Inventory Decision Model of Single-echelon and Two-indenture Repairable Spares

Chenyu Liu, Bin Zhou, Feng Guo, Suqin Zhang2012 · DOI

Abstract

According to the three conditions that the variance-to-mean ratio of the spares demand is larger than, equal to or less than 1, put forward the three demand distributions such as the Negative binomial distribution, Poisson distribution and Binomial distribution. For the problem that the two-level materials demand follows the Poisson distribution, use the Negative binomial distribution to improve the forecast accuracy. In the given total security funds as constraint conditions, through total shortfall minimum instead of supply availability maximum to simplify the objective function, establish the spares inventory decision model, and optimized by the marginal analysis and compile algorithm. The example proves that the model has good accuracy and practicality.

Introduction

In aerospace logistics, managing "repairable spares"—high-value components that are refurbished rather than discarded—is a critical balancing act. Insufficient stock leads to grounded aircraft (backorders), while excessive stock ties up millions in capital. Traditionally, these systems rely on the Poisson distribution, assuming the average demand equals its variance. However, real-world data often shows "bursty" or highly predictable demand patterns that standard models fail to capture.

This paper proposes a more flexible framework that selects between Negative Binomial, Poisson, and Binomial distributions based on the variance-to-mean ratio ($V$). Furthermore, it addresses the "two-indenture" problem, where the availability of a main component (Line Replaceable Unit, or LRU) depends on the availability of its sub-components (Shop Replaceable Units, or SRUs).

Methodology: Distribution Selection and Two-Indenture Modeling

The core of the proposed model is the selection of the demand distribution based on the variance-to-mean ratio ($V$):

  1. $V > 1$ (High Variance): The Negative Binomial distribution is used. This often occurs for LRUs because shortages in SRUs add delays and "noise" to the repair pipeline.
  2. $V = 1$ (Standard): The Poisson distribution is applied.
  3. $V < 1$ (Low Variance): The Binomial distribution is used for parts with more deterministic wear-out patterns.
Table 1: Parameters for the five spare parts used in the case study, including the variance-to-mean ratio ($V$).

As shown in Table 1, the model accounts for the hierarchical relationship between parts. For instance, LRU2 is linked to SRU1 and SRU2. If an SRU is out of stock, the LRU cannot be repaired, effectively increasing its "pipeline" time and shifting its demand distribution toward a Negative Binomial shape.

Optimization Strategy

To optimize the inventory, the researchers focused on minimizing the Total Expected Backorders (EBO) under a fixed budget constraint. They employed a marginal analysis algorithm, a greedy optimization technique that iteratively allocates funds to the spare part providing the greatest reduction in total shortage per dollar spent.

Table 2: Iterative optimization results using marginal analysis to minimize total shortage under budget constraints.

Table 2 illustrates this process. As the budget ("total costs") increases, the algorithm selectively increases the stock levels ($s_i$) of specific LRUs and SRUs to drive down the "total shortage."

Key Results and Practical Insights

The model was validated using a 10-year historical dataset (2000–2009) for a fleet of 24 aircraft. Key findings include:

  • High Predictive Accuracy: The model predicted a supply availability of 92.78% for a $90 million budget. Real-world performance in 2010 showed a "normal" availability of 91.84%, representing a relative error of only 0.94%.
  • The Poisson Trap: Relying solely on Poisson distributions led to an overly optimistic prediction of 99.89% availability, which would have resulted in significant under-stocking in practice.
  • Diminishing Returns: The study suggests that aiming for near-100% availability through inventory alone is cost-prohibitive. A more efficient strategy is to target ~91% availability and handle the remainder through emergency orders or part cannibalization.

Conclusion

By moving beyond the standard Poisson assumption and accounting for the LRU/SRU indenture relationship, this model provides a highly accurate tool for aerospace inventory management. While it currently assumes single-echelon logistics and no part cannibalization, its performance in real-world case studies demonstrates significant practical value for optimizing multi-million dollar spare parts budgets.

Summary updated

Abstract

According to the three conditions that the varianceto-mean ratio of the spares demand is larger than, equal to or less than 1, put forward the three demand distributions such as the Negative binomial distribution, Poisson distribution and Binomial distribution. For the problem that the two-level materials demand follows the Poisson distribution, use the Negative binomial distribution to improve the forecast accuracy. In the given total security funds as constraint conditions, through total shortfall minimum instead of supply availability maximum to simplify the objective function, establish the spares inventory decision model, and optimized by the marginal analysis and compile algorithm. The example proves that the model has good accuracy and practicality.
BibTeX
@misc{liu2012inventory,
  author = {Chenyu Liu and Bin Zhou and Feng Guo and Suqin Zhang},
  title = {{Inventory Decision Model of Single-echelon and Two-indenture Repairable Spares}},
  year = {2012},
  doi = {10.2991/iccasm.2012.61},
}
Figure