Five to fifteen configurations is the honest output of a typical design phase at a small aerospace company. That range appears across team sizes, problem types, and tool stacks. Compute is not usually the limiting factor. The constraint is the engineer time needed to prepare, run, interpret, and act on each simulation.
Five to fifteen configurations cannot describe a substantial design space. A structural bracket with 625 configurations (four free geometry parameters, five levels per parameter) already exceeds what a team can evaluate by hand. So engineers narrow the search, select the likeliest options, and accept the strongest result among them. That choice is rational under time pressure, but it can leave aerospace hardware heavier or less efficient than the physics allows.
Where Simulation Time Goes
Solver runtime is rarely the main simulation bottleneck. For many component-scale FEA problems, current workstation hardware returns results within minutes or a few hours. The time-consuming work surrounds the solver:
Geometry preparation is substantial. Imported CAD often needs cleanup before meshing: small features must be removed, gaps at assembly interfaces resolved, and cosmetic details suppressed. For a first-time component evaluation, preparation alone can take an hour or more per configuration.
Mesh generation still requires engineering judgment. An automated tool can create a mesh, but an analyst must assess element quality, aspect ratios in critical regions, and whether mesh density matches the expected stress gradient. Automated checks can pass while the mesh remains too coarse in areas where stress changes rapidly.
Boundary conditions depend on the physical situation. Load magnitude and direction are usually clear. Support conditions are harder. Pinned, fixed, and semi-rigid supports can produce materially different results, and the CAD model does not always reveal which condition is correct.
Post-processing adds another block of work. Someone must extract maximum Von Mises stress, locate it, check displacement at specified points, compare results with allowables, and record the comparison for review. None of these tasks is especially difficult, but each requires an engineer who is also handling other work.
How Teams Work Around It
Small aerospace teams use practical workarounds. These choices fit the constraint, but they narrow the design search and can produce weaker outcomes.
The most common workaround is conservative design: add mass margin rather than test the structural limit. When the optimum is uncertain, extra margin provides room for error. The cost is mass. For a launch vehicle or satellite, that cost spreads through mission analysis. A 10% structural mass penalty translating to 5-8% payload mass penalty at system level can materially affect the result.
Another workaround is geometric anchoring. Teams start from a proven configuration and make incremental changes. The baseline mesh and boundary conditions reduce preparation time, but the search becomes narrower by definition. Engineers refine a known concept instead of looking broadly for a better one.
A third workaround is disciplinary decomposition. Teams separate the coupled problem into single-domain analyses and use simplified models at the discipline boundaries. This works when coupling is weak. When coupling is strong, as in a thermal-structural problem in a compact propulsion system, the method can understate interaction effects. Each discipline may appear optimized while the coupled system is not.
Why the Bottleneck Is Growing
The bottleneck is longstanding, but several structural trends are making it more acute for small aerospace teams.
Problems are becoming harder. Tighter mass targets and higher performance requirements push designs beyond comfortable margins, making multidisciplinary optimization necessary for compliant solutions. Additive manufacturing also enlarges the available geometry space, creating forms without prior design history that require first-principles simulation rather than empirical lookup tables.
Simulation capability has expanded faster than analyst capacity. Small teams can access more tool capability per seat, while the same engineers must still prepare and interpret the analyses. Compute costs per analysis have fallen sharply. Analyst time has not. The constraint has moved from solver cost to analyst time.
What Automation Addresses
Automation has real value, with clear limits. It handles candidate evaluation, but does not replace engineering judgment in problem formulation or design selection.
Automation addresses per-candidate setup. If geometry preparation, meshing, boundary conditions, solver execution, and result extraction run without analyst intervention for each candidate, throughput per analyst-hour rises. A process requiring four hours of analyst time per candidate evaluation can theoretically support ten candidates per analyst-hour when per-candidate cost drops to six minutes of setup time for automated exception handling.
Automation does not replace initial problem formulation. Engineers must define load cases, boundary conditions, allowables, and design constraints. Automating evaluation helps only when that formulation is correct. Higher throughput does not repair an incorrect model.
Automation also does not replace final selection. A set of 200 physics-valid candidates still requires an engineer to weigh manufacturing constraints, integration needs, program risk, and factors outside the physics model. The engineer spends less time running analyses and more time making decisions, but that decision remains essential.
Throughput Numbers
Consider a growing UAV development company with four engineers, ten-week design phase for a new structural component. At fifteen to twenty configurations (traditional cadence), the team sees only a narrow sample. With several hundred configurations (automated cadence), the same team can review a broader candidate set and make its selection with evidence from the actual Pareto frontier.
The candidates omitted from the first process did not cease to be options. They were hidden by the throughput limit. Expanding the visible set does not change what can be designed in ten weeks, but it increases the chance that the selected design is close to the true optimum.
This is the problem P1 addresses: not replacing simulation with something faster and less accurate, but removing per-candidate analyst time so engineers can review more candidates before selecting a design.