Fix C++ CFD Solver Accuracy
Budget: ₹1,500 – ₹12,500 INR
I have a legacy C++-based CFD solver that compiles and runs without crashing, yet the results are clearly off. Pressures drift, residuals refuse to settle, and key boundary layers look nothing like the benchmark data. No recent edits have been made, so the issue is embedded somewhere in the existing numerical implementation rather than a new bug I introduced.
The main problems I can see are:
• Physical accuracy – predicted flow fields deviate significantly from published results for identical cases.
• Numerical stability – residuals plateau well above the expected threshold and sometimes grow after a few hundred iterations.
• Boundary condition handling – inflow and wall conditions appear to be applied, but the resulting gradients tell a different story.
I need someone to trace the root causes, correct the offending sections, and leave me with a solver that reproduces standard test cases (e.g., lid-driven cavity, backward-facing step) within accepted error bands. You’ll be working directly in modern C++ (11/14), and familiarity with common CFD data structures, finite-volume schemes, and typical linear solvers (GS, CG, BiCGStab) is essential.
Deliverables and acceptance tests:
• Clean, documented source code that compiles with g++/clang on Linux.
• A concise report outlining each bug fixed or algorithm adjusted and why.
• Verification results showing converged solutions for at least two benchmark cases matching reference values within 2 % for global quantities and displaying stable residual decay.
If you are comfortable diving deep into C++ numerical kernels, pressure-velocity coupling routines, and subtle boundary logic, I would love to work with you.
The main problems I can see are:
• Physical accuracy – predicted flow fields deviate significantly from published results for identical cases.
• Numerical stability – residuals plateau well above the expected threshold and sometimes grow after a few hundred iterations.
• Boundary condition handling – inflow and wall conditions appear to be applied, but the resulting gradients tell a different story.
I need someone to trace the root causes, correct the offending sections, and leave me with a solver that reproduces standard test cases (e.g., lid-driven cavity, backward-facing step) within accepted error bands. You’ll be working directly in modern C++ (11/14), and familiarity with common CFD data structures, finite-volume schemes, and typical linear solvers (GS, CG, BiCGStab) is essential.
Deliverables and acceptance tests:
• Clean, documented source code that compiles with g++/clang on Linux.
• A concise report outlining each bug fixed or algorithm adjusted and why.
• Verification results showing converged solutions for at least two benchmark cases matching reference values within 2 % for global quantities and displaying stable residual decay.
If you are comfortable diving deep into C++ numerical kernels, pressure-velocity coupling routines, and subtle boundary logic, I would love to work with you.