Change is safer: a dynamic safety stock model for inventory management of large manufacturing enterprise based on intermittent time series forecasting

As a key issue of inventory management for enterprise after-sales service, safety stock is dedicated to ensuring maintenance reliability while keeping low inventory cost. Existing researches of safety stock are mainly designed for frequent and stable spare parts demand (SPD) rather than fluctuating and intermittent demands in practical operation and maintenance of large manufacturing enterprises. Consequently, emergency restocking will be frequently triggered, or an over-safe inventory level with excessive inventory cost will be set. This paper proposes a dynamic safety stock model for after-sales service of large manufacturing enterprise. First, this paper builds a multi-objective safety stock optimization model based on a three-level warehousing architecture. The multi-objective genetic algorithm is adopted to calculate the static values of reorder point (RP) and maximum stock (MS) by simultaneously minimizing excessive inventory cost and shortage cost. Second, this paper proposes a robust forecasting algorithm for multi-variable intermittent time series by integrating the inter-sequence structured information and temporal evolution information of sequence itself. Finally, this paper designs a dynamic information fusion mechanism by integrating the demand prediction values into the static stock. An actual after-sales spare parts dataset from a Chinese large rail transit manufacturing enterprise is introduced for validation. The experimental results show that the proposed model can accurately predict the demand trend of spare parts, and improve the inventory turnover rate as well as requirement coverage. This work is then believed to provide a new idea for inventory management: dynamic information can boost maintenance efficiency and safety, that is, Change is Safer.