Multicriteria optimisation in design for reliability

A novel methodology for series-parallel systems’ reliability optimisation has been proposed developed and tested. It formulates the problem as a multi-criteria optimisation, to maximise the subsystem reliabilities while minimising the system cost modelled as a penalty function of component reliabilities, with lower bound constraints on the subsystems’ reliabilities. The goal was to find the Pareto optimal component reliability values that at least yielded a system reliability target. This problem usually occurs at the system design stage. The resultant continuous optimisation model was solved by the Weighted Sum method. The methodology was applied to hypothetical problems and to cases derived from published work concerned with life support and electricity transmission systems’ reliability. It was also tested on a gas transmission system. A comparison of the results with those for a single criterion optimisation model of the life support system indicated that higher reliability could be generated for the components/system under the methodology. The relative levels of the component reliability values were consistent with those achieved under the single criterion formulation. The reliability values allocated to a component was also consistent with their importance. The cost/penalty increased with increase in component reliabilities, becoming indeterminate as component reliability approached its maximum value.