Optimizing the Machinery System Maintenance Policy Using Vibration Monitoring and Proportional Hazard Model

摘要

The time-dependent proportional hazards model (PHM) is a popular model in survival analysis. The merit of the time-dependent PHM lies that its ability to relate the failure probability to both age and condition variables, so that one can assess the failure probability with given machine condition at any specified age. The PHM can combine deterioration and failure information effectively of system or components to assemble the current reliability to determine the optional maintenance policy. Obviously, the PHM is more practically significant and attractive. The explanatory covariates in the PHM are time-dependent and its transformation can identify deterioration of system or component. Adequate covariates must be guaranteed to sufficiently reveal current condition of the component, thus, principal component analysis (PCA) can be applied to covariates (measurements) to reduce the number of variables included in the model, as well as to eliminate possible co-linearity between the covariates. A hidden Markov model (HMM) consists of two stochastic processes: a Markov chain with finite number of states describing an underlying mechanism (state transformation sequence) and an observation process depending on the hidden state (observation sequence). Due to the characteristic mentioned above, the covariate process is represented by a nonhomogeneous discrete Markov process with a finite state space. A practical example of workbench of mill-turn machining center is given to demonstrate the rationality and effectiveness of this method.