Add Herschel-Bulkley non-Newtonian viscosity model#1298
Add Herschel-Bulkley non-Newtonian viscosity model#1298jasontruong2707 wants to merge 11 commits intoMFlowCode:masterfrom
Conversation
Review Summary by QodoAdd Herschel-Bulkley non-Newtonian viscosity model with dynamic Reynolds number computation
WalkthroughsDescription• Implements Herschel-Bulkley non-Newtonian viscosity model with Papanastasiou regularization for power-law, Bingham, and HB fluids • Replaces static viscous Reynolds arrays (Res_viscous, Res_gs, Res_pr) with dynamic computation via new m_re_visc module • Adds per-fluid non-Newtonian parameters: yield stress (tau0), consistency index (K), flow behavior index (nn), viscosity bounds (mu_max, mu_min, mu_bulk), and regularization parameter (hb_m) • Integrates non-Newtonian viscosity computation into time-stepping CFL calculations, Riemann solvers, and data output stability criteria • Adds validation for non-Newtonian fluid parameters in input checker • Extends MPI communication across pre-processing, simulation, and post-processing modules to broadcast non-Newtonian properties • Supports non-Newtonian viscosity in immersed boundary method calculations • Includes two validation cases: 2D Poiseuille flow and 2D lid-driven cavity with non-Newtonian fluids • Adds non-Newtonian parameter definitions to toolchain configuration • Maintains backward compatibility with Newtonian flows (non-Newtonian disabled by default) Diagramflowchart LR
A["Input Parameters<br/>non_newtonian, tau0, K, nn"] --> B["m_hb_function<br/>Herschel-Bulkley<br/>Viscosity Model"]
B --> C["m_re_visc<br/>Dynamic Re Computation"]
C --> D["Time Stepping<br/>CFL Calculation"]
C --> E["Riemann Solvers<br/>Viscous Fluxes"]
C --> F["Data Output<br/>Stability Criteria"]
G["Velocity Gradients"] --> C
H["Strain Rate Tensor"] --> B
File Changes1. src/simulation/m_time_steppers.fpp
|
Code Review by Qodo
1. non_newtonian lacks case validation
|
![qodo-code-review[bot]](https://avatars.githubusercontent.com/in/484649?s=60&v=4){kind=small}
| _r(f"{px}non_newtonian", LOG, {"viscosity"}) | ||
| for a in ["tau0", "K", "nn", "mu_max", "mu_min", "mu_bulk", "hb_m"]: | ||
| _r(f"{px}{a}", REAL, {"viscosity"}) |
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
src/simulation/m_time_steppers.fpp
Outdated
| if (any_non_newtonian) then | ||
| if (igr) then | ||
| call s_compute_re_visc(q_cons_ts(1)%vf, alpha, j, k, l, Re_visc_per_phase) | ||
| else | ||
| call s_compute_re_visc(q_prim_vf, alpha, j, k, l, Re_visc_per_phase) | ||
| end if | ||
| call s_compute_mixture_re(alpha, Re_visc_per_phase, Re) |
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
|
Caution Review failedFailed to post review comments 📝 WalkthroughWalkthroughAdds Herschel–Bulkley non‑Newtonian support: eight new fields to the derived type physical_parameters (non_newtonian, tau0, K, nn, mu_min, mu_max, mu_bulk, hb_m); new modules m_hb_function and m_re_visc for HB viscosity and viscous/Re computations; input validation via s_check_inputs_non_newtonian; MPI broadcasts and GPU updates extended to the new fields; time‑stepper dt computation updated to use per‑phase viscous terms; pressure‑relaxation internals simplified; two example case scripts added; and CI workflow scripts reference shared frontier scripts. 🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Tip Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs). Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Format branch before merge to avoid whitespace conflicts in m_riemann_solvers.fpp. Add new modules to module_categories.json. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude Code ReviewHead SHA: ab1a4c8 Key files: Summary
FindingsCritical / Correctness1. Hard-coded shear-rate clamp is unphysical and problem-specific shear_rate = min(max(shear_rate, 1.0e-2_wp), 1.0e5_wp)This clamps 2. Riemann-solver index mapping for NORM_DIR=2 and NORM_DIR=3 is incorrect The Riemann solver loops over rotated index triplets call s_compute_re_visc(q_prim_vf, alpha_L, k, j, l, ...)but inside 3. Significant4. Performance regression on Newtonian-only cases 5. 6. IBM non-Newtonian viscosity uses dynamic_viscosities(fluid_idx) = f_compute_hb_viscosity(..., 1._wp, ...)The IBM force/torque integration evaluates viscosity at a fixed shear rate of 1 s⁻¹ rather than the actual local shear rate at each IB ghost-cell. This will give incorrect viscous forces on immersed boundaries in non-Newtonian flows with spatially varying Minor7. File header says !! @file m_hb_function.f90 ! should be .fpp
!! @file m_re_visc.f90 ! should be .fpp8. 9. Missing regression test for non-Newtonian viscous flows The IBM torque bug fix, the |
Add Herschel-Bulkley parameter descriptions to descriptions.py. Add fluid materials table entries and detailed section in case.md with model equation, parameter table, and special cases. Improve PHYSICS_DOCS entry with fuller explanation. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude Code ReviewHead SHA: 4d042f5 Key files:
Summary:
Findings1. Shear-rate lower clamp is physically wrong —
|
Fix .f90 -> .fpp in Doxygen headers for m_hb_function and m_re_visc. Add shear-thinning and Bingham regression tests to CI test suite. Re-merge with upstream/master after ruff PR (MFlowCode#1300) landed. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Code Review — Non-Newtonian ViscosityThorough investigation of all findings from the automated review. Each item was verified against the source code. Critical / Correctness1. Index mapping bug in Riemann solver for NORM_DIR=2,3
For NORM_DIR=2 (y-normal faces), the Riemann solver calls: call s_compute_re_visc(q_prim_vf, alpha_L, k, j, l, Re_visc_per_phase_L)Inside Impact: Wrong shear rate → wrong viscosity at y-normal and z-normal Riemann interfaces in 2D/3D non-Newtonian simulations. Most severe on anisotropic grids. Newtonian fluids are unaffected (early exit before gradient computation). Affected locations in Fix options:
2. GPU coherence:
GPU kernels in Impact: Silent wrong results on all GPU builds when Fix: Add Significant3. Hard-coded shear-rate clamp is not parameterizable
shear_rate = min(max(shear_rate, 1.0e-2_wp), 1.0e5_wp)The bounds Recommendation: Either make these user-configurable ( 4. Missing pre-computed gradients in Riemann solver calls All calls to Impact: Performance regression (redundant gradient computation) and slight inconsistency with the viscous flux gradients. Not a correctness bug. 5. IBM non-Newtonian viscosity uses fixed
dynamic_viscosities(fluid_idx) = f_compute_hb_viscosity( &
..., 1._wp, ...) ! Fixed shear rate = 1The IBM force/torque calculation evaluates viscosity at a constant shear rate of 1 s⁻¹ rather than the local shear rate at each IB ghost cell. The comment "compute reference viscosity at gdot=1" acknowledges this but doesn't explain the physical reasoning. Impact: Incorrect viscous forces on immersed boundaries in flows with spatially varying shear rate. Low severity if IBM + non-Newtonian is not a primary use case. 6. Newtonian performance regression ✅ No issue Verified: Minor (fixed in latest push)7. File headers say
8. Boundary shear rate asymmetry
9. Missing regression test ✅ Fixed Added two non-Newtonian test cases to
Both run in the single-fluid viscous test block for each dimension. Summary
Items 1 and 2 should be addressed before merge — they produce silently wrong results in 2D/3D and GPU builds respectively. |
Claude Code ReviewHead SHA: Files changed: 28 Key files:
Summary
Findings🔴 CRITICAL — Missing
|
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude Code ReviewHead SHA: Summary
Findings🔴 Critical — Missing namelist bindings in
|
Generated from 6 test cases (shear-thinning + Bingham, 1D/2D/3D). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude Code ReviewHead SHA: 9c15db4 Key files
Summary
Findings🔴 Critical — Missing namelist bindings in
|
2 participants
Summary
New files
Test plan