Purchasing And Production Inventory Multi-Period Optimization Of Steel Enterprise

Due to the global economic crisis in2008and the European debt crisis in2009,the world steel industry is influenced greatly and Chinese large and medium-sized steel enterprise commonly appear profits decline or losses. Domestic steel enterprises have been always "heavy production light circulation", however smelting technology reform often costs huge but obtains benefits after many years, so more and more steel enterprises search profits from the logistics field. Procurement and inventory cost cover more than60%of total production cost in steel enterprise logistics, so optimization research is not only important to enrich purchase inventory theory,but also has practical significance to lower steel enterprise logistics cost, expanding the profit space.The dissert’s main content is as follows:Raw materials of Steel enterprise have the character of multiple kinds and large quality, different material has different purchasing environment and management is very difficult, so reasonable classification of goods is needed. First, Kraljic model is adopted for materials classification and management from two aspects of profit impact and supply risk, and a complete attribute index system is established. Then a linear programming model based on DEA is put forward to determine total index value. Through the model, positions of steel enterprise materials in the Kraljic matrix are obtained, and raw materials are divided into four classes:strategic items, bottleneck items,leverage items and non-critical items, and the corresponding purchasing strategies for each type of item are put forward.Before purchasing optimization of the raw materials procurement, steel enterprise managers need to analyze future prices of raw materials. Iron ore which has the most complex purchase environment and largest cost is taken for an example. Price forming mechanism and influential factors of iron ore are analyzed. When the influential factors can not be quantized directly, the expert system and event data analysis method are combined for quantization. Partial correlation coefficient and the sequence grey relational analysis are applied for influence strength analysis, then corresponding countermeasures and suggestions are put forward for steel enterprise.In order to overcome the poor fitting quality of general regression model, the paper proposes the nonlinear semi-parametric model to predict iron ore prices. Although the fitting precision has greatly increased, nonlinear semi-parametric model’s prediction results are not very ideal. So from the time series angle, the influence of various factors is taken as a whole variable for research. Support vector machine method based on particle swarm optimization (PSO-SVM) is adopted to forecast iron ore price. Comparing among auto-regression intergrated moving average model, the BP neural network method and PSO-SVM method, the results shows that PSO-SVM method’s prediction accuracy is higher, and more tally with the actual situation.Facing the fluctuation of raw material prices and random purchasing lead time, the dissert has established a stable demand purchase and inventory optimization model suitable for small steel enterprises. If demand is also random, the dissert has also established a purchasing stock optimization model suitable for large and medium-sized enterprises. From the aspect of muti-periods combination optimization, the two models have considered the influence of raw material price’s fluctuations. These two models aim to reduce the unite cost of purchasing, transportation and warehousing under muti-periods decision. Because of the model’s nonlinear characteristic, this dissert has put forward the improved particle swarm algorithm based on the price and steel demand forecasting. Steel enterprises L is taken for an example, results showed that the unit cost of raw materials of multi-period combination optimization is lower than the single period economic order quantity and the purchasing strategy to replenish inventories. It has proved the validity of the model and the algorithm. According to the purchasing decision table, the enterprise can save a lot of purchasing cost.The influence of raw material prices fluctuations on the steel production is the fluctuation of unit production variable cost. A dynamic optimization model for steel production is established under stochastic demand. The model considers multi-period production decision to deal with fluctuation of unit variable cost. Because of model’s large variables, particle swarm optimization algorithm is easy to fall into the local optimum, therefore the combination method of the simulated annealing and the particle swarm algorithm is put forward for solution. Take steel enterprise L as an example, the dissert proves effectiveness of the model and the convergence of the algorithm.Considering the interaction relationship between the purchase of raw material optimization and steel production optimization, this dissert establishes an integrated optimal model based on bi-level programming to describe the cyclic variation process.Through the case of steel enterprise L, concentrated decisions have more profits than scattered decisions.