A Methodology Research On Robust Design Optimization Considering Aleatory And Epistemic Uncertainty

As an engineering design theory which can effectively improve the quality of productions, robust optimization design technique has been demanding crucially and gaining wide attention in academic community. The basic concept of robust design is to ensure the product performances to be insensitive to uncertainties existing in the design rather than eliminating the uncertain factors. Even though the traditional methods of robust optimization design take into account the aleatory uncertainty of design variables, the ignoring the epistemic uncertainty in calculation and design analysis process makes the estimation of performance variation incomprehensive and the criterion for optimization iteration inaccurate, all of which directly affect the robust solution in turn. Therefore. it is of great value in theoretical and engineering aspects to explore a robust optimization design method which can handle both epistemic and aleatory uncertainty together.In this study, the conventional robust optimization design concept was extended. A robust design optimization theory and its corresponding technical means were developed, in which not only aleatory uncertainty, but also epistemic uncertainty can be tackled more comprehensively and the robust optimum solution would be more consistent with the actual situation. The main contents are as follows:Firstly, the traditional robust design optimization theory was extended, a robust design optimization framework orienting aleatory uncertainty and epistemic uncertainty was proposed. which is short for augmented robust design optimization (ARDO). The process of ARDO consists of design modeling, establishing surrogate model, optimization iteration and final optimal solution. In optimization iteration, both kinds of uncertainty were quantified together and the design variables were changed to make the robustness evaluation index satisfy the iteration termination prerequisite. The corresponding optimized design model, implementation process and key technology of the ARDO framework were put forward.Secondly, in order to quantify multiple uncertainties in the ARDO, based on the current technique on quantification and propagation of the aleatory uncertainty and the characteristic of epistemic uncertainty, evidence theory was adopted to represent both aleatory uncertainty and epistemic uncertainty and the sampling method, interval arithmetic or the optimization method was utilized to propagate both types of uncertainty to obtain the evidence structure of performance response evidence space.Thirdly, according to the evidence structure of response evidence space, two kinds of robustness evaluation criterions taking in account both types of uncertainties were presented to assess the robustness in ARDO. One criterion based on the aggregated interval median and deviation of performance response, the other based on belief functionBel(Q) and plausibility function Pl(Q), in which Q represents the proposition that the performance response is eligible.Finally, this article took a typical structural system composed by bearing bracket, vibration isolator and function component as an example to demonstrate the realization process of the presented approach in detail.