Reliability Prediction using the Cox Proportional Hazards Model

Currently, for a variety of mechatronic systems and components, sufficient failure behaviour data are not available. Endurance tests at customer-specific operating conditions provide manufacturers with specific failure time data. However, they are timeconsuming and expensive. Findings gained through experiments are valid only for the applied test conditions and loads. On the other hand, developers require, as early as possible, meaningful key figures characterizing the applied components to determine the overall reliability. Often, modified components using the same technology basis are applied with other load profiles, so that available test data can not be used without further steps. Alternatively, one can try to derive sufficiently precise predictions for newly developed components or new application environments from a variety of existing data sets from endurance tests of similar components and other load cases. To this end, well-known regression models of survival analysis have been developed further. To illustrate the transferability to applications for reliability prediction, test data of DC motors from inhouse experiments and simulated data sets are adapted to a Cox proportional hazards model. Index Terms – Reliability prediction, Cox proportional hazards model, DC motors, mechatronic systems, regression