Reliability analysis and maintenance scheduling considering operating conditions

Abstract: In reliability analysis failure times are often considered as the only factor that can explain the reliability characteristics of a system. This may be insufficient. Operating conditions and other factors (e.g. humidity, dust and vibration) may also influence the reliability of a system. All those factors that may influence the reliability of a system are referred to as covariates or explanatory variables. The reliability characteristics of a system can be explained in a better way by considering the effects of covariates. Identification of the effects of covariates on reliability may help in the selection of a suitable design of a system under given operating conditions and in controlling the critical covariates for optimising the performance of a system. This thesis deals with reliability analysis and maintenance scheduling considering the effects of covariates. A brief survey of available reliability models for analysing the effects of covariates is presented. Among the many non-parametric models, the Proportional Hazards Model (PHM) and some of its extensions appear to be suitable models for practical applications. Some of the research work performed in the area of the PHM is summarised. Some of the advantages and drawbacks of the PHM are also discussed. The PHM has been used for analysing the effects of covariates on the reliability characteristics of the power supply cables of electric mine loaders which are used in underground mines. The effects of covariates such as fault types, cable types and the number of repairs done are identified. The PHM and some of its extensions are useful in analysing the effects of covariates under different assumptions about the failure process of repairable systems. A case study for analysing the effects of time-dependent covariates has been carried out using a graphical method based on a linear regression model. The linear regression model should be used as a supplement to the PHM for checking the variation in the effects of covariates over time. The effects of time-dependent covariates may be misinterpreted in the PHM. A continuous estimator for the cumulative baseline hazard rate in the PHM is suggested. It is based on the piecewise exponential estimator. This estimator is useful in the case of a small sample size and highly censored data. The thesis discusses approaches for maintenance scheduling of a system considering the effects of covariates under the age replacement and the block replacement policies. New applications of some graphical methods in maintenance scheduling considering the effects of covariates are suggested. These methods have been considered earlier based on failure times only. The application of a graphical method called total time on test plot is illustrated with an example. Approaches used in the graphical methods for maintenance scheduling are extended for estimating the threshold value of a monitored parameter.