Estudo comparativo de modelos de estoques num ambiente com previsibilidade variável de demanda.

Abstract

The objective of this study is to compare the performance of inventory models in a large range of costs and demand environments. The compared models are the traditional periodic and continuous Up to Maximum Inventory Level, Base Stock and Fixed Lot Size, and another model, based on the Material Requirements Planning (MRP) logic and here referred to as Requirements Planning, which uses demand forecast to quantify the acquisition decisions. In the first step, simulation and neighborhood search are used to select the best of 4 forecasting models, which generates the forecasts to the Requirements Planning model. Single Exponential Smoothing, Holt’s linear method, Single Exponential Smoothing with seasonality, and Holt-Winters’ trend & seasonality method are these 4 models. In the second step, simulation and neighborhood search are used again to optimize the inventory models parameters. The items’ demand time series are based on trends and seasonality defined arbitrarily plus the addition of a irregular random component. The period average Purchase, holding, shortage and total costs are calculated and the models are ranked, based on the total cost criterion. The results show the superior performance of the Requirements Planning model in practically all tested conditions, with the continuous Up to Maximum Inventory Level in a secondary position. The results show, too, the poor performance of the continuous Base Stock model, base of the Kanban system, in opposition to its actual hard recommendation as robust model. This study has, as major contribution, the evidence of the Requirements Planning model superior performance compared with the traditional inventory models.