Shelf Life Management System for Military Aircraft Components

Ng Choon Leong

Abstract

In the airforce, aircraft and missile component manufacturers often stipulate the shelf life of the components. This is to ensure that such items remain serviceable in storage before being used. In the event that the manufacturers do not specifically state the shelf life, problem such as the stability of the product arises. Hence, the objective of this project is to explore the methodologies adopted by the industries that are used to study shelf life.

Two related industries are selected. These are the food science and technology, and the pharmaceutical industries. For the food industry, accelerated shelf life testing (ASLT) and the microbiological predictive models are adopted. The ASLT is especially practical when the actual storage time is long. The basic premise of the ASLT is by changing a storage condition, the chemical or physical process that leads to deterioration is accelerated, and that a predictive shelf life relationship related to ambient conditions can be defined. On the other hand, microbiological predictive model is used for predicting the rates of deteriorative change by inoculating one or more microorganism into the test medium and observing its behaviors in the presence of intrinsic and extrinsic factors. The pharmaceutical industry, similar to the food industry, also uses the ASLT method. However, the pharmaceutical industry will have to conduct a long-term stability test to support the claim of the ASLT.

The second objective of this project is to propose a general approach to shelf life study. This general approach consists of six steps. These are the product evaluation, experimental design, data collection, predictive models, model validation and software systems. In the product evaluation stage, areas such as the product characteristics, the influencing factors, and the mode of deterioration are taken into consideration. In the design of experiment, two commonly used designs for the study of shelf life are proposed. These are the 2n factorial design and the fractional factorial design. For collection of data, techniques used include physical measurement, chemical measurement, instrumental method or even measures taken from sensory panels. The predictive models (Arrhenius equation, Eyring equation, etc.) then relate the accelerating variables to the time acceleration.

Lastly, a simple Bayesian network (BN) of a hydraulic hose using Hugin Lite 5.7 is built and modeled. Prior information based on the author's past experience in the Air Force is taken into account. From this prior information, the effect of the warehouse and storage condition on the shelf life of the hydraulic hose is then computed. Finally, the backward propagation of the Bayesian network is accomplished based on the desired shelf life of a product.