Funded research is producing several highly specialized tools in different code-bases written with different languages. The work presented here demonstrates a proof-of-concept application connecting distinct, and possibly distributed, research codes that provide individual components of a multi-physics, coupled, or systems-level simulation. While interoperability of research codes providing distinct physics is becoming more common, this work focuses on providing a turn-key solution to adjoint sensitivity analysis. The example uses three distinct instances of a multi-physics solver to separately evaluate the fluid, mesh displacement, and structural equations. The computed sensitivities are used in a gradient based optimization algorithm to find the optimal topology of a structure under air loads. The objective considered is minimum weight and the constraint considered is maximum tip displacement. Time for serialization and deserialization of data is reported compared to the time to: 1) evaluate the finite element equations, and 2) solve the resulting global linear system. Time spent transferring data is approximately one-twentieth of the time to solve the linear system, but is within one-sixth of the time to evaluate the finite element equations.
展开▼
