Available at: https://digitalcommons.calpoly.edu/theses/3403
Date of Award
6-2026
Degree Name
MS in Industrial Engineering
Department/Program
Industrial and Manufacturing Engineering
College
College of Engineering
Advisor
Byeongmok Kim
Advisor Department
Industrial and Manufacturing Engineering
Advisor College
College of Engineering
Abstract
To navigate severe supply chain disruptions and volatile market dynamics where historical data distributions fail to predict sudden environment shifts, traditional frameworks, such as reorder point (s,Q) policies are often inadequate in managing inventory. This thesis develops and evaluates an advanced simulation-based optimization framework to determine inventory control variables under severe, uncertain demand and bimodal lead-time conditions. Using Python-based discrete-event simulation, a reinforcement learning (RL) proximal policy optimization (PPO) is constructed and tested against traditional analytical baseline calculation, and two metaheuristic algorithms – simulated annealing (SA) and Genetic Algorithm (GA). The model specifically incorporates a highly uncertain demand environment and multi-component lognormal lead-time mixtures to simulate the vendor capacity losses and sudden demand surges frequently seen in high-tech manufacturing sectors. The experimental results demonstrate that while traditional metaheuristics offer marginal adjustments to the baseline policy, a reinforcement learning framework dynamically adjusts to changes in the system – significantly reducing total operational costs and backorder penalties while maintaining target stock levels. This research contributes an isolated, robust, inventory control agent capable of managing material planning in a highly volatile procurement environment.
Included in
Industrial Engineering Commons, Operational Research Commons, Other Operations Research, Systems Engineering and Industrial Engineering Commons