Censored Field Data: Impact and Consideration on Reliability Analytics in Product life cycle - A Case Study on Light Electric Vehicles

The technical reliability of components in technical systems and products has a significant impact on the long-term reliability of products and thus on the sustainable use of energy and resources. Long-term reliability of components means that the product needs to be replaced later and that the components are potential candidates for carry-over parts (COP) regarding the next product generation, and can also be used after refurbishment as part of a second life phase (e.g. fair value repair). However, this requires a precise Model of the failure behaviour respectively the technical reliability of the component. The technical reliability of a component or product can be mapped using failure models based on field data (e.g. operating data, load collectives, damage data). In most cases, censored field data is a challenge and must be taken into account. Depending on the degree of censoring, failure models are subject to uncertainty regarding the expected failure rate or probability of failure. The study which is presented in this paper examines the impact of various failure models on the example of predictive product maintenance cycles based on censored field data. The aim is to compare commonly used prediction models and determine their effects on component reliability predictions. In a first step, various methods for modelling failure behaviour in the case of censored data (Kaplan-Meier estimator, Johnson method, Eckel candidate method) are compared. In a second step, the influence on an exemplary predictive maintenance scenario is determined. The study is conducted using a mechanical transmission component from a light electric vehicle (LEV), for which field data from an LEV fleet in the usage phase was recorded over two years of operation.

Keywords: Light Electric Vehicle Case Study, Censored Field Data, Weibull Distribution, Johnson Method, KaplanMeier Method, Eckel Method.