoring the completion of the production plan, and provides the basis for adjusting production plans.