AI-orchestrated EM simulation: the NASA PEC Almond benchmark

What this benchmark is

The NASA PEC Almond (Problem IIIA) is a canonical scattering target from the UT Austin Computational Electromagnetics Benchmark Suite, with calibrated range measurements the CEM community uses to check solver accuracy. Nullspace EM ran it across four frequencies and two solver configurations, with every result compared against the UT Austin measurement data.

Nullspace EM computed the physics: a full-wave Method of Moments solver with proprietary Fast Direct Compression, running the largest case at 37,502 unknowns on a standard laptop, with every RCS value a direct full-wave solution inside the ±1 dB measurement band. An AI coding assistant handled the workflow around it, script generation through report assembly, through the Nullspace Python API. That split is the point: the solver runs the physics, AI runs the workflow, and neither approximates the other.

Inside the benchmark

• The benchmark definition: geometry, frequencies, mesh density, and solver configurations

• The six-stage automated pipeline, from prompt to finished deck, with per-stage timing

• Full performance metrics across all eight runs: fill time, solve time, peak RAM

• Monostatic RCS comparison plots against UT Austin measurement, VV and HH, with RMS error