Performing accurate design optimization, sensitivity analysis, and uncertainty quantification of large- and multi-scale electromagnetics problems has far-reaching applications. Due to computational cost, physics tools with limited accuracy, and a lack of optimization and uncertainty quantification (O&UQ) frameworks, such activities are often ignored or abandoned shortly after they have begun. Lack of robust O&UQ tools leads engineers and scientists to inadvertently create fragile solutions, leading to a greater number of design/fabrication/test cycles at high cost and schedule impact. Designers are not usually able to quantify the impact of uncertainty in design inputs and fabrication processes in order to improve the robustness of a design or at least better understand key sensitivities. Critical to solving these challenges is incorporation of advanced O&UQ capability with modern HPC-enabled computational electromagnetics (CEM) software, all while improving user accessibility of these functions. During this program, IERUS will incorporate the O&UQ utilities of DAKOTA into our O&UQ software. IERUS has already developed an initial uncertainty modeling capability for EM software in a tool called MURPH. However, MURPH implements a naïve Monte Carlo approach to UQ. By offering the features available in DAKOTA to our customers through V-Lox (with MURPH), IERUS will eliminate the expertise required to setup and use the complex O&UQ features available in DAKOTA. This will bring our customers an expert-level O&UQ toolset for CEM problems without the need for expert-level knowledge of the underlying algorithms. This work has four major objectives. First, the team will demonstrate the feasibility of integrating the optimization and uncertainty quantification tools available in the Dakota toolkit into a Monte Carlo uncertainty modeling software framework and 3D CEM software package. This will provide confidence toward the ultimate goal of a commercial HPC-enabled simulation software with built-in access to expert-level optimization. Second, the team will deploy the optimization and uncertainty quantification capability to cloud computing resources to support large scale analyses. The team will augment Dakota with a fast and expressive surrogate modeling capability. Finally, the team will demonstrate the prototype using several real-world, large-scale optimization examples relevant to commercialization interests. The long-term goal, initiated by the work of this Phase I program, is to provide users with a powerful HPC-enabled design optimization and sensitivity analysis capability. After initial release of these features into commercial products, a design-as-a-service product leveraging Dakota in an HPC environment for automated device design will be developed. This will provide users with turn-key operation of software with state-of-the-art design capability and reduce the overall workload of design engineers while also achieving design performance previously unrealized. These new capabilities will also enable the academia, government, and small business to provide high performance results using a simulation-as-a-service business model since not all users may have access to large-scale computational resources.
