TL;DR
This paper tackles the complex problem of optimizing planned lead times in multi-level assembly systems where component lead times are uncertain. The authors propose a novel mathematical model and a Branch and Bound algorithm to minimize total inventory and backlogging costs, offering a practical approach for contract negotiation in assembly-to-order environments.
What Problem Does This Paper Address?
In today's globalized marketplace, supply chains are increasingly complex and vulnerable to disruptions, particularly due to uncertain lead times. For assembly-to-order (ATO) systems, where products are assembled only after customer orders are received, accurately planning lead times is crucial for efficient operations and customer satisfaction. Traditional approaches often simplify these uncertainties by treating lead times as deterministic, which can lead to suboptimal decisions and increased costs. This paper addresses the challenge of optimizing planned lead times in multi-level assembly systems where lead times for individual components are stochastic, and the assembly structure involves multiple tiers. The authors aim to provide a method for determining optimal order release dates for components at the final assembly level to minimize expected total costs, considering both inventory holding and backlogging costs.
Methods
The core of this paper lies in its novel mathematical model and optimization algorithm. The authors propose a general probabilistic model for multi-level assembly systems with stochastic lead times. A key innovation is the development of a recursive function to calculate necessary probability distributions. For optimization, they introduce a Branch and Bound algorithm. This algorithm is designed to work without restrictive assumptions on costs that were present in prior work, making it more general. The paper also introduces techniques to reduce the search space for the optimization problem, including decomposing the multi-level system into linear chains and utilizing properties of the Newsboy model. They also develop methods for calculating upper and lower bounds within the Branch and Bound framework, considering vertical and horizontal dependencies between nodes in the search tree.
Datasets
Not discussed in the paper. The paper focuses on a theoretical model and uses a numerical example for illustration and testing.
Key Insights and Ideas
- Decomposition for Tractability: Breaking down a complex multi-level assembly system into simpler, linear supply chains is a powerful strategy for managing complexity and enabling optimization.
- Lifting Cost Restrictions: The proposed model's ability to handle arbitrary cost structures, rather than relying on specific relationships between component and finished product costs, makes it more broadly applicable.
- Newsboy Model Adaptation: Applying the principles of the discrete Newsboy model to determine optimal order release dates for components at the final level is an elegant way to leverage existing optimization theory for this specific problem.
- Impact of b/r Ratio: The ratio of backlogging cost (b) to inventory holding cost (r) for the finished product significantly influences the search space and the effectiveness of optimization techniques, highlighting the importance of cost structure in supply chain design.
- Robustness to Lead Time Variance: The study demonstrates that the expected total cost is relatively insensitive to moderate variations in lead times, suggesting a degree of robustness in the proposed optimization approach.
Key Results
The paper presents a Branch and Bound algorithm that effectively minimizes the expected total cost for multi-level assembly systems under lead time uncertainty. Numerical experiments show that the proposed method, particularly when combined with the reduced space of research (RSR) technique, significantly reduces the computational effort required to find optimal solutions. For instance, in some cases, the RSR technique can reduce the search space by over 99%. The study also quantifies the impact of the b/r ratio on the efficiency of the algorithm, showing that for high b/r ratios, the Branch and Bound procedure is highly effective. The analysis of lead time variance indicates that the expected total cost is not overly sensitive to variations within a reasonable range (e.g., ±200%).
Limitations and Future Work
The authors acknowledge that their model is based on a one-period planning horizon and assumes infinite assembly capacity at all levels. While the approach is efficient for small to medium-sized problems, they suggest that for systems with a very large number of components and levels, metaheuristics might be necessary to find good approximate solutions within reasonable timeframes. Future work could involve coupling the analytical method with a genetic algorithm and extending the model to multi-period planning scenarios, which would be crucial for applications like MRP parameterization. Integrating demand uncertainty into the model is also identified as a potential area for future research.
Summary updated