Steady-state CFD with k-ω SST turbulence model on a 7.8M-cell mesh identified the dominant loss contributors in an automotive coolant loop. Four targeted geometry changes reduced total pressure drop by 30% — from 18.6 kPa to 13.1 kPa — cutting pumping power from 132 W to under 95 W.
The client's thermal system team had sized the coolant pump for a 95 W budget at a nominal flow rate of 0.35 kg/s through a 1.2 m assembly. Early pressure estimates based on straight-pipe correlations came in 39% over target — but without CFD, there was no way to know which of the six bends, the T-junction manifold, or the area transitions was driving the overrun, or what to change.
Committing to tooling without resolving this would have locked in a larger, heavier pump — a direct energy and cost penalty in both EV and ICE configurations. The objective was to quantify each loss contributor, identify which geometry changes offered the best reduction per unit complexity, and deliver a simulation-validated design recommendation before any metal was cut.
SIMFORGE built a steady-state incompressible CFD model in ANSYS Fluent. The geometry — three pipe diameters (18, 20, 22 mm), six bends, one T-junction manifold, and four area transitions — was meshed with a polyhedral core and 10 structured prism layers at all walls to resolve the near-wall velocity gradient accurately at the target y⁺ of 30–80.
Turbulence closure: k-ω SST, selected for its well-documented superiority over k-ε in flows with adverse pressure gradient and moderate flow curvature — the dominant physics in bend-and-junction assemblies. Working fluid properties (density, viscosity, thermal conductivity) were modelled as temperature-dependent polynomials calibrated to the 50/50 water-glycol mixture at 65–90 °C.
Loss attribution was performed by integrating static pressure over control volumes isolating each feature class — pipe segments, individual bends, the T-junction, and each area change — giving a direct, actionable contribution breakdown.
A 1.2 m coolant transfer assembly connecting the pump outlet to the battery thermal plate and ICE radiator circuit, with six mitre-entry bends and a T-junction manifold splitting flow between the two circuits.
| Geometry feature | Value |
|---|---|
| Total assembly length | 1.2 m |
| Pipe inner diameter | 18–22 mm (three sections) |
| Number of bends | 6 — mitre entry, R/D ≈ 1.5 |
| Junction type | T-junction manifold |
| Area changes | 4 — abrupt transitions |
| Working fluid | Water-glycol 50/50 |
| Operating temperature | 65–90 °C |
| Mesh parameter | Value |
|---|---|
| Total cell count | 7.8 M |
| Cell type | Polyhedral core + prism layers |
| Prism layers | 10 — growth ratio 1.2 |
| y⁺ target | 30–80 |
| Refinement zones | Bends, junction, area changes |
| Turbulence model | k-ω SST |
| Convergence criterion | Residuals < 10⁻⁵ · ΔP monitor |
| Boundary | Condition | Value |
|---|---|---|
| Inlet | Mass flow rate | 0.35 kg/s |
| Outlet | Gauge pressure | 0 Pa |
| Fluid temperature | Uniform | 80 °C |
| Pipe walls | No-slip, adiabatic | — |
| Fluid properties | Temperature-dependent polynomial | W-G 50/50 @ 65–90 °C |
Pressure loss was integrated over isolated control volumes for each feature class. The T-junction manifold and bends together account for 59% of total loss — both are geometry-driven and directly addressable without changing pipe routing or flow rate.
Each modification was assessed independently in the CFD model to isolate its contribution, then combined in a single optimised geometry for final validation. No pipe re-routing required — all changes are local to the bend radii, junction fillet, and transition profiles.
| Configuration | ΔP | Pump power | Reduction |
|---|---|---|---|
| Baseline | 18.6 kPa | 132 W | — |
| Increased bend radius only | 15.3 kPa | 109 W | −18% |
| + Junction fillet | 13.5 kPa | 96 W | −27% |
| + Smooth transitions | 12.7 kPa | 90 W | −32% |
| Full optimised (+ diameter) | 13.1 kPa | 93 W | −30% |
At 31% of total loss, the T-junction manifold is far more significant than the six bends combined (28%). A fillet radius of just 5 mm at the branch entry eliminates the separation bubble responsible for 85% of the junction loss — a sub-£50 tooling change in a cast housing.
Standard Darcy-Weisbach estimates predicted 28% of total loss from pipe friction. CFD shows it is actually 32% — close, but the more important error is the 31% T-junction loss that correlations do not model at all. Early design sizing based on correlations alone was guaranteed to miss the pump budget.
A 2 mm diameter increase (18→20 mm) reduces friction loss by 10% but adds packaging volume and weight. CFD showed that adding the junction fillet and bend radius changes first brings pump power to 96 W — within budget — without touching the diameter. Diameter increase becomes an optional reserve margin, not a first-resort fix.
Dean vortices in each bend create a distorted velocity profile that persists 5–8 diameters downstream. This means the T-junction, positioned immediately after a bend, sees a non-uniform inlet profile that amplifies separation. Increasing bend radius reduces both the local bend loss and the downstream junction loss — a multiplicative effect that linear superposition of correlations cannot capture.
| Deliverable | Format |
|---|---|
| CFD simulation report — methodology, results, loss attribution, design recommendations | PDF / DOCX |
| Pressure and velocity contour plots — baseline & optimised | PNG · CFD-Post |
| Loss contribution breakdown — per feature class | Excel + PDF |
| Optimisation summary — ΔP vs geometry change matrix | Excel + PDF |
| Mesh and solver settings file | .cas / .dat (Fluent) |
| Geometry-ready CAD with recommended changes annotated | STEP + PDF redline |
| Tool | Usage |
|---|---|
| ANSYS Fluent | Steady-state CFD solver, k-ω SST |
| ANSYS Meshing / Fluent Meshing | Polyhedral core + prism layer generation |
| SpaceClaim / CATIA | CAD geometry prep, transition smoothing |
| CFD-Post | Post-processing, loss integration, contour plots |
| Python | Automated control-volume pressure integration, report generation |
From EV battery thermal management to ICE cooling circuits and industrial heat exchangers — if it flows, we can size it, optimise it, and deliver a fixed-fee CFD report with actionable geometry recommendations before your tooling decision.
Initial response · 24 h · NDA available · Fixed-fee proposal within 48 h