Modeling and predicting inventory variation for multistage steel production processes based on a new spatio-temporal Markov model

Abstract

Inventory control and variation reduction are critical and complicated issues for multistage production processes (MPP) because reasonable inventory is key to ensuring continuous production and on-time order delivery in iron and steel enterprises. However, due to the uncertainties in production environments, physics-based models cannot accurately and efficiently approximate inventory variation propagation in MPP. Moreover, classical statistical models usually fail to consider material production processes with spatio-temporal movement characteristics. Therefore, in this study, a spates-temporal Markov model (STMM) with the probability chain adjustment (STMMPC) is developed to predict states of inventory variation and analyze inventory variation propagation in multistage steel production processes. Firstly, the STMM is established, where the expression of the state transition probability matrices is derived based on both spatial and temporal dimensions. Secondly, probability chains and probabilities of joint states following the chains are defined and used to further improve the prediction accuracy of the STMM. Finally, a differential evolution algorithm with self-adaptive mutation strategies is adopted to optimize the weights of the probabilities in STMMPC. The results based on actual steel production data demonstrate that the STMMPC is superior to STMM and regular Markov models, and the model is relatively stable against changes in the weight parameters. Furthermore, the proposed method can assist managers with better production plans to maintain optimal inventory balance.