Vehicle reliability field data analysis — Best practise at Mercedes-Benz cars

Summary form only given. Product reliability has ultimately to be proven under real life stress in the field. Due to weakness of the design, unforeseen vehicle driving profiles or unsteady production processes, components will fail unexpectedly early in vehicle lifetime. For a quality oriented company like Daimler it is mandatory to find and eliminate rapidly the root cause of the underlying failure mechanism, and, to forecast as soon as possible the upcoming number of affected vehicles (up to 10 years). But, other than in product validation fleets, there is no complete information above time in service and driven mileage for each vehicle in the field. So missing values have to be estimated and the fact of censored data has to be taken into account for all mathematical models. Often a clear seasonal impact of the failure occurrence can be observed and this has to be included in the forecasting model. To optimize preventive field repair scenarios, one is interested to estimate the probability that a specific vehicle will fail during the next weeks.