The move toward electrified mobility has made efficient thermal management one of the most critical issues in the design of electric vehicles. My work is motivated by the continuous increase in energy density and charge-discharge power of lithium-ion batteries, which directly raises the heat flux that must be rejected from battery modules. Traditional air-cooling strategies cannot always maintain the surface temperature and temperature uniformity required by modern electric vehicle battery packs. Therefore, I focused on microchannel heat exchangers because they have a very large surface-area-to-volume ratio, a compact structure, and the ability to support both single-phase and two-phase cooling schemes.
In an electric vehicle battery thermal management system, the working fluid must keep the cell temperature below about 50 °C while maintaining the temperature difference inside the module below roughly 5 °C. Two-phase microchannel cooling can exploit latent heat to absorb a large amount of heat at a nearly constant saturation temperature. Consequently, I selected low-boiling-point electronic fluorinated fluids as the coolants. These fluids have dielectric properties, chemical inertness, and low global warming potential, making them suitable for direct or indirect cooling of electric vehicle battery cells. In particular, I experimentally investigated the flow boiling behavior of HFE-649 and Noah-2100A in pin-fin microchannel heat exchangers.
Background and Motivation
Electric vehicles offer a route to reduce greenhouse gas emissions in the transportation sector. However, battery performance, life, and safety all depend on temperature. Lithium-ion cells generate heat during charging and discharging because of ohmic resistance, polarization, and electrochemical reactions. If the cooling system cannot dissipate that heat effectively, the pack temperature can exceed the recommended window. A severe condition called thermal runaway can then occur, resulting in smoke, fire, or explosion. My literature review revealed that passive air cooling has a relatively low heat-transfer coefficient. Liquid cooling can increase the heat-transfer coefficient several times compared with air cooling. Phase-change materials and heat pipes also show promise, but they often face challenges in terms of weight, cost, and long-term reliability.
I summarized the main technologies in Table 1. Each method has distinct advantages and limitations for electric vehicle battery thermal management. In my opinion, microchannel liquid cooling combined with phase change through fluorinated fluids represents an attractive solution because it couples high heat transfer coefficients with acceptable pumping power. The objective of this thesis is to understand how pin-fin geometry and fluid properties affect heat transfer, pressure drop, bubble dynamics, and critical heat flux in such exchangers.
| Method | Heat-transfer enhancement | Uniformity | Pumping or parasitic energy | Key challenge |
|---|---|---|---|---|
| Air cooling | Low | Moderate | Low to moderate | Limited by ambient temperature and low heat capacity |
| Liquid cooling | High | Good | Moderate | Need for pump, manifold, and leak management |
| Phase-change material | Passive, moderate | Very good | None | Low thermal conductivity and leakage risk |
| Heat pipe | High | Good | None | Fabrication cost and orientation constraints |
| Microchannel flow boiling | Very high | Excellent | Moderate | Flow instability and critical heat flux limits |
To evaluate microchannel exchangers for electric vehicle applications, I designed an experimental loop that permitted controlled mass flow, inlet temperature, inlet vapor quality, and heating power. I used transparent windows to observe bubble formation and high-speed imaging to quantify the bubble dynamics. The measured variables included temperatures, pressures, flow rate, wall temperature, and heat flux. The loop contained a gear pump, a mass flow meter, a thermostatic bath, a plate heat exchanger, a condenser, a test section, and a data-acquisition system based on LabVIEW. A direct-current power supply sent power to cartridge heaters located beneath a copper heating block.
Experimental System and Procedures
The microchannel test section is a critical component. I used a copper bottom plate with a machined array of pin fins. The top cover included a quartz glass window, which made it possible to observe the two-phase flow patterns. A silicone O-ring and several screws sealed the assembly. The main channel had a total heated length of 50 mm and width of 50 mm, while the flow height above the base was 1 mm. Three pin-fin shapes were studied: circular, diamond, and droplet-shaped. I kept the same nominal fin spacing and pin height for all shapes so that geometry effects could be compared directly. Table 2 gives the relevant dimensions used in the coordinate system of the test section.
| Pin shape | W1 (mm) | W2 (mm) | C (mm) | S (mm) |
|---|---|---|---|---|
| Circular | 3.4 | 7.2 | 9.425 | 1 |
| Diamond | 3.4 | 7.2 | 8.485 | 1 |
| Droplet-shaped | 3.4 | 7.2 | 10.609 | 1 |
The coolant properties are important for interpreting the results. I selected HFE-649 and Noah-2100A because both fluids have boiling points close to the desired temperature range for electric vehicle battery thermal management. HFE-649 boils at 49 °C, while Noah-2100A boils at 47 °C. Their saturated liquid densities are similar, but their vapor densities, latent heats, viscosities, and thermal conductivities differ enough to change the flow-boiling behavior. Table 3 summarizes the physical properties obtained from REFPROP.
| Property | HFE-649 | Noah-2100A |
|---|---|---|
| Boiling point (°C) | 49 | 47 |
| Liquid density (kg/m³) | 1602 | 1601 |
| Vapor density (kg/m³) | 12.552 | 11.316 |
| Dynamic viscosity (Pa·s) | 6.4121 × 10⁻⁴ | 5.70276 × 10⁻⁴ |
| Specific heat (kJ/(kg·K)) | 1.103 | 1.279 |
| Surface tension (N/m) | 10.8 × 10⁻³ | 11.44 × 10⁻³ |
| Latent heat of vaporization (kJ/kg) | 93.22 | 88 |
| Thermal conductivity (W/(m·K)) | 0.0587 | 0.0609 |
| GWP | 1 | 20 |
Before performing detailed experiments, I checked the airtightness of the loop by charging it with dry air to about 1.5 bar and monitoring the pressure for 12 hours. A leak was not acceptable because the working fluids are volatile and because mass-flow stability is necessary for accurate boiling heat-transfer measurements. I also carried out repeated experiments under identical conditions. The maximum deviations in wall temperature and pressure drop were about 2.5 °C and 0.3 kPa, respectively, which confirmed that the bench could provide repeatable results.
Data Reduction and Uncertainty
I used standard equations to compute the hydraulic diameter, Reynolds number, heat flux, heat-transfer coefficient, and Nusselt number. The hydraulic diameter of the pin-fin array was calculated from the flow volume and the wetted area. The porosity is:
$$
D_h = \frac{4 V_{\text{flow}}}{A_{\text{HT}}}
$$
where \(V_{\text{flow}}\) is the fluid volume within the heated region and \(A_{\text{HT}}\) is the total wetted surface area of the channel base and fins. The inlet velocity is based on the inlet cross-sectional area:
$$
v_{\text{in}} = \frac{\dot{m}}{\rho A_{\text{in}}}
$$
The mean longitudinal velocity in the fin array can be estimated by dividing by the porosity:
$$
v = \frac{\dot{m}}{\rho A_{\text{in}}\varepsilon}
$$
where \(\varepsilon\) is the fin-array porosity:
$$
\varepsilon = 1 – \frac{N A_{\text{fin}}}{L_{\text{ch}} W_{\text{ch}}}
$$
The single-phase Reynolds number is:
$$
Re = \frac{\rho v D_h}{\mu}
$$
For the two-phase tests, a two-phase Reynolds number was also useful:
$$
Re_{\text{tp}} = \frac{D_h G}{\mu_l + \alpha(\mu_g – \mu_l)}
$$
Here, \(G\) is the mass flux, \(\alpha\) is the void fraction, and \(\mu_l\) and \(\mu_g\) are the liquid and vapor dynamic viscosities. The heat flux was obtained by assuming one-dimensional steady conduction through the copper block. The measured temperature difference between two rows of thermocouples gave the local heat flux:
$$
q = k_{\text{Cu}} \frac{T_2 – T_1}{\Delta y}
$$
where \(k_{\text{Cu}}\) is the thermal conductivity of copper, \(T_1\) and \(T_2\) are the mean temperatures of the two thermocouple rows, and \(\Delta y\) is the vertical distance between them. The corresponding bottom wall temperature of the microchannel is:
$$
T_w = T_1 – \frac{q}{k_{\text{Cu}}} \delta
$$
where \(\delta\) is the distance from the uppermost thermocouple row to the channel base. The reference fluid temperature was taken as the arithmetic mean of the inlet and outlet bulk temperatures:
$$
T_r = \frac{T_{\text{in}} + T_{\text{out}}}{2}
$$
The total heat absorbed by the fluid is:
$$
\dot{Q} = \dot{m}(h_{\text{out}} – h_{\text{in}})
$$
In the subcooled and saturated regions, I calculated the heat-transfer coefficient from:
$$
h = \frac{q A_d}{(T_w – T_r) A_{\text{eff}}}
$$
where \(A_d\) is the projected heated area and \(A_{\text{eff}}\) is the effective heat-transfer area. Because the fins contribute to heat transfer, I applied the usual fin efficiency:
$$
\eta_{\text{fin}} = \frac{\tanh(m_f H_f)}{m_f H_f}
$$
$$
m_f = \sqrt{\frac{h L_w}{\lambda_{\text{fin}} A_f}}
$$
Therefore, the two-phase heat-transfer coefficient \(h_{\text{tp}}\) was obtained as:
$$
h_{\text{tp}} = \frac{q_{\text{eff}} A_d}{(T_w – T_{\text{sat}})\left(A_{\text{base}} + N A_{\text{fin}}\eta_{\text{fin}}\right)}
$$
In addition to the Reynolds number, I used the boiling number, Prandtl number, Weber number, and Martinelli parameter to develop predictive correlations. The boiling number is:
$$
Bo = \frac{q}{G h_{fg}}
$$
The liquid-only and vapor-only Reynolds numbers are:
$$
Re_l = \frac{G(1-x) D_h}{\mu_l}
$$
$$
Re_g = \frac{G x D_h}{\mu_g}
$$
The liquid Weber number is:
$$
We_l = \frac{G^2 D_h}{\rho_l \sigma}
$$
All measured variables had small uncertainties. I propagated the instrument tolerances using a standard root-sum-square method. The relative uncertainty of the hydraulic diameter was about 1.1%; the heat flux uncertainty was 2.9%; the heat-transfer coefficient uncertainty was 2.3%; and the Nusselt number uncertainty was 2.7%. Table 4 lists the direct and indirect uncertainties.
| Quantity | Uncertainty |
|---|---|
| Length and width measurement | ±0.01 mm |
| DC power measurement | ±0.4% |
| Mass flow rate | ±0.1% |
| Pressure transducer | ±0.08% |
| Temperature sensor | ±0.1% |
| Hydraulic diameter | 1.1% |
| Heat flux | 2.9% |
| Heat-transfer coefficient | 2.3% |
| Nusselt number | 2.7% |
Flow Visualization Results for Different Pin-Fin Shapes
The first part of my experiments was carried out with HFE-649 in circular, diamond, and droplet-shaped microchannel heat sinks. I used a high-speed camera to record the flow patterns at different heat fluxes and mass flow rates. The key parameter used to quantify the bubble population was the equivalent bubble diameter, which was obtained from the measured projected bubble area. The tests covered mass flow rates from 0.6 g/s to 1.4 g/s, a constant inlet temperature of 30 °C, and imposed heating powers from 20 W to 140 W.
I observed that the bubble diameter increased when the mass flow rate increased. The reason is that higher liquid velocity produces larger drag and promotes the coalescence of bubbles. This trend was evident for all three pin-fin shapes. At a given mass flow, the diamond-shaped fins produced the largest equivalent bubble diameters. The sharp corners of the diamond fins generated separated flow regions and low-velocity zones that trapped vapor bubbles. The circular fins produced intermediate bubble sizes because their symmetrical shape provided stable wake regions that allowed bubbles to grow. The droplet-shaped fins generated the smallest bubbles, as their streamlined profile reduced wake recirculation and helped detach bubbles quickly.
Table 5 summarizes the effect of mass flow rate on bubble equivalent diameter in a qualitative but systematic way. I normalized the values to the bubble diameter at the lowest flow rate. The droplet geometry showed the largest relative increase over the tested mass-flow range. This outcome suggests that the droplet-shaped pin fins can effectively reduce the vapor slug length and improve flow stability in electric vehicle battery cooling applications.
| Pin-fin shape | Relative increase from 0.6 g/s to 1.4 g/s | Maximum bubble size ranking |
|---|---|---|
| Circular | 292% | Intermediate |
| Diamond | 135% | Largest |
| Droplet | 601% | Smallest |
When the heat flux was increased while mass flow was fixed, the bubble equivalent diameter first increased and then decreased. In the low-heat-flux region, greater superheat activates more nucleation sites and produces larger bubbles. In the high-heat-flux region, intense nucleation and strong bubble interactions cause bubbles to depart much sooner and break into smaller vapor fragments. This nonmonotonic behavior affects pressure drop and heat transfer in a complex way. The trend is therefore relevant to the thermal design of electric vehicle battery cooling plates because local hot spots can shift the bubble dynamics into different regimes.
Effect of Pin-Fin Shape on Pressure Drop
Measuring pressure drop was necessary because the pumping power in an electric vehicle thermal management system must remain acceptable. My results show that the pressure drop increased with mass flow rate for all geometries. Because the Reynolds number increases with flow rate, wall friction increases and momentum changes become larger. The pressure drop from 0.6 g/s to 1.4 g/s grew by 183%, 147%, and 184% for the circular, diamond, and droplet-shaped microchannel heat sinks, respectively.
The diamond pins gave the largest pressure drop among the three structures. This observation is attributed to the sharp leading edges and abrupt flow separations at the corners of the diamond pins. The wake behind each diamond fin is wide, and bubble accumulation in the corners narrows the effective flow area. In contrast, the droplet-shaped pin fins have a streamlined front profile, which suppresses separation and produces a lower pressure penalty. The circular pins lie between the two other configurations. At a mass flow rate of 1.4 g/s, the pressure drop of the diamond geometry was approximately 188% higher than that of the circular geometry and about 319% higher than that of the droplet geometry.
Pressure drop also increased with bubble size. Larger vapor bubbles occupy a greater fraction of the channel area and increase the interfacial shear. In extreme cases, elongated bubbles can bridge the channel and create temporary high-resistance vapor slugs. Since the diamond geometry promotes large bubbles, its pressure drop is especially sensitive to heat flux. For electric vehicle battery thermal management, the droplet-shaped microchannel would offer the lowest pressure-drop penalty, but its heat-transfer performance must also be considered.
I plotted the local wall temperature distribution along the heated surface at a fixed flow rate. The wall temperature first increased from the inlet to the middle section and then decreased near the outlet. Near the outlet, strong nucleate boiling and high vapor velocity enhanced heat removal. The circular geometry had the lowest peak wall temperature, whereas the diamond geometry had the highest. The circular pins produced better temperature uniformity, which is beneficial for electric vehicle battery lifetime and safety.
Critical Heat Flux for Different Pin-Fin Geometries
Critical heat flux defines the upper operating limit of a boiling heat exchanger. Beyond the critical heat flux, a vapor film covers the heated wall and causes a sharp decrease in heat-transfer efficiency. In an electric vehicle fast-charging scenario, the heat flux can become very high, so the critical heat flux of the cooling structure must be sufficiently large. I identified the critical heat flux point by increasing the heating power until the heat flux began to decrease even as wall superheat increased.
For all three geometries, the critical heat flux increased with mass flow rate. From 0.6 g/s to 1.4 g/s, the critical heat flux rose by approximately 118.8% for the circular pin fins, 80.9% for the diamond pin fins, and 117.6% for the droplet-shaped pin fins. The circular pin geometry produced the highest critical heat flux at all tested flow rates. At 1.4 g/s, the circular critical heat flux was 27.3 W/cm², about 43.3% higher than the diamond geometry and 15.1% higher than the droplet geometry. Table 6 shows these critical heat flux values at 1.4 g/s.
| Pin-fin shape | Critical heat flux (W/cm²) | Relative to circular |
|---|---|---|
| Circular | 27.3 | 1.000 |
| Diamond | ≈19.0 | 0.698 |
| Droplet | ≈23.7 | 0.869 |
The circular geometry appeared to provide additional secondary flow paths around the fins, allowing vapor to escape more easily. Because the diamond geometry trapped vapor in the corner regions, the local dry-out happened earlier. The droplet geometry performed better than the diamond geometry because the streamlined shape delayed flow separation and reduced vapor accumulation. Therefore, for conditions requiring a high critical heat flux in electric vehicle battery cooling, a circular pin-fin arrangement is probably the most reliable choice.
Comparison of HFE-649 and Noah-2100A
After determining that the circular microchannel geometry offered the best overall performance, I compared two electronic fluorinated fluids in that geometry. The operating parameters included mass flow rates from 0.8 g/s to 2.4 g/s, inlet temperatures of 20 °C and 30 °C, and inlet vapor qualities of 0.1, 0.2, and 0.3. I measured the heat-transfer coefficient, pressure drop, and critical heat flux. The purpose was to understand whether Noah-2100A could be used as an alternative to HFE-649 for electric vehicle battery thermal management.
Figure 4-1 shows that the Nusselt number increased with Reynolds number for both fluids. When the Reynolds number increased from about 24 to 70, the Nusselt number increased by approximately 97.4% for Noah-2100A and by 114.8% for HFE-649. At a fixed Reynolds number, Noah-2100A exhibited a slightly higher Nusselt number than HFE-649. This difference is attributed to the higher thermal conductivity and specific heat of Noah-2100A.
The heat-transfer coefficient also increased with mass flow rate. The stronger convective motion and increased bubble generation at higher flow rates enhance the heat-transfer process. At an inlet temperature of 20 °C and a mass flow rate of 2.4 g/s, the heat-transfer coefficient of Noah-2100A reached approximately 2500 W/(m²·K), which was about 4.1% higher than that of HFE-649. Figure 1 of the original data shows that inlet temperature has only a weak effect on heat-transfer coefficient in the tested range. In contrast, inlet vapor quality had a negative effect. When the inlet vapor quality increased from 0.1 to 0.3, the heat-transfer coefficient of Noah-2100A and HFE-649 decreased by 3.5% and 9.3%, respectively. Higher inlet quality reduces the ability of the fluid to absorb heat through evaporation near the upstream part of the channel.
Table 7 summarizes the heat-transfer and pressure-drop trends for the two coolants. I include the qualitative effect of mass flow, inlet temperature, and inlet vapor quality so that the ranking between the fluids can be easily seen.
| Parameter | Increasing mass flow | Increasing inlet temperature | Increasing inlet vapor quality | Fluid with higher value |
|---|---|---|---|---|
| Heat-transfer coefficient | Increases | Slight decrease | Decreases | Noah-2100A |
| Pressure drop | Increases | Increases | Increases | Noah-2100A |
| Critical heat flux | Increases | Not shown separately | Decreases | Noah-2100A |
The pressure-drop behavior is particularly important for pump sizing in an electric vehicle cooling circuit. My measurements showed that the pressure drop of both fluids increased strongly with mass flow rate. At an inlet temperature of 30 °C and a mass flow rate of 2.4 g/s, the pressure drop of Noah-2100A was about 40.5% higher than that of HFE-649. This result is partly caused by the lower vapor density of Noah-2100A. At a fixed heat flux and quality, the vapor volumetric flow rate is larger when the vapor density is smaller, which increases the flow velocity and frictional pressure losses.
The inlet temperature effect on pressure drop was also noticeable. At a fixed mass flow rate, pressure drop is higher when the inlet temperature increases from 20 °C to 30 °C. The warmer fluid reaches the saturation temperature earlier, which increases the length of the two-phase region. The resulting acceleration pressure drop and friction pressure drop are therefore larger. Table 8 gives the pressure-drop ratios that I observed at representative conditions.
| Condition | Noah-2100A / HFE-649 pressure-drop ratio |
|---|---|
| Inlet 30 °C, mass flow 2.4 g/s | 1.405 |
| Inlet vapor quality 0.3, mass flow 2.4 g/s | ≈2.20 |
The critical heat flux was also higher for Noah-2100A than for HFE-649. As mass flow increased, both fluids exhibited a higher critical heat flux because the replenishment of liquid at the heated wall was faster. As inlet vapor quality increased, the critical heat flux decreased. When the inlet quality changed from 0.1 to 0.3, the critical heat flux of Noah-2100A decreased by 12.9%, while that of HFE-649 decreased by 10.3%. At a mass flow rate of 2.4 g/s and an inlet quality of 0.1, Noah-2100A produced a critical heat flux about 19.8% higher than HFE-649.
Comprehensive Performance Evaluation
A coolant with a higher heat-transfer coefficient is attractive, but not if the pressure drop becomes unacceptable. Therefore, I used a thermal-performance ratio to evaluate the tradeoff between heat transfer and friction. The performance factor \(TPR\) is:
$$
TPR = \frac{Nu_1 / Nu_2}{\left(f_1 / f_2\right)^{1/3}}
$$
where subscript 1 refers to Noah-2100A and subscript 2 refers to HFE-649. If \(TPR > 1\), Noah-2100A offers a better net thermal-hydraulic performance. I found that \(TPR\) was above unity for all of my test points. The average value was about 1.29. Therefore, although Noah-2100A has a higher pressure drop and a slightly higher global warming potential than HFE-649, it still gives a better overall thermal performance in the circular pin-fin microchannel.
From an electric vehicle system perspective, the increased pressure drop of Noah-2100A would require a more powerful coolant pump. However, the improved heat-transfer coefficient and critical heat flux could allow a lower volumetric coolant flow rate or a smaller heat-exchange area. I therefore believe that Noah-2100A is a promising alternative for future electric vehicle battery thermal management.
Development of Heat-Transfer and Pressure-Drop Correlations
The design of microchannel heat exchangers for electric vehicle battery thermal management requires reliable correlations. Existing correlations for two-phase flow in small channels were not always accurate for my fluids and geometry. I therefore used dimensional analysis to construct a generalized Nusselt-number correlation. The variables governing heat transfer were assumed to appear through the two-phase Reynolds number, the Prandtl number, and the fin-array porosity. The proposed correlation is:
$$
Nu = 0.5057 \, Re_{\text{tp}}^{0.7481} \, \varepsilon^{2.922} \, Pr^{1.22}
$$
This correlation was fitted using the complete experimental data set for both fluids. Figure 8 in the original comparison showed that all data points for Noah-2100A fell within ±20% of the predicted values. The mean absolute percentage error was 7.4% for Noah-2100A and 13.9% for HFE-649. The mean absolute percentage error is defined as:
$$
MAPE = \frac{100\%}{n}\sum_{t=1}^{n}\left|\frac{A_t – F_t}{A_t}\right|
$$
where \(A_t\) is the measured value and \(F_t\) is the predicted value. The new Nusselt correlation is suitable for preliminary design calculations in the range of parameters tested here.
Pressure-drop prediction was more difficult. I used a Chisholm-type separated-flow model. The two-phase frictional pressure gradient is:
$$
\left(\frac{dp_f}{dz}\right)_{\text{tp}} = \Phi_l^2 \left(\frac{dp_f}{dz}\right)_l
$$
where the liquid-only frictional pressure gradient is:
$$
\left(\frac{dp_f}{dz}\right)_l = f_l \frac{2 G (1-x)}{D_h \rho_l}
$$
The two-phase multiplier is:
$$
\Phi_l^2 = 1 + \frac{C}{X} + \frac{1}{X^2}
$$
where \(X\) is the Martinelli parameter based on the liquid and vapor pressure gradients:
$$
X = \frac{\left(dp_f/dz\right)_l}{\left(dp_f/dz\right)_g}
$$
I initially tried to fit one Chisholm parameter \(C\) for both fluids. The general correlation had the form:
$$
C = 3.39 \times 10^{-6} \, We_{\text{tp}}^{3.305} \, Re_{\text{tp}}^{-1.33}
$$
The general correlation predicted most points within ±30%, but the mean absolute percentage error was still about 10%. Because the two fluids have different densities and surface-tension effects, I then modified the C-parameter separately for each fluid. The relation was:
$$
C = a_1 \, We_{\text{tp}}^{a_2} \, Re_{\text{tp}}^{a_3}
$$
The coefficients are listed in Table 9. The separation of the coefficients substantially improved the accuracy. Noah-2100A produced a mean absolute percentage error of 4.9%, while HFE-649 produced a mean error of 6.2%.
| Fluid | a₁ | a₂ | a₃ |
|---|---|---|---|
| Noah-2100A | 0.03086 | 1.351 | -0.4512 |
| HFE-649 | 0.000745 | 1.880219 | -0.877709 |
Both correlations are applicable to the pin-fin microchannel geometry and to the inlet temperature, mass-flux, and vapor-quality ranges covered in my experiments. I expect these correlations to be useful in system-level models of electric vehicle battery thermal management, especially when evaluating tradeoffs between heat-transfer surface area and pumping power.
Numerical Simulation of the Microchannel Heat Exchangers
To further understand the local velocity, temperature, and pressure fields that cannot be directly measured in such a small channel, I built three-dimensional computational fluid dynamics models with the software Fluent. I used the volume-of-fluid method for the two-phase flow and the Lee model to account for evaporation and condensation. The computational domain consisted of a 50 mm long, 50 mm wide, and 5 mm high solid copper block, with a 1 mm high flow channel cut into it. The pin fins were placed on the bottom surface. A constant heat flux was imposed at the lower boundary, and the top surface was modeled as an adiabatic wall. The inlet boundary condition was specified by mass flow rate and inlet temperature; the outlet boundary was set to a pressure outlet.
I tested several mesh densities for each pin-fin shape. For the circular geometry, four meshes with about 1.2 million to 6.25 million elements were evaluated. The wall temperature and pressure drop changed very little when the mesh was refined beyond 4.33 million elements. The selected final mesh had about 4.33 million cells. For the diamond pin-fin model, I selected a mesh of about 4.25 million cells. For the droplet geometry, a mesh of about 3.2 million cells was sufficient. Table 10 summarizes the mesh independence study.
| Geometry | Selected mesh size (cells) | Maximum variation in wall temperature | Maximum variation in pressure drop |
|---|---|---|---|
| Circular | 4.33 × 10⁶ | 0.6% | 1.6% |
| Diamond | 4.25 × 10⁶ | 0.2% | 0.4% |
| Droplet | 3.20 × 10⁶ | 0.5% | 1.9% |
I used HFE-649 as the working fluid in the simulations. The saturated liquid and vapor properties were entered as constant values because the pressure variation was relatively small. The latent heat was incorporated into the standard-state enthalpy difference between vapor and liquid. Air or noncondensable gases were not included. The operating boundary conditions are given in Table 11.
| Boundary | Value |
|---|---|
| Inlet temperature | 30 °C |
| Inlet mass flow rate | 0.8 – 2.4 g/s |
| Bottom heat flux | 20 – 120 kW/m² |
| Outlet pressure | Atmospheric |
| Backflow temperature | 49 °C |
The simulation results were validated by comparing the predicted heat-transfer coefficient and pressure drop with the measured data. The maximum discrepancy was approximately 15% for the heat-transfer coefficient and less than 10% for the pressure drop. The mean absolute percentage error for both parameters was below 8%. This gave me confidence that the model captured the essential physics of the flow-boiling problem.
After validation, I analyzed the velocity contours at a mass flow rate of 2.0 g/s. The diamond pin geometry produced the strongest velocity variation. High-velocity regions appeared near the side surfaces of the diamond fins, while large wakes developed downstream. The circular pins generated moderately uniform velocity distributions. The droplet-shaped pins produced the smoothest velocity field. Figure 10 from the simulation displays the three-dimensional streamlines. From those contours, I could see that the diamond fins obstruct the flow much more than the droplet fins. For an electric vehicle cooling plate, such strong obstruction could create hot zones on the heating surface.
The wall temperature contours showed that the circular microchannel had the most uniform temperature distribution. The diamond geometry had distinct high-temperature zones behind the fins, which were caused by trapped vapor and poor liquid replenishment. The droplet geometry performed well but still had slightly higher temperature peaks than the circular geometry. Thus, the simulation confirmed the experimental conclusion that circular pins provide the best combination of heat-transfer performance and temperature uniformity.
Pressure contours showed high pressure on the leading edge of each pin and lower pressure in the wake. The diamond pins generated the largest pressure difference from inlet to outlet. The droplet pins generated the lowest pressure drop. The circular pins were intermediate. These local pressure insights are useful when designing manifolds and pumps for electric vehicle battery thermal management.
I also ran the numerical model over a range of heat fluxes. The heat-transfer coefficient first increased with heat flux and then decreased when the heat flux was high. The initial increase is caused by the activation of more nucleation sites and the growth of vapor bubbles. The later decrease is related to the large vapor fraction near the heated wall, which reduces the effective liquid contact area. In the simulation, the pressure drop increased monotonically with heat flux for all structures because the vapor void fraction increased. Table 12 shows the relative pressure-drop increase from 20 kW/m² to 120 kW/m² for the three geometries at a mass flow rate of 2.6 g/s.
| Pin-fin shape | Pressure-drop increase |
|---|---|
| Circular | 279% |
| Diamond | 302% |
| Droplet | 271% |
The numerical results therefore support the experimental conclusion that circular pin fins offer a robust solution for microchannel heat exchangers in electric vehicle battery thermal management. In particular, the pressure-drop penalty is moderate, the wall-temperature field is uniform, and the critical heat flux is high. If the design must minimize pressure drop above all other constraints, droplet-shaped fins may be considered, but the circular fins appear to be the best compromise.
Discussion of Practical Implications
Electric vehicle battery packs are often cooled by cold plates mounted beneath or between cells. The heat flux generated by the batteries can vary with state of charge, state of health, and charge rate. During fast charging, the heat generation can be two or three times higher than during steady driving. A microchannel heat exchanger using a dielectric fluid could be integrated directly with the battery pack or used in an indirect cooling loop. The current results show that the choice of pin geometry and coolant affects both the maximum allowable heat flux and the pumping power.
For example, if HFE-649 is used with a diamond pin-fin microchannel, the vapor tends to accumulate in the sharp corner regions. This accumulation reduces the critical heat flux and increases wall-temperature nonuniformity. In an electric vehicle application, such nonuniformities accelerate battery aging and may even trigger thermal runaway. The circular pin-fin geometry should be preferred because it creates a more stable bubble-removal path. The droplet-shaped geometry is also attractive but had a lower critical heat flux than the circular geometry in my tests. Therefore, when designing a microchannel cooling plate for electric vehicle battery packs, the evaporation-side pressure drop and critical heat flux should be evaluated together rather than treating heat-transfer coefficient as the only indicator.
The fluid comparison between HFE-649 and Noah-2100A has direct implications for electric vehicle system design. Noah-2100A offers higher heat-transfer coefficients and a higher critical heat flux than HFE-649. However, its pressure drop is also higher. If the electric vehicle cooling system has a pump that can deliver the required flow and pressure head, Noah-2100A may enable a more compact cold plate. Its slightly higher GWP may still be acceptable because the absolute GWP value is very low compared with many hydrofluorocarbon refrigerants. This tradeoff should be evaluated through a system-level electric vehicle thermal-management model.
Conclusion
In this thesis, I investigated the enhancement of microchannel heat exchangers for electric vehicle battery thermal management through experimental observation, theoretical analysis, and computational simulation. The main conclusions can be summarized as follows.
First, flow visualization experiments in circular, diamond, and droplet-shaped pin-fin microchannels showed that the equivalent bubble diameter increases with mass flow rate. The largest bubbles occurred in the diamond-fin geometry, while the smallest bubbles occurred in the droplet-shaped geometry. Bubble size first increased and then decreased with heat flux because strong nucleation at high heat flux promotes early bubble departure and breakup.
Second, pressure drop in microchannel heat exchangers is strongly affected by pin shape and vapor behavior. The diamond geometry exhibited the highest pressure drop, followed by the circular geometry, while the droplet geometry had the lowest pressure drop. The pressure drop increased with mass flow rate and with bubble equivalent diameter. Large vapor slugs increase resistance and require additional pumping power in an electric vehicle cooling loop.
Third, the circular pin-fin geometry produced the highest critical heat flux among the three structures. At 1.4 g/s, the circular geometry reached 27.3 W/cm², which was about 43.3% higher than the diamond geometry and 15.1% higher than the droplet geometry. The circular geometry also gave the most uniform wall temperature distribution, making it the preferred geometry for electric vehicle battery thermal management.
Fourth, the comparison of Noah-2100A and HFE-649 in the circular microchannel showed that Noah-2100A has higher heat-transfer coefficients, higher pressure drops, and higher critical heat fluxes than HFE-649. Increasing mass flow rate increases heat transfer and pressure drop. Increasing inlet temperature has a weak negative effect on heat transfer but increases pressure drop. Increasing inlet vapor quality reduces heat-transfer performance and critical heat flux but increases pressure drop. The comprehensive thermal-performance ratio of Noah-2100A was above 1.0 over the entire range, indicating that Noah-2100A may be a promising alternative for high-power electric vehicle cooling.
Fifth, I developed empirical correlations for the Nusselt number and two-phase frictional pressure drop. The Nusselt correlation had mean absolute percentage errors below 12% for the tested fluids. After modifying the Chisholm parameter separately for each coolant, the pressure-drop correlations achieved mean errors of 4.9% for Noah-2100A and 6.2% for HFE-649. These correlations can be used in future design calculations for microchannel cold plates used in electric vehicle battery thermal management.
Finally, three-dimensional numerical simulations revealed the local mechanisms behind the experimental trends. The circular pins generated a smooth but sufficiently agitated flow field, promoting bubble departure and heat transfer. The diamond pins generated strong wakes and localized high-temperature regions. The droplet pins created the lowest pressure drop but also a somewhat lower heat-transfer enhancement. The simulated heat-transfer coefficients and pressure drops agreed with experiments within acceptable error. This validated model can be used to optimize new pin-fin shapes for even more demanding electric vehicle battery thermal management scenarios.
Future work should focus on extending these results to full-scale electric vehicle battery modules, dynamic load profiles, and fast-charging conditions. The interaction between multiple cold plates and the thermal behavior of complete electric vehicle battery packs should also be investigated. A reliable system-level model must combine the new correlations with cell heat-generation models to evaluate the thermal performance over the entire vehicle driving cycle. I hope that the data and correlations reported in this thesis will support the development of next-generation thermal management solutions for electric vehicles.