This paper analyzes the value of different sources of installed base information for spare part demand forecasting and inventory control. The installed base is defined as the set of products (or machines) in use where the part is installed. Information on the number of products still in use, the age of the products, the age of their parts, as well as the part reliability may indicate when a part will fail and trigger a demand for a new spare part. The current literature is unclear which of this installed base information adds most value – and should thus be collected – for inventory control purposes. For this reason, we evaluate the inventory performance of eight methods that include different sets of installed base information in their demand forecasts. Using a comparative simulation study we identify that knowing the size of the active installed base is most valuable, especially when the installed base changes over time. We also find that when a failure-based prediction model is used, it is important to work with the part age itself, rather than the machine age. When one is not able to collect information on the part age, a logistic regression on the machine age might be a valuable alternative to a failure-based prediction model. Our findings may support the prioritization of data collection for spare part demand forecasting and inventory control.

本文分析了不同来源的安装基础信息对备件需求预测和库存控制的价值。已安装基础定义为安装部件的使用中的产品（或机器）集。有关仍在使用的产品数量、产品使用年限、部件使用年限以及部件可靠性的信息可能表明部件何时会出现故障并引发对新备件的需求。目前的文献不清楚哪些已安装的基础信息为库存控制目的增加了最大的价值，因此应该收集。出于这个原因，我们评估了八种方法的库存绩效，这些方法在其需求预测中包含了不同的安装基础信息集。通过比较模拟研究，我们发现了解活跃安装基础的规模最有价值，尤其是当安装基础随时间发生变化时。我们还发现，当使用基于故障的预测模型时，重要的是使用零件年龄本身，而不是机器年龄。当人们无法收集有关零件年龄的信息时，机器年龄的逻辑回归可能是基于故障的预测模型的一种有价值的替代方法。我们的发现可能支持对备件需求预测和库存控制的数据收集进行优先排序。当人们无法收集有关零件年龄的信息时，机器年龄的逻辑回归可能是基于故障的预测模型的一种有价值的替代方法。我们的发现可能支持对备件需求预测和库存控制的数据收集进行优先排序。当人们无法收集有关零件年龄的信息时，机器年龄的逻辑回归可能是基于故障的预测模型的一种有价值的替代方法。我们的发现可能支持对备件需求预测和库存控制的数据收集进行优先排序。