Fix ND problems stack overflow in Enright-Pryce benchmarks

[ChrisRackauckas-Claude]

Summary

• Fixed stack overflow errors in ND1-ND5 problems in the Enright-Pryce benchmark suite
• Simplified symbolic expressions to avoid ModelingToolkit compilation issues
• Maintained mathematical correctness while improving compilation stability

Problem

The ND problems (orbital mechanics/Kepler orbit benchmarks) were causing stack overflow errors during compilation. The root cause was the complex symbolic expression:

r = sqrt(y[1]^2 + y[2]^2)^3

When used in the gravitational force equations (-y[1]) / r and (-y[2]) / r, this created overly complex symbolic manipulations that caused ModelingToolkit's symbolic engine to exceed stack limits during compilation.

Solution

Replaced the complex expression with mathematically equivalent intermediate variables:

# Before (causes stack overflow):
r = sqrt(y[1]^2 + y[2]^2)^3

# After (compiles successfully):
r_squared = y[1]^2 + y[2]^2
r_cubed = r_squared * sqrt(r_squared)  

This represents the same physics (gravitational force law F ∝ 1/r³) but breaks down the symbolic computation into manageable steps.

Test Plan

• Verified ND problems compile without stack overflow
• Confirmed mathematical equivalence of the expressions
• All five ND problems (ND1-ND5) with different eccentricities should now work
• Run full benchmark suite to verify performance characteristics unchanged

The fix specifically addresses the compilation issue reported in the Enright-Pryce benchmarks documentation where all ND problems showed "StackOverflowError".

🤖 Generated with Claude Code