Research On Theory And Application Of Data-driven Engineering Design Under Uncertainty

It is widely recognized in the domain that uncertainty influences the process and results of engineering design evidently.Besides,uncertainty influences the evolution and development of design methodologies profoundly.But its universality,diversity and complexity properties promote deeper research on uncertainty design theories.At the same time,various data are involved in an engineering design process.However,the role of data in dealing with uncertainty has to be further investigated.In this dissertation,based on the two most important concepts of uncertainty and data in engineering design,we propose the theory of data-driven engineering design under uncertainty.Its application to practical engineering design problems is also studied.The research achievements are summarized as follows:First,a theoretical hierarchy for data-driven engineering design under uncertainty is constructed,including the core of the whole theory,two supporting branches and two basic elements.The core of the whole theory is data-driven engineering design under uncertainty,which is the ultimate focus of this research.The two supporting branches are data-driven engineering optimization under uncertainty and data-driven engineering decision under uncertainty,of which the former is the basis of the latter.The two basic elements are fundamentals of data science and uncertainty theory,which promote the representation and manifestation of the proposed theory.Second,the theory of data-driven engineering optimization under uncertainty is investigated.An optimization model involving two levels of uncertainty is established and analyzed.Existence theorem of the optimum solution,as well as the basic and varied forms of Kuhn-Tucker theorem of the established optimization model is proved.By adopting Neumaier clouds to describing the parameter uncertainty,a data-driven optimization model involving two levels of uncertainty is further established.The specific optimization form aiming at solving decision problems is also investigated and the strategy of coupling subjective engineering knowledge in the decision optimization model is also proposed.Solving strategy of the proposed data-driven engineering optimization under uncertainty is studied,and three algorithms are conceived according to dimensions of optimization variables and parameters.Initialization principles for establishing data-driven optimization models under uncertainty as well as principles for model updating are studied,and the effects of model updating on optimum solutions are also analyzed.Third,the theory of data-driven engineering decision under uncertainty is investigated.Aiming at design decision problems in engineering,a data-driven engineering decision model involving two levels of uncertainty is established.The relationship between the data-driven engineering optimization under uncertainty and the data-driven engineering decision under uncertainty is elaborated.For the data-driven decision model updating,a separation updating strategy is proposed.Updating methods for each separated part of the data-driven engineering decision model are given in detail.According to the background of engineering design in real life,four basic decision rules for applying the proposed data-driven engineering decision under uncertainty are summarized.Their mathematical implications are all rigorously elaborated and corresponding conclusions are also proven mathematically in detail.Fourth,the theory of data-driven engineering design under uncertainty is proposed.The characteristics of stage-based and improvised engineering design paradigms are compared.A tensor representation theorem involving the transformation from non-numerical data to numerical data is proposed,unifying data forms without losing any information.The roles of data and knowledge in design process are analyzed by investigating their respective importance,concluding that engineering design is inherently a data-driven process.Finally,a complete framework of data-driven engineering design under uncertainty is presented.Fifth,the proposed theory of data-driven engineering design under uncertainty is applied to a small Automatic Identification System(AIS)satellite mission design problem.According to the conceived mission goal,design database and knowledge repository are initialized.The problems of orbit determination,design objective and constraints modeling in early design stages,as well as the problems of overall scheme design and power subsystem design in middle design stages are investigated.The obtained optimum solutions are assessed.To sum up,this dissertation conducted a creationary research on how data and uncertainty affect the engineering design process,proposed the theory of data-driven engineering design under uncertainty,established an integrated theoretical hierarchy of the proposed theory,and applied it to an engineering design case.The work in this dissertation has laid a consolidated foundation for further research of the proposed theory,and served as a favorable reference experience in other application fields.