Çok ölçütlü eniyi bileşen sınama planlarının tasarlanması

Testing the system as a whole might be found economically infeasible or physically impossible in many cases. For instance, testing a nuclear device is currently banned by international agreements or testing a space shuttle might be found too risky because of its financial consequences. Instead, various components can be tested separately to meet some desired level of performance for the whole system, while achieving a minimal total testing cost. This problem is called system based component testing problem and is investigated in this thesis.Although system reliability is the only considered measure in the available literature on this problem, there exist other more practical measures such as mean time to failure or availability worth to take into account. Here we extend previous studies by incorporating various performance measures separately or jointly.The problem is formulated as a semi-infinite linear programming problem, and the optimum component test times are obtained by combining the well-known cutting plane method with the well-known column generation technique. The columns are generated by solving two different optimization subproblems which are proved to be d.c. (difference of two convex functions) programming or signomial geometric programming problems depending on the performance measures examined. These subproblems are solved to optimality by adapting an outer approximation method and by a special branch and bound technique. Several numerical examples are provided to illustrate the approach.