Material Requirements Planning under uncertainty

Abstract

Material Requirements Planning (MRP) is widely used to translate production plans into procurement and inventory decisions. Its performance can deteriorate in environments characterised by demand uncertainty and frequent material changeovers. In such settings, deterministic MRP models often set safety stocks for changeover materials to zero in order to minimise material obsolescence. This increases vulnerability to stockouts, minimum-order-quantity rush orders and last-minute procurement costs. This thesis investigates whether simulation-based optimisation can be used to parametrise safety stock levels for changeover materials in a deterministic mixed-integer programming MRP model to improve rolling-horizon performance in a stochastic production environment.

A framework is developed in which the manufacturer’s existing MRP optimisation model is treated as a black box and is embedded in an Optimisation with Simulation-based Iterations (OSI) structure. To better represent robustness in the MRP logic, a hard safety stock constraint is introduced through which safety stock levels are optimised for grouped materials. Demand uncertainty is modelled using historical weekly production plans, analysed through a method based on block bootstrapping and kernel density estimation (KDE). This analysis showed that nearly all observed variability in the production plans is demand-driven rather than induced by planning decisions.

The framework is tested using a Full Enumeration optimisation technique, and experimented on different simulation settings, material grouping strategies, uncertainty variants and safety stock variants. Although the proposed approach is technically feasible and yields a reproducible method for evaluating safety stock policies under uncertainty, the final optimisation does not deliver an obsolete stock cost improvement caused by the addition of safety stock. Low safety stock levels consistently performed best, suggesting that a generic safety stock policy for changeover materials is not effective under the current objective structure and framework setup. The main contribution of this thesis is methodological: it demonstrates how simulation-based optimisation can be applied to an industry-scale, black-box MRP environment. It also highlights the challenges of translating robustness into measurable system-wide gains when only obsolete stock costs are optimised. The findings indicate that future research should adopt broader objective functions and further refine material grouping and uncertainty scenarios to better capture material- and scenario-specific trade-offs.