Impact of uncertainty quantification on design: an engine optimisation case study

The method for solving design optimisation problems when some or all design variables and/or parameters are not deterministic depends on how we quantify uncertainty. Probabilistic design methods can be employed when sufficient information is available. In reality, however, we often do not have enough knowledge and/or data to conduct statistical inference. The amount of available information about the uncertain quantities may be limited to ranges of values. Possibility theory may then be employed to reformulate and solve the optimal design problem. In this paper, we use both probability and possibility theories to determine optimal values of engine characteristics for a hydraulic-hybrid powertrain of a medium-sized truck while accounting for the most significant modelling uncertainties. A worst-case optimisation using interval analysis is considered as a special case of possibilistic design. We contrast the two sets of results, draw some conclusions and discuss features of the two approaches.