Forecasting spare part demand with installed base information: A review

Abstract

Service maintenance is commonly used to extend the lifetime of capital assets, such as manufacturing equipment or heavy infrastructure. When a service part, necessary to perform the maintenance action, is required but not immediately available, the incurred shortage costs may be substantial. For this reason, companies keep large stock buffers to deal with uncertain demand of these spare parts. Specialized service parts models should therefore focus on improving the availability of parts whilst limiting the investment in inventories. An important characteristic of most service parts is their intermittent demand pattern, for which specific forecasting techniques have been developed (see e.g., the review of Boylan and Syntetos, 2010). Many of these methods, however, rely on the time series of the historical demand and do not take into account the factors that generate the spare part demand: the failure behaviour of the components, the maintenance policy, etc. We refer to these factors as the Installed Base information. In our work we provide an overview of the papers which use installed base information for forecasting future service parts demand and we develop a new model which incorporates this information. Dekker et al. (2013) define installed base information as the information on the set of systems or products for which a company provides after sales services. It can include the number of installed and serviced machines (i.e. the size of the installed base), its evolution over time, the failure behaviour of the parts, part age information, and the part replacement probability. In addition to that, it is also possible to include information on the sudden and scheduled service needs of the products. Because the maintenance policy in use has an impact on the demand of spare parts, taking this information into account will improve the predictability of service parts demand. We aim to provide a new model which uses installed base information to predict future demand. This model combines information on the maintenance policy, the size of the installed base and its evolution over time, the part failure behaviour, and the replacement probability, in order to capture the full picture of the demand generating process.