Rack Liquid Cooling: Cold-Plate Flow, Pressure Drop and CDU Sizing

Size rack cooling flow, pressure drop and pump duty. Calculate CDU exchanger approach with ε–NTU and fouling, then check the accelerator coolant inlet limit.

#CDU #Churchill correlation #cold plates #Darcy–Weisbach #liquid cooling #pump sizing #rack cooling #rack liquid cooling

Use this rack liquid cooling calculator to size cold-plate flow, pressure drop, pump duty and the liquid-to-liquid CDU heat exchanger. It connects the facility-water inlet to the actual accelerator coolant inlet through the complete thermal chain. Data-center engineers can compare duty points and degraded conditions; students can follow the heat balances and counterflow effectiveness–NTU method.

Start with sustained IT heat, liquid capture fraction, heat reserve and rack topology. Set the parallel branches per rack and the number of equally loaded cold plates in each series string. The calculator selects the larger of the flow needed for the allowed rack temperature rise and the combined vendor minimum plate flows. Every series plate carries its branch flow. The schematic shows representative paths.

Enter a measured cold-plate pressure drop at a reference flow, its fitted exponent, hose dimensions and fitting coefficients. Supply and return lengths are combined within each section. The pump duty includes one branch path, its series plates, common piping and the CDU auxiliary loss. Parallel branch pressures are not added. The system curve gives the required pressure at each flow; it does not predict the operating point without a pump curve.

Facility water temperature is not the GPU coolant inlet temperature. Enter facility flow and properties, the exchanger's clean UA, heat-transfer area and combined area-based fouling resistance. The counterflow ε–NTU model uses both heat-capacity rates and the reduced UA to solve the secondary return and exchanger outlet temperatures needed to reject the rack heat plus pump allowance. The pump is placed after the exchanger; its conservative heat allowance raises the delivered secondary supply slightly.

Set the accelerator's qualified coolant inlet limit. In a series string, the last accelerator receives coolant already warmed by upstream plates. The result reports that inlet, its temperature headroom and whether this particular limit is met. It does not calculate chip junction temperature. The facility-flow and fouling charts show how the thermal margin disappears while the hydraulic sizing can remain unchanged.

The defaults are an assumed design example, not a GPU or CDU product rating. At 30 °C facility inlet, a clean 18 kW/K exchanger and 150 L/min facility flow deliver approximately 34.67 °C secondary supply. Two series plates put the last inlet at approximately 39.67 °C against an assumed 40 °C limit. Reducing facility flow to 100 L/min, or adding 0.0003 m²·K/W fouling on the 10 m² area, raises the approach to roughly 7–7.4 K and exceeds that inlet limit.

Use vendor UA and pressure-loss data at the actual coolant and both flows. The sensitivity plots hold clean UA fixed; changing flow can also change film coefficients in a real exchanger. This model excludes branch maldistribution, bypass control, transients, junction temperatures, primary pressure losses, cavitation, condensation control and coolant chemistry. Check phase, freeze protection and component ratings separately. Installation work belongs to a qualified person.

How it is calculated

Reserved liquid heat is Pd=NrPrfliqMQP_d=N_rP_rf_{liq}M_Q. With N=NrNbN=N_rN_b balanced paths, select V˙s=max⁡[Pd/(ρcpΔTmax),Nqmin]\dot V_s=\max[P_d/(\rho c_p\Delta T_{max}),Nq_{min}] and V˙b=V˙s/N\dot V_b=\dot V_s/N. Rack rise is Pd/(ρV˙scp)P_d/(\rho\dot V_sc_p); equal plate loads give ΔTplate=ΔTrack/Ns\Delta T_{plate}=\Delta T_{rack}/N_s.

Circular-pipe losses use Δp=(fL/D+ΣK)ρv2/2\Delta p=(fL/D+\Sigma K)\rho v^2/2 with Churchill's all-regime Darcy friction factor (ASHRAE Fundamentals 2025, Chapter 3, equations 30–34). One plate uses the local measured fit Δpcp=Δpref(qb/qref)n\Delta p_{cp}=\Delta p_{ref}(q_b/q_{ref})^n. Add the series plates, one branch hose path, common pipes and quadratic auxiliary loss; then apply the pressure reserve. Parallel path losses are not summed. Electrical pump duty is Pe=ΔppumpV˙s/ηwP_e=\Delta p_{pump}\dot V_s/\eta_w. Exchanger duty is Q=Pd+PeQ=P_d+P_e.

For the liquid-to-liquid CDU, Cs=m˙scp,sC_s=\dot m_sc_{p,s}, Cf=m˙fcp,fC_f=\dot m_fc_{p,f}, Cmin=min⁡(Cs,Cf)C_{min}=\min(C_s,C_f) and Cr=Cmin/CmaxC_r=C_{min}/C_{max}. Combined fouling resistance RfR_f is referred to the stated heat-transfer area AA:

1UAdirty=1UAclean+RfA,NTU=UAdirtyCmin.\frac{1}{UA_{dirty}}=\frac{1}{UA_{clean}}+\frac{R_f}{A},\qquad NTU=\frac{UA_{dirty}}{C_{min}}.

The counterflow relation is

ε=1−e−NTU(1−Cr)1−Cre−NTU(1−Cr),ε∣Cr=1=NTU1+NTU.\varepsilon=\frac{1-e^{-NTU(1-C_r)}}{1-C_re^{-NTU(1-C_r)}},\qquad \varepsilon\big|_{C_r=1}=\frac{NTU}{1+NTU}.

The implementation uses an equivalent small-argument expansion to remain continuous and accurate around equal capacity rates. NPTEL Heat Exchangers, PDF pp. 11–12 and 27–29, supplies the resistance and ε–NTU relations.

Close the recirculating secondary loop at its imposed heat load:

Th,in=Tf+QεCmin,Th,out=Th,in−QCs.T_{h,in}=T_f+\frac{Q}{\varepsilon C_{min}},\qquad T_{h,out}=T_{h,in}-\frac{Q}{C_s}.

All pump electrical power is conservatively added to the liquid after the exchanger: Ts=Th,out+Pe/CsT_s=T_{h,out}+P_e/C_s. Rack return is Ts+Pd/CsT_s+P_d/C_s, equal to Th,inT_{h,in}. Facility return is Tf+Q/CfT_f+Q/C_f. Both exchanger terminal differences are calculated. The delivered supply approach Ts−TfT_s-T_f includes pump heat; the separate exchanger cold-end approach Th,out−TfT_{h,out}-T_f excludes it. This distinction follows the FWS/TCS supply-temperature convention in OCP's CDU Test Methodology, revision 1.0 (2024), §3.3.2.1.

For equal series plate heat loads, the warmest device inlet is Tin,max=Ts+(Ns−1)ΔTplateT_{in,max}=T_s+(N_s-1)\Delta T_{plate}. Headroom is the entered accelerator inlet limit minus this temperature. Positive headroom checks only coolant inlet temperature, not junction temperature.

OCP OAI Liquid Cooling Guidelines (2023), PDF p. 7, informs topology and the example 10 K rack-rise target. NIST water properties at 30 °C and 1 bar supply the water defaults. APALTEK's eight-GPU example checks thermal flow and topology; ASHRAE Example 2 checks hydraulic loss. CMU's published countercurrent example checks the equal-capacity ε–NTU limit. Sensitivity curves use 101 directly evaluated points.

An independent distributed energy-balance boundary-value integration (SciPy 1.15.3 solve_bvp) verifies the reduced-flow, fouled and equal-capacity closed-loop temperatures without using ε–NTU. These regressions are identified separately from published reference data.

Assumptions

Steady, incompressible, single-phase Newtonian fluids with constant properties at representative mean temperatures. Primary and secondary properties are independent. Identical racks, balanced branches, equal heat per plate, no bypass, leakage or pipe heat exchange. The pump allowance enters immediately after the exchanger and before the rack.

The exchanger is an ideal two-stream counterflow unit with uniform effective UA, no axial conduction and no heat loss to the room. Enter clean UA at the intended facility and selected secondary flows. Fouling from both sides is combined on one common area basis; if the areas differ, convert each resistance to that basis before summing. Enter clean rather than already fouled UA to avoid counting fouling twice.

Clean UA stays fixed when varying facility flow, heat load or rack-rise target. The model captures finite primary/secondary heat-capacity rates and added fouling resistance; it does not infer flow-dependent film coefficients from exchanger geometry. Re-enter a flow-specific clean UA when assessing another final duty point.

The default 18 kW/K UA, 10 m² area, 40 °C accelerator inlet limit, clean condition, 150 L/min facility flow, component losses, geometry and pump efficiency are illustrative design inputs, not a manufacturer's ratings. The 40 °C limit applies at every accelerator inlet in this example and must be replaced with the chosen device's qualified limit.

Full-flow common pipe length excludes the distributing portions of a manifold. Branch hose length includes both legs. The cold-plate power law and quadratic auxiliary loss are local fits for the entered coolant. The plate reference-flow warning is a proximity reminder; the plotted hydraulic sweep is not a wider validation of that fit. Heat and pressure reserves are explicit design choices.

Where it applies

Use for duty-point planning of balanced rack loops. A passed coolant inlet check does not establish acceptable junction temperature or guarantee all branches receive the calculated flow. Validate the hottest downstream accelerator with vendor thermal data. The bulk coolant rise is not a cold-plate thermal resistance.

Counterflow ε–NTU is appropriate to the stated idealized exchanger, not an uncorrected multipass or crossflow unit. Validate effective UA with supplier performance data. Flow sweeps hold clean UA fixed: reduced facility flow can also reduce convective heat transfer and raise the real approach further. Fouling is thermal resistance only; blockage and its hydraulic penalty are not inferred. No CDU control valve, bypass, dew-point setpoint, fouling growth rate or transient thermal storage is solved.

At very low facility flow or UA the required steady temperatures may be outside a fluid's liquid range or equipment ratings. Such results indicate an unsuitable duty point; this constant-property model has no phase-equilibrium or pressure-rating data. Check both exchanger terminal differences, temperature limits, freeze protection, pressure ratings and condensation separately.

Churchill friction does not replace measured losses for corrugated hoses, developing passages or cold-plate microchannels. Check Reynolds-dependent fitting coefficients and transitional flow. A filled closed loop has no net circulation elevation head; fill pressure, absolute pressure, NPSH, expansion, relief and surge are outside this model.

Pump pressure includes auxiliary losses and reserve. Compare the external circuit loss with a supplier's available external differential pressure without subtracting internal losses twice. A pump curve must meet the selected flow and pressure at the actual efficiency and redundancy configuration. CDU heat duty includes reserve once and the conservative full electrical pump allowance. Primary-side pressure losses and facility pump sizing are not calculated. Check against the standard in force where applicable. Installation work belongs to a qualified person.

References

  • OCP, OAI System Liquid Cooling Guidelines, March 2023, §1.2–1.3, PDF p. 7. https://www.opencompute.org/documents/oai-system-liquid-cooling-guidelines-in-ocp-template-mar-3-2023-update-pdf
  • ASHRAE Handbook—Fundamentals, 2025 edition, Chapter 3 Fluid Flow, equations (30)–(34) and Example 2. https://handbook.ashrae.org/Handbooks/F25/IP/f25_ch03/f25_ch03_ip.aspx
  • NIST Chemistry WebBook, Thermophysical Properties of Fluid Systems, water at 30 °C and 1 bar, isobaric query (accessed 30 October 2026). https://webbook.nist.gov/cgi/fluid.cgi?ID=C7732185&Action=Page
  • APALTEK, Liquid Cold Plate Design Calculations, 12 October 2026, fluid properties and worked eight-GPU loop example. https://www.apaltek.com/liquid-cold-plate-design-calculations/
  • NPTEL / IIT, Heat Exchangers, course 112108149, Module 7 teacher slides, PDF pp. 11–12 and 27–29: fouling resistance and effectiveness–NTU method. https://archive.nptel.ac.in/content/storage2/courses/112108149/pdf/M7/M7TeacherSlides.pdf
  • OCP, Liquid to Liquid CDU Test Methodology and Performance Rating, revision 1.0, 2024, §3.3.2.1 and §3.3.2.3: FWS/TCS approach definition, flow and performance reporting. https://www.opencompute.org/documents/ocp-wp-l-lcdu-test-methodology-performance-rating-r1-pdf
  • Carnegie Mellon University, Kitchin Research Group, Differentiable Flowsheets, Heat Exchanger Examples §2.1, accessed 1 November 2026. https://kitchingroup.cheme.cmu.edu/differentiable-flowsheets/examples/07_heat_exchangers.html
  • Independent counterflow boundary-value verification, Python 3 / SciPy 1.15.3 solve_bvp, 1 November 2026; attached numerical results, not published measurements. https://docs.scipy.org/doc/scipy-1.15.3/reference/generated/scipy.integrate.solve_bvp.html

Inputs

InputSymbolUnitDefault
Racks on this CDU
Identical racks; the common run carries the combined flow.
N_r1
IT heat per rackP_rkW100
Fraction of IT heat captured by liquidf_liq0.8
Heat-load sizing factor
1.15 adds 15% reserve before flow sizing.
M_Q1.15
Parallel branches per rack
Identical, balanced branches. All flow passes through these branches.
N_b48
Cold plates in series per branch
Each plate carries the branch flow; plate pressure drops add.
N_s2
Maximum rack coolant risedT_maxK10
Facility-water inlet temperature
Facility inlet to the CDU exchanger, before its temperature approach.
T_f°C30
Accelerator coolant inlet limit
Use the qualified limit at every accelerator's cold-plate inlet.
T_limit°C40
Coolant density
Water at 30 °C, 1 bar (NIST). Replace with actual coolant properties at mean temperature.
rhokg/m^3995.65
Coolant specific heat
Water at 30 °C, 1 bar (NIST). Fixed property; it does not change with supply temperature.
c_pJ/(kg*K)4179.8
Coolant dynamic viscosity
Water at 30 °C, 1 bar (NIST). Glycol needs its own viscosity and measured component curves.
mumPa*s0.79722
Facility-water flow through CDUq_fL/min150
Clean exchanger UA
Use the supplier's clean UA at the entered primary and secondary flows.
UA_cleankW/K18
Exchanger heat-transfer area
Use the same area basis as the combined fouling resistance.
A_hxm^210
Combined fouling resistance
Sum both sides on the stated area basis; zero represents a clean exchanger.
R_fm^2*K/W0
Facility-water densityrho_fkg/m^3995.65
Facility-water specific heatcp_fJ/(kg*K)4179.8
Minimum flow through each plate
Vendor minimum. Series plates carry the same flow, so do not multiply this by the series count.
q_minL/min1
Plate reference flowq_refL/min1.5
Pressure drop of ONE plate at reference flow
Design-basis input, not a product rating. Replace with a measured value for the chosen coolant.
dp_refkPa20
Plate pressure-drop exponent
Fit Δp ∝ flowⁿ to vendor data near the duty point; 1.8 is an assumed example fit.
n_dp1.8
Branch hose length, supply + return
Total wetted hose length in ONE complete branch, excluding the plates.
L_bm4
Branch hose inside diameterD_bmm10
Branch fitting loss coefficient sum
Use branch-hose velocity. Include connectors and balancing valves; do not count plate losses again.
K_b6
Common pipe length, supply + return
Full-flow common run before the split and after the merge. Do not multiply by the rack count.
L_cm10
Common pipe inside diameterD_cmm40
Common fitting loss coefficient sum
Use common-pipe velocity. Include only fittings not already in the auxiliary curve.
K_c4
Pipe and hose absolute roughness
Assumed smooth-tube design basis; use measured or supplier roughness for both modeled runs.
epsmm0.0015
Auxiliary reference TOTAL flowq_auxL/min150
Auxiliary loss at reference total flow
Combined CDU heat exchanger, filter and other common losses, modeled ∝ flow². Omit anything counted in pipe K values.
dp_auxkPa20
Pump-pressure sizing factor
Pressure reserve at the selected flow, not an extra flow margin.
M_p1.15
Pump wire-to-water efficiency
Combined pump, motor and drive efficiency at the duty point.
eta_w0.55

Results

With the default inputs above, the calc gives:

ResultValue
Total secondary flow132.6 L/min
Flow through each cold plate2.76 L/min
CDU thermal duty incl. pump allowance92.7 kW
Secondary supply to rack34.7 °C
Delivered supply approach to facility4.67 K
Last accelerator coolant inlet39.7 °C
Accelerator inlet headroom0.33 K
Coolant inlet checkInlet limit met
Rack return temperature44.7 °C
Facility-water return38.9 °C
Exchanger UA with fouling18 kW/K
Exchanger effectiveness0.687
Total pump differential, with reserve175.3 kPa
External circuit loss, before reserve136.8 kPa
Pump electrical duty705 W
Rack coolant rise10 K

Checked against sources

Each case below was calculated by this calc and agrees with the value from its source (within the tolerance where one is shown).

Published eight-GPU example: four parallel strings, two plates each

Source: APALTEK, Liquid Cold Plate Design Calculations (12 Oct 2026), worked 8-GPU loop, 8 kW, PG25, 40 °C inlet, 4×2 topology: total flow 12 L/min, plate flow 3 L/min, mean plate rise about 5 K. Uses its tabulated rho=1.024 kg/L and cp=3.92 kJ/(kg K). Its 12 L/min is a rounded design target; these properties give 11.958 L/min for exactly 10 K, so ±0.06 L/min allows the published rounding. Hydraulics are zeroed; this case verifies only thermal sizing and topology. https://www.apaltek.com/liquid-cold-plate-design-calculations/ The published 40 °C rack inlet is prescribed, whereas this tool now calculates it from UA and facility conditions; only flow, topology and temperature rise are compared.

Inputs: IT heat per rack = 8 kW; Fraction of IT heat captured by liquid = 1; Heat-load sizing factor = 1; Parallel branches per rack = 4; Coolant density = 1,024 kg/m³; Coolant specific heat = 3,920 J/(kg·K); Coolant dynamic viscosity = 2 mPa·s; Minimum flow through each plate = 0 L/min; Plate reference flow = 3 L/min; Pressure drop of ONE plate at reference flow = 0 kPa; Branch hose length, supply + return = 0 m; Branch fitting loss coefficient sum = 0; Common pipe length, supply + return = 0 m; Common fitting loss coefficient sum = 0; Auxiliary loss at reference total flow = 0 kPa; Pump-pressure sizing factor = 1; the others as above.

ResultSourceThis calc
Total secondary flow12 ± 0.06 L/min11.958 L/min
Flow through each cold plate3 ± 0.02 L/min2.989 L/min
Mean coolant rise per series plate5 ± 0.03 K5 K

Published eight-GPU example: eight parallel plates

Source: APALTEK, Liquid Cold Plate Design Calculations (12 Oct 2026), same worked 8-GPU load and PG25 properties, all eight plates in parallel: 12 L/min total, 1.5 L/min per plate, approximately 10 K per plate return based on a specified inlet. The published flow is rounded; the energy balance retains rho=1024 kg/m³ and cp=3920 J/(kg K). No hydraulic claim is made. https://www.apaltek.com/liquid-cold-plate-design-calculations/ The published 40 °C rack inlet is prescribed, whereas this tool now calculates it from UA and facility conditions; only flow, topology and temperature rise are compared.

Inputs: IT heat per rack = 8 kW; Fraction of IT heat captured by liquid = 1; Heat-load sizing factor = 1; Parallel branches per rack = 8; Cold plates in series per branch = 1; Coolant density = 1,024 kg/m³; Coolant specific heat = 3,920 J/(kg·K); Coolant dynamic viscosity = 2 mPa·s; Minimum flow through each plate = 0 L/min; Pressure drop of ONE plate at reference flow = 0 kPa; Branch hose length, supply + return = 0 m; Branch fitting loss coefficient sum = 0; Common pipe length, supply + return = 0 m; Common fitting loss coefficient sum = 0; Auxiliary loss at reference total flow = 0 kPa; Pump-pressure sizing factor = 1; the others as above.

ResultSourceThis calc
Total secondary flow12 ± 0.06 L/min11.958 L/min
Flow through each cold plate1.5 ± 0.01 L/min1.495 L/min
Mean coolant rise per series plate10 ± 0.05 K10 K

ASHRAE 2025 fluid-flow Example 2(a): pipe and fitting head loss

Source: ASHRAE Handbook—Fundamentals 2025, Chapter 3, Example 2(a): Q=2.1 ft³/s, D=6 in, L=340 ft, roughness=0.01 in, rho=62.4 lbm/ft³, nu=1.08e-5 ft²/s, summed K rounded to 11.1. Published Darcy f=0.023 and friction head loss=47.5 ft (14.478 m). The calculator retains the full Churchill value instead of rounding f to 0.023; its head is about 14.435 m. The tolerance allows that rounding (<0.6%). Only the friction-loss part of this open-reservoir example is tested; its elevation and discharge kinetic head are not part of a closed rack loop. https://handbook.ashrae.org/Handbooks/F25/IP/f25_ch03/f25_ch03_ip.aspx

Inputs: IT heat per rack = 1 kW; Fraction of IT heat captured by liquid = 1; Heat-load sizing factor = 1; Parallel branches per rack = 1; Cold plates in series per branch = 1; Minimum flow through each plate = 3,568 L/min; Plate reference flow = 3,568 L/min; Pressure drop of ONE plate at reference flow = 0 kPa; Auxiliary loss at reference total flow = 0 kPa; Branch hose length, supply + return = 103.6 m; Branch hose inside diameter = 152.4 mm; Branch fitting loss coefficient sum = 11.1; Common pipe length, supply + return = 0 m; Common pipe inside diameter = 152.4 mm; Common fitting loss coefficient sum = 0; Pipe and hose absolute roughness = 0.254 mm; Coolant density = 999.6 kg/m³; Coolant dynamic viscosity = 1.003 mPa·s; Pump-pressure sizing factor = 1; the others as above.

ResultSourceThis calc
Branch Darcy friction factor0.023 ± 2 × 10⁻⁴0.02286
Branch hose and fitting loss as coolant head14.478 ± 0.08 m14.435 m

CMU counterflow benchmark: equal capacity rates, NTU = 2

Source: CMU Kitchin Research Group, Differentiable Flowsheets, Heat Exchanger Examples §2.1: both streams 10 mol/s × 75 J/(mol K) give C = 750 W/K, UA = 1500 W/K, NTU = 2 and published effectiveness 0.667. Equivalent mass-based properties reproduce those capacity rates. The rack heat is reduced to 7.5 kW to stay within this tool's 10 K rise; only C, NTU and effectiveness are compared, not the example's outlet temperatures or its 75 kW duty. https://kitchingroup.cheme.cmu.edu/differentiable-flowsheets/examples/07_heat_exchangers.html

Inputs: IT heat per rack = 7.5 kW; Fraction of IT heat captured by liquid = 1; Heat-load sizing factor = 1; Parallel branches per rack = 1; Cold plates in series per branch = 1; Coolant density = 1,000 kg/m³; Coolant specific heat = 1,000 J/(kg·K); Minimum flow through each plate = 45 L/min; Plate reference flow = 45 L/min; Pressure drop of ONE plate at reference flow = 0 kPa; Auxiliary loss at reference total flow = 0 kPa; Branch hose length, supply + return = 0 m; Branch fitting loss coefficient sum = 0; Common pipe length, supply + return = 0 m; Common fitting loss coefficient sum = 0; Pump-pressure sizing factor = 1; Facility-water flow through CDU = 45 L/min; Facility-water density = 1,000 kg/m³; Facility-water specific heat = 1,000 J/(kg·K); Clean exchanger UA = 1.5 kW/K; Facility-water inlet temperature = 26.85 °C; Accelerator coolant inlet limit = 100 °C; the others as above.

ResultSourceThis calc
Secondary heat-capacity rate750 W/K750 W/K
Facility heat-capacity rate750 W/K750 W/K
Number of transfer units22
Exchanger effectiveness0.6670.6667

Distributed-balance verification: Reduced facility flow

Source: Independent Python 3 / SciPy 1.15.3 solve_bvp run of distributed counterflow energy balances with heat-load and facility-inlet boundary conditions; no ε–NTU formula used. See attached settings and results. This is a numerical regression, not a published device measurement. Pump heat is the specified 704.6963906573898 W hydraulic-model allowance; that hydraulic duty is not independently verified by this thermal test. Solver residual <9e-11. https://docs.scipy.org/doc/scipy-1.15.3/reference/generated/scipy.integrate.solve_bvp.html

Inputs: Facility-water flow through CDU = 100 L/min; the others as above.

ResultSourceThis calc
Secondary supply to rack37.045 °C37.045 °C
Last accelerator coolant inlet42.045 °C42.045 °C
Delivered supply approach to facility7.045 K7.045 K
Accelerator inlet headroom-2.045 K-2.045 K

Distributed-balance verification: Fouled exchanger

Source: Independent Python 3 / SciPy 1.15.3 solve_bvp run of distributed counterflow energy balances with heat-load and facility-inlet boundary conditions; no ε–NTU formula used. See attached settings and results. This is a numerical regression, not a published device measurement. Pump heat is the specified 704.6963906573898 W hydraulic-model allowance; that hydraulic duty is not independently verified by this thermal test. Solver residual <9e-11. https://docs.scipy.org/doc/scipy-1.15.3/reference/generated/scipy.integrate.solve_bvp.html

Inputs: Combined fouling resistance = 3 × 10⁻⁴ m²·K/W; the others as above.

ResultSourceThis calc
Secondary supply to rack37.439 °C37.439 °C
Last accelerator coolant inlet42.439 °C42.439 °C
Delivered supply approach to facility7.439 K7.439 K
Accelerator inlet headroom-2.439 K-2.439 K

Distributed-balance verification: Equal capacity rates

Source: Independent Python 3 / SciPy 1.15.3 solve_bvp run of distributed counterflow energy balances with heat-load and facility-inlet boundary conditions; no ε–NTU formula used. See attached settings and results. This is a numerical regression, not a published device measurement. Pump heat is the specified 704.6963906573898 W hydraulic-model allowance; that hydraulic duty is not independently verified by this thermal test. Solver residual <9e-11. https://docs.scipy.org/doc/scipy-1.15.3/reference/generated/scipy.integrate.solve_bvp.html

Inputs: Facility-water flow through CDU = 132.6 L/min; the others as above.

ResultSourceThis calc
Secondary supply to rack35.227 °C35.227 °C
Last accelerator coolant inlet40.227 °C40.227 °C
Delivered supply approach to facility5.227 K5.227 K
Accelerator inlet headroom-0.2269 K-0.2269 K

Related Resources

Responses

Your email address will not be published. Required fields are marked *

L ading...