A user-interactive, response surface approximation-based framework for multidisciplinary design

Multidisciplinary Design Optimization (MDO) focuses on reducing the time and cost required to design complex engineering systems. One goal of MDO is to develop systematic approaches to design which are effective and reliable in achieving desired performance improvements. Also, the analysis of engineering systems is potentially expensive and time-consuming. Therefore, enhancing the efficiency of current design methods, in terms of the number of designs that must be evaluated, is desirable.; A design framework, Concurrent Subspace Design (CSD), is proposed to address some issues that are prevalent in practical design settings. Previously considered methods of system approximation and optimization were extended to accommodate both discrete and continuous design variables. Additionally, the transition from one application to another was made to be as straightforward as possible in developing the associated software.; Engineering design generally requires the expertise of numerous individuals, whose efforts must be focused on and coordinated in accordance with a consistent set of design goals. In CSD, system approximations in the form of artificial neural networks provide information pertaining to system performance characteristics. This information provides the basis for design decisions. The approximations enable different designers to operate concurrently and assess the impact of their decisions on the system design goals.; The proposed framework was implemented to minimize the weight of an aircraft brake assembly. An existing industrial analysis tool was used to provide design information in that application. CSD was implemented in a user-interactive fashion that permitted human judgement to influence the design process and required minimal modifications to the analysis and design software. The implications of problem formulation and the role of human design experts in automated industrial design processes were explored in the context of that application.; In the most recent testcase, CSD was able to reduce the weight of the brake assembly from 589 to 292 pounds and the final design met all imposed performance requirements. This was achieved at modest computational expense relative to the cost of conventional optimization techniques. These results demonstrate that the CSD approach to design is of potential value in practical environments.