A mixture Weibull proportional hazard model for mechanical system failure prediction utilising lifetime and monitoring data

• We construct a mixture Weibull proportional hazard model to predict the failure time of a mechanical system with multiple failure modes.
• The historical lifetime and monitoring data of multiple failure modes are combined to estimate the system failure probability density and reliability.
• Monitoring data are input into the MWPHM to predict the failure time.

As mechanical systems increase in complexity, it is becoming more and more common to observe multiple failure modes. The system failure can be regarded as the result of interaction and competition between different failure modes. It is therefore necessary to combine multiple failure modes when analysing the failure of an overall system. In this paper, a mixture Weibull proportional hazard model (MWPHM) is proposed to predict the failure of a mechanical system with multiple failure modes. The mixed model parameters are estimated by combining historical lifetime and monitoring data of all failure modes. In addition, the system failure probability density is obtained by proportionally mixing the failure probability density of multiple failure modes. Monitoring data are input into the MWPHM to estimate the system reliability and predict the system failure time. A simulated sample set is used to verify the ability of the MWPHM to model multiple failure modes. Finally, the MWPHM and the traditional Weibull proportional hazard model (WPHM) are applied to a high-pressure water descaling pump, which has two failure modes: sealing ring wear and thrust bearing damage. Results show that the MWPHM is greatly superior in system failure prediction to the WPHM.