A Multi-Stage Supply Chain Network Optimization Using Genetic Algorithms

In today’s global business market place, individual firms no longer compete as independent entities with unique brand names but as integral part of supply chain links. Key to success of any business is satisfying customer’s demands on time which may result in cost reductions and increase in service level. In supply chain networks decisions are made with uncertainty about product’s demands, costs, prices, lead times, quality in a competitive and collaborative environment. If poor decisions are made, they may lead to excess inventories that are costly or to insufficient inventory that cannot meet customer’s demands. In this work we developed a bi-objective model that minimizes system wide costs of the supply chain and delays on delivery of products to distribution centers for a three echelon supply chain. Picking a set of Pareto front for multi-objective optimization problems require robust and efficient methods that can search an entire space. We used evolutionary algorithms to find the set of Pareto fronts which have proved to be effective in finding the entire set of Pareto fronts.

💡 Research Summary

The paper addresses the increasingly integrated nature of modern supply chains, where firms operate not as isolated entities but as interconnected links within a network. Recognizing that timely fulfillment of customer demand is pivotal for cost reduction and service level improvement, the authors develop a bi‑objective optimization model for a three‑echelon supply chain consisting of factories, logistics centers, and distribution centers. The two objectives are (1) minimization of total system cost—including production, transportation, inventory holding, and ordering costs—and (2) minimization of delivery delays to distribution centers, measured as average lead time from order placement to receipt.

The model incorporates realistic constraints such as production capacity, transportation limits, inventory balance, demand satisfaction, and lead‑time variability, thereby capturing the inherent non‑linearity, discreteness, and uncertainty of real‑world supply chain decisions. Because traditional exact methods (e.g., linear or mixed‑integer programming) struggle with such complexity, the authors turn to evolutionary computation, specifically a multi‑objective Genetic Algorithm (GA). The GA maintains a population of candidate solutions, applies binary crossover and Gaussian mutation, and uses a non‑dominated sorting with crowding‑distance based selection—essentially an NSGA‑II‑style approach—to evolve a diverse set of Pareto‑optimal solutions. Elitism ensures that the best discovered fronts are preserved across generations.

Parameter settings (population size 200, crossover probability 0.9, mutation probability 0.1, 500 generations) are explored through sensitivity analysis. Experiments on synthetic data representing a three‑stage network demonstrate that the GA reliably uncovers a well‑distributed Pareto front, outperforming single‑objective linear programming baselines in both solution diversity (spread, hyper‑volume) and the ability to capture the trade‑off region between cost and delay. Robustness tests with demand fluctuations of ±20 % show that the GA‑derived solutions maintain stable performance, indicating resilience to uncertainty.

The authors acknowledge limitations: the model assumes static demand forecasts, simplifies resource availability to binary constraints, and the GA’s computational burden may become prohibitive for very large networks. They propose future work that includes dynamic demand forecasting, integration of local search heuristics to form hybrid meta‑heuristics, parallel implementation on cloud/edge platforms for real‑time scalability, and extension of the objective set to incorporate sustainability metrics such as carbon emissions.

In conclusion, the study demonstrates that a multi‑objective GA can effectively generate a comprehensive set of Pareto‑optimal configurations for a three‑echelon supply chain, enabling decision makers to balance cost efficiency against service responsiveness. This approach offers a practical decision‑support tool for firms seeking to navigate the complex trade‑offs inherent in modern, globally‑distributed supply networks.