Priority fractional rationing (PFR) policy and a hybrid metaheuristic for managing stock in divergent supply chains

Kurian John,Brijesh Paul,Chandrasekharan Rajendran,Hans Ziegler

DOI: https://doi.org/10.1007/s12046-022-02011-0

2022-11-26

Abstract:A distributor catering to demands of multiple retailers is considered in this paper and stock-management in this divergent supply chain is achieved through the deployment of periodic review base-stock (i.e. ( R , S ) policy) policy at every member. In the model of the supply chain considered in this study, in every time-period, an attempt is made by the distributor to first transport backlogged-demands from the downstream members till the distributor's previous instance-of-review even before considering demands from downstream members in recent time-periods. This practice of the distributor attempting first to satisfy backlogged-demands till its last instance of review will ensure that the shipment will reach the retailer, contemplating whom the order was made by the distributor. A Mixed Integer Linear Programming (MILP)-based mathematical formulation of the supply chain (with the objective of minimizing the Total Supply Chain Cost (TSCC)), to obtain optimum policy ( R , S ) parameters at each member of the supply chain, and inherently performing the allocation and rationing of stock over a finite planning horizon, is proposed through this paper. A new heuristic allocation and rationing mechanism for the distributor to distribute stock among retailers during the occurrence of shortage named as Priority Fractional Rationing (PFR) policy is also introduced in this study. A heuristic methodology which is a hybrid of Genetic Algorithm and Particle Swarm Optimization algorithm ( HGA-PSO ) combined with PFR policy is proposed through this study after analyzing the computational difficulty encountered and lack of tractability of stock-allocation and stock-rationing mechanism while the MILP-based mathematical formulation is solved to optimality. A local search technique as part of the Particle Swarm Optimization (PSO) algorithm, similar to mutation operation in Genetic Algorithm (GA) is introduced due to the observation of inferior results during pilot studies. The performance evaluation studies of the HGA-PSO with that of stand-alone GA, stand-alone PSO with new local search and the exact solution obtained by solving the MILP-based mathematical formulation is presented. The results indicate the superior performance of the hybrid algorithm in comparison with stand-alone GA and PSO.