Supporting MDO through dynamic workflow (re)generation

The use of advancements in computing technology has enabled designers to perform more thorough and more detailed design studies. Multidisciplinary Design Optimization (MDO) architectures provide users with guidelines on how to structure their MDO problem, including the linking of disciplines and how to perform the optimization. However, complex MDO problems can consist of tens of disciplines and hundreds of design variables. Thus, the set-up of these problems can be complex and time consuming. In an attempt to reduce the time required and complexity of this set up, the main goal in this thesis is: "To develop and demonstrate a methodology for automatic workflow (re)generation to support MDO". The method to fulfill these requirements consists of three main steps. The first is the automatic generation of microworkflows, workflows representing the different disciplines of the problem. The user will need to specify the inputs, outputs and operations, after which the workflows are automatically generated. The second step involves the automatic storage of workflows. Workflows are stored in a graph database, allowing the addition of semantics to the data. Adding semantics allows a reasoner to understand what the data means, enabling the inferring of data not explicitly defined. OWL (Web Ontology Language) ontologies are used to supply structure to the workflow data and add semantics. In addition, materialization scripts are present to regenerate stored workflows. The final step of the implementation involves the automatic generation of simulation workflows according to different MDO architectures. This generation involves the materialization and adjustment of microworkflows and the creation of a ‘higher level’ workflow that links the disciplines and performs the optimization. The implementation of the automatic architecture generation has been validated using three case studies of varying complexity, amount of disciplines and discipline couplings. These case studies have shown a reduction of 93 to 98 % of time spent on the generation of simulation workflows representing the problem using an MDO architecture. In addition, the approach reduces the required user expertise and minimizes the amount of information the user needs to provide.Aerospace EngineeringFlight Performance and Propulsio