Statistical methods for lifetime verification and failure rate analysis

Abstract: In the powertrain of a heavy vehicle, several sensors and electronic control units are mounted. These components need to have a long lifetime and high reliability. The quality is secured by robust design and is verified by different tests. After production launch the product quality can be investigated using warranty data and repair and maintenance (R\&M) data. In this study, both warranty data and R\&M data from Scania are used. The warranty data covers all claims during the first year for analyzed vehicles in any country. A R\&M contract can be signed in some countries and may extend for varying number of years. From the R\&M data, information for six European countries are used. This study focuses on investigating the NOx sensor which measures nitrogen oxides in the exhaust gas. The aim is to investigate patterns in the failure behavior depending on different factors that may affect the failure. Factors chosen for the analysis are vehicle type, exhaust outlet direction, country and the engine stroke volume. By a descriptive analysis using warranty data an overview over the claims is found. A survival analysis using the R\&M data is done. This is due to that the R\&M data covers claims over a longer period than one year. A regression analysis is used to investigate which factors of the chosen ones that affects the sensor the most. The descriptive analysis shows that claim rates differs a lot between countries. It can be due to various aspects, for example differences in knowledge about the NOx sensor in different workshops. Or the differences reflect real variations of the failures. There is also differences between the engine types and between the two vehicle types haulage trucks and construction vehicles. The results from the parametric regression analysis shows that different stroke volumes and countries affect the failure pattern the most. A distribution model including multiple failure modes seems to fit the data best. This imply that there is not one single failure mode that most failures arise from. The failure rate seems to increase over time meaning that most failures are due to wear-out. It is important to understand the data used. Aspects like if claims always are reported in all countries have to be taken under consideration. Parts of the data might have to be omitted. Knowledge about the data is important if one wants to be able to draw conclusions from the analyzes of the claim rate of components.