Text

theses

A new modeling approach has been developed to optimally design a degradable system structure and maintenance plans for applications exposed to distinct stresses under different environments and diverse usage conditions. When component stresses change or shift in the future and diverse operating conditions, equipment or components may not be operating with the same stress profile that was presented when the historical data was collected and analyzed. A more detailed mathematical perspective is described to analyze the uncertainty of actual system usages in future scenarios that take variations and uncertainties explicitly into consideration. The integrated development of cost-reliability optimization models and a system maintenance policy is an interesting and challenging problem. A new integrated system optimization planning model is developed and applied to provide an insightful mathematical representation of the process that more directly represents the system reliability design process. A four-stage optimization model is constructed to accommodate sequences of decisions over time with random future usage scenarios. In the first and second stage, a two-stage stochastic model with recourse is formulated with a system cost, reliability and maintenance combined objective function to determine an initial design structure and corrective or preventive maintenance intervals. In the third-stage, the system is fielded and data is collected and analyzed to update model parameters and coefficient estimates, and adaptive preventive maintenance optimization is performed. Bayesian posterior distributions are constructed to provide the model with timely and improved estimates to reflect the fielded data. In the fourth-stage, a cost saving strategy is implemented to decide whether the current system design or a new or revised system design can provide sufficient cost savings beyond a threshold to justify design changes. The optimization models are expressed as nonlinear stochastic integer programming models where several tools are used for solving problems. In particular, pattern-search (from MATLAB toolbox), iterative coordinate descent and neighborhood heuristic search are the methods used for searching the feasible region to obtain optimal or recommended model solutions. Theoretical and simulated examples are solved to demonstrate the modeling approach and results.

ETD_6229

Ph.D.

Includes bibliographical references

by Nida Chatwattanasiri

doi:10.7282/T3319XQC

ETD doctoral

The author owns the copyright to this work.