Spare parts planning with a vision: Probabilistic life expectancy forecast
Predicting the service life of a vehicle component enables manufacturers to plan the future demand for spare parts in good time, to control developments in a problem-specific way and to estimate expected warranty costs more precisely. In particular, the exact prediction of the necessary parts reserves enables a significant cost reduction, as unnecessary storage costs or a far disproportionately expensive subsequent production can be avoided.
Technology and methodology
Vehicle components are exposed to a certain risk of failure due to aging effects. The level of this risk depends on the vehicle-specific aging and the component’s failure statistics. For probabilistic life expectancy, both the vehicle-specific mileage and the failure statistics are mapped with the aid of Bayesian models. Finally, both probabilistic models are merged to predict the failures.
The data used for the forecast must only contain the mileage and the status repair event yes/no for each vehicle. They are available to manufacturers in the form of workshop data.
• The process is available as Software as a Service (SaaS) in the Microsoft Azure cloud but can also be used in another form.
• Efficient implementation with short computing times of only a few minutes, even if a minimum configuration of compute resources is used
• For example, even if large fleets of more than 1 million vehicles are used, the forecast will be completed in less than 2 hours.