Author: Albert Frank Lowas (III)

Publisher:

Published: 2015

Total Pages: 316

ISBN-13:

Part demand forecasting methods assume that the demand for a part over time follows a predictable pattern, and that the patterns observed in historical data provide a reliable indication of future demands. Generally, forecasting studies focus on topics such as: the span of time from which to sample the historical data, an assessment of data in order to find weekly or annual patterns, and the assignment of probabilities of different demand quantities in any given time period. Using the demand models derived from these forecasting methods, inventory decisions are made--decisions which directly impact operating cost and equipment availability. Like most general part demand forecasting methods, aircraft spare part demand forecasting considers historical trends in order to predict future demand. It is a well-known practical observation that aircraft spare part demands are often very erratic (quantity variability), intermittent (variable in timing), and otherwise unpredictable. However, contemporary science does not explain the causes of these variations, and suffers from very poor forecasting accuracy. The objective of this research is to study the likely causes of the variations in demand quantity and from that understanding to develop forecasting methods which are more appropriate for the wearout characteristics and high reliability of many aircraft parts. As a first look at the problem, models of part failure are developed. These models are used to simulate multiple simultaneous parts operating identically. The simulations found that aircraft spare parts demands tend to be lumpy, and that this lumpiness tends to consist of two parts: a random element (called noise), and a cyclic element (called signal). These simulation results are compared to existing aircraft spare parts demand data, and similar lumpy characteristics are identified. The research then more deeply understands these elements of spare part demand lumpiness by developing equations explaining this lumpiness. These equations find that the same factors (quantity of parts operating simultaneously and reliability of those parts) both impact the average demand interval and the demand coefficients of variance, and that they impact these demand characteristics so similarly that demand lumpiness should be expected. Having determined that lumpiness is to be expected, the research proceeds to find forecasting methods that best account for this lumpiness. It is theorized that no one forecasting method would best account for signal lumpiness, noise lumpiness, and smooth demands; thus, the study develops a heuristic to select the best forecasting method based upon key part characteristics (reliability and quantity). The forecasting heuristic development uses Monte Carlo simulations to find ranges of part characteristics for which certain forecasting methods and parameters are most likely to provide the lowest error forecasts. Developing this forecasting method selection heuristic uncovers additional new and unique information, as follows: - The best error in many cases is 100% error, showing the need to move beyond forecasting for inventory management of many parts. - The forecasting error computation method used by the forecasting professional strongly influences the selection of the best forecasting method. - Certain elementary forecasting methods (e.g. naive or always zero) produce lower errors than any complex methods in some aircraft parts management conditions. - The selection of forecasting method parameters is as important as the selection of forecasting methods. This dissertation makes an important and unique contribution to the science of aircraft spare parts forecasting in creating a method to develop heuristics to select the lowest error forecasting methods. However, this dissertation makes a simultaneously important contribution in developing the inherent limits of forecasting accuracy, and finding the ranges of conditions (part reliability and quantity in service) under which no forecasting method will be effective. This study develops a new and unique understanding necessary for aircraft spare parts management.