There is a world-wide drive to optimize maintenance decisions in an increasingly competitive manufacturing industry. Preventive maintenance if often the most organized and cost efficient strategy to follow, but a decision still has to be made on the optimal instant to perform preventive maintenance. Use based preventive maintenance decisions have been optimized through statistical analysis of failure date while predictive preventive maintenance (condition monitoring) has been optimized by utilizing more sophisticated technology. Very little work has however been done to combine the advantages of the two schools of thought. This thesis originated from a realization of the potential improvement in maintenance practice by combining use based preventive maintenance optimization techniques with high technology condition monitoring. In this thesis an approach is developed to estimate residual life of industrial equipment dynamically by combining statistical failure analysis and sophisticated condition monitoring technology. The approach is based on failure intensity proportions determined from historic survival time information and corresponding diagnostic information such as condition monitoring. Combined Proportional Intensity Models (PIMs) for non-repairable and repairable systems, containing the majority of conventional PIM enhancements as special cases, with numerical optimization techniques to solve for the regression coefficients, are derived. In addition to the residual life estimation approach, a user-friendly graphical method with which residual life estimates can be presented was also developed. This method is natural and easy to comprehend, even by inexperienced data analysts. The residual life estimation approach is applied to a typical data set from a South African industry and results are compared to those obtained from a similar, established maintenance decision support tool. This comparison showed that the approach developed in this thesis is relevant, practical and marginally better than the established decision support tool for certain criteria.
Thesis (PhD (Industrial Engineering))--University of Pretoria, 2006.