Modeling and optimization of maintenance policies for computers

The performance of a computer deteriorates over time. It is important that the computer performance stays above a required minimum performance level, Pmin, in order not to affect the working efficiency. Therefore, a good maintenance policy is necessary. In this thesis we determine the optimal maintenance policy for computers for a period of T years, in the sense of guaranteeing a computer performance above or equal to Pmin at minimized costs. We assume that there are 3 types of computers available and that each type of computer consists of 4 different components. Because a computer performance mainly depends on the performance of its components, we first model the component performance. Study showed that the components exponentially deteriorate [5], which is therefore taken into account. We propose to model the computer performance at time t as the average of all component performances at time t. To determine the optimal maintenance policy, two types of policies are implemented, namely deterministic and stochastic policies. Each type of policy considers replacing components, which is called updating, or replacing the computer. The deterministic policies execute maintenance in a fixed order, while the stochastic policies determine a maintenance policy with a decision rule. We propose 2 decision rules: the performance – cost ratio and the lifetime extension – cost ratio. Stochastic policies in general provide better maintenance policies than deterministic policies. Therefore, we determine which decision rule delivers the optimal maintenance policy to make a recommendation which type of computer the best option to buy is. At last sensitivity analysis is implemented to test whether the choice of the optimal decision rule changes when the settings change.