AbstractScheduling and inventory control problems have been amongst research topics on designing and optimising Supply Chains (SCs) which have attracted a very high interest to both academics and practitioners for many decades. Without a doubt, the body of scientific literature and the real-world applications deliver a proof that these problems hold the interest not only of researchers and mathematicians, but also industry practitioners which deal with scheduling and inventory control decisions every day. Interest in these problems have resulted in many state-of-the-art mathematical models programmed to find optimal solutions and simulation frameworks allowing to observe parameters of complex environments in real time manner. However, these models very often do not consider uncertainties which are inherent in SCs.
The goal of this research is to create a new model supporting decisions for coordinated inventory and scheduling problems in a dynamic SC environment facing uncertainty in demand. A control scheme using fuzzy logic for modelling uncertainty was developed for a four echelon SC including Suppliers, Manufacturer, Distribution Centre and Customer. Fuzzy sets enable use of expert knowledge which allows representation of imprecise or vague data. The new method developed in this research proposes the decision support system which determines scheduling of orders by prioritising jobs to the resources available to the specific echelon with simultaneous determination of replenishment levels and order quantities of required raw materials. The objectives are to minimise the total holding cost of the inventory along SC and to minimise the delays of orders delivered to the customer. Simulation-optimisation approach was employed to test knowledge extraction capabilities of the proposed models, aiming to propose a robust control-schemes which is less sensitive to the changing demand. Four control-schemes were developed. Crisp dispatching rules (DRs) and two sets of fuzzy dispatching rules (FDRs) were used to provide inventory and scheduling control. First set of FDRs were delay-focused, so higher inventory levels were kept by echelons to quickly satisfy the demand, the second set were holding cost-focused FDRs where inventory levels were kept lower to minimise holding cost of additional stock. To determine the optimal control the search process was guided by NSGAII and to increase the robustness of the model, a Monte Carlo simulation was conducted within NSGAII creating MCNSGAII control scheme. A benchmark scenario and a number of experiments with varying due dates, order sizes, processing times and order intensities were carried out. The results obtained are analysed and provide an insight into SC performances with uncertainty in demand and changing SCs parameters. Non-information sharing policy between echelons was employed and varying order intensity was simulated multiple times in various scenarios to test the proposed control schemes.
FDRs for both subproblems including inventory control and scheduling outperformed standard policies based on continuous review policy (CRP) and crisp DRs for scheduling. Uncertainty in both; demand for both subproblems was addressed by applying a multi-objective optimisation. NSGAII performed better than both manually determined FDRs leading to a decrease in the delay in delivering orders to customers by 66% in comparison to delay-focused FDRs, while keeping a very similar holding cost level.
Rule bases generated by MCNSGAII led to improvement of both objectives by capturing dynamics of changing demand offering robust solutions with a low standard deviation from the average objectives’ values. A further decrease of the average holding cost by 8.2% and the average delay by 5.2% were also observed comparing to the standard NSGAII. The novel developed methodology displays robustness of solutions and success in making trade-offs between holding cost and delay offering an independent and flexible control for both scheduling and inventory control problems across multiple echelons.
|Date of Award||Jun 2021|
|Supervisor||Dobrila Petrovic (Supervisor) & Ammar Al Bazi (Supervisor)|