Operations‑Research Methods for Large‑Scale Supply‑Chain Network Optimization

In the supply‑chain domain, operations‑optimization techniques are crucial for making optimal decisions across procurement, production, inventory, and logistics to minimize cost, maximize efficiency, and meet customer demand, including inventory level determination and risk‑scenario simulation.

A typical logistics network planning model aims to minimize total logistics cost (transportation plus transshipment). The objective function sums vehicle transport costs for each route and adds transshipment costs when a route is selected. Constraints enforce unique routing per flow, vehicle capacity, vehicle‑type selection only on opened routes, single vehicle type per route, logical relations between route opening and flow, and minimum flow requirements for opened routes.

When applied to a real case (Company C), the model contains millions of integer variables, and commercial solvers fail to find a feasible solution after ten hours. A column‑generation‑based approach combined with a partitioned solving strategy is adopted to generate a high‑quality initial solution.

Using partitioned solving, an initial solution is obtained in about three hours with a 30 % gap to the relaxed optimum, whereas direct Gurobi solving yields no feasible solution after ten hours, demonstrating the efficiency of the partitioned method.

After obtaining the initial solution, convergence remains slow; applying column generation further accelerates the solving process.

The column‑generation workflow starts with a restricted master problem built from routes in the initial solution, relaxes integer variables to continuous, solves for dual values, evaluates non‑included candidate routes using reduced costs, adds promising routes (negative reduced cost), and iterates until no improving routes remain, then restores integrality to obtain a high‑quality final solution.

Effectiveness of column generation depends on the gap between the MILP and its LP relaxation. In this problem, a large‑M formulation makes the relaxed model too loose, so a tighter constraint (z_l ≥ y_r) is introduced to improve performance.

With column generation, the massive network planning problem is solved, revealing an “hourglass” structure: three layers of routes (5‑10 primary‑to‑hub branches × 40‑60 hub‑to‑hub main lines). The optimized network averages 2.49 transshipments per parcel, requiring only a few main lines, making many existing facilities suitable.

The supply chain can be viewed as a chain of material, logistics, and information flows, with operations acting as the rivets that ensure global optimality, turning complex business processes into decision variables, objectives, and constraints for clear analysis.

Improving supply‑chain decisions across all stages—strategic planning, tactical planning, and execution—helps manage revenue, cost, and the flow of goods, information, and funds, achieving cost reduction and efficiency gains.

Strategic planning (road‑building) integrates resources and sets long‑term routes, influencing decisions over 3‑5 years and requiring careful consideration of market uncertainties.

Tactical planning (traffic control) covers production, inventory, and procurement plans over quarters to a year, focusing on real‑time market factors such as sales trends, seasonality, and supply availability.

Execution (driving) involves real‑time implementation of plans, with decisions made at hour, minute, or second granularity, monitoring production efficiency, inventory levels, and supplier performance to ensure smooth operations.

This phased optimization enables enterprises to address challenges at different time scales, combining long‑term development with short‑term responsiveness.

To cope with growing supply‑chain complexity, advanced algorithms and technologies—cloud computing, artificial intelligence, machine learning, and data mining—are employed to create intelligent supply chains capable of prediction, classification, and clustering for various practical problems.

The content is excerpted from the book Intelligent Supply Chain: Operations‑Optimization Theory and Practice , which comprehensively introduces optimization algorithms and industry cases for solving supply‑chain management problems.

At the end of the article, readers are invited to participate in an interactive giveaway where three commenters can win a signed copy of the book by leaving relevant remarks in the comment section.