Case Studies
Air Heat Exchanger—Thermal Fin Design/Optimization
The first case study looks at thermal fin design for compact heat exchangers applied to indoor air conditioning units. Complex articulated patterns are used today, pressed in thin aluminum fins to promote heat transfer in such units, by acting on the air boundary layer formation (interruption), affecting the transition to turbulent flow and the fin tube wake area on the fin surface which is otherwise a poorly utilized heat transfer area. Fin designs continue to be studied today both numerically and experimentally, even for a technology which is arguably very mature. This is due to the need to continuously improve energy efficiency in space heating or industrial applications, which can be realized through changes to the fin configuration (adding vortex generators has been studied for many years) or more fundamental changes to the heat exchanger tube shape/layout. In this study (full details can be found in [8]), various experimental setups were used; the focus of this section is on the wind tunnel experiments with its scaled-up model of the fin configuration.
The intent of this experiment was to determine both the global averaged heat transfer coefficient of the fin assembly, the local visualization of the boundary layer on each of the louver elements and to explore the impact of key fin parameters such as the fin spacing and louver angle. As the louvers are normally in the order of 1 mm in length, the setup considered a scaled- up version of the fin geometry to enable access to the louvers with thermocouples. The tube wall was not considered in this study, thus avoiding the complexity of the wall-bounded vortices and their scale-up. In order to study the tube effect on the louvered fins, the heat exchanger tubes would need to be included in a sufficient number to ensure periodic effects, as illustrated, for example, in [7].
Figure 1.1a schematically shows the wind tunnel, which consists of a fan sending air through the scaled-up fin sample (as shown in Figure 1.1b). The dimensions of the test setup were selected based on a trade-off between scaling-up of the fin design (roughly 12/1 to place up to 10 thermocouples on a single louver measuring wall temperatures on both sides of the louver) and the available space in the laboratory considering the air mass flow metering requirements through an orifice plate. This measurement requires sufficient length of undisturbed flow. The total test setup length exceeded 8 m. Uncertainty analysis indicated that mass flow (Reynolds number) was a key property to get right, so rather than relying on the operating point of the fan, the flow rate is measured with up to <2% accuracy through an orifice plate (designed to ISO standard 5167). Computational studies were performed to assess the flow behavior and deflection in the fin structure to avoid the impact of the channel walls on the central fin structure where
FIGURE 1.1
(a) Experimental test setup used to study a scaled-up fin configuration; (b) inclined louvered fin geometry studied in the test section (7) showing three fin rows [9].
measurements would take place. The results showed that at least six fin rows would be needed to ensure periodic-like boundary conditions were present, which in turn constrained the maximum fin spacing that could be studied.
The wind tunnel was designed to provide flow with a low turbulent intensity by use of a diverging/converging section with, at its center, a 50-mm- thick honeycomb structure, with a cell size of ~10 mm. At the connection between the fan and the wind tunnel, a grate with circular drilled holes was inserted to ensure an even flow distribution. This was a common laboratory practice used to adjust for maldistribution in the fan exit plane. To even out the flow distribution, the holes were machined in such a manner to add more resistance locally where the flow rate was too high and to smoothen the velocity profile. In this specific case, the applied hole sizes were 10,16, and 25 mm, set in vertical even spaced rows. The verification of the inflow conditions was done through 2D hot-wire anemometry. The results are presented in [9] for reference.
The local heat transfer coefficient hloca[ is defined in Eq. (1.1). The test setup was built to measure the local wall temperature on the louvers while imposing a constant surface heat flux q" through electrical current heating. The local fluid bulk temperature was computed based on thermal balance considering the upstream heat input, but other reference values, e.g., the inlet temperature, can be used as well. The louvers were manufactured from balsa wood, which is lightweight and easy to handle. It also is electrically and thermally insulating, which is important considering that the temperatures at the top and bottom of a louver would differ due to flow phenomena such as impingement or local wakes. Figure 1.2 shows the cross section of the measurement louver, [10]. Thin (0.25-mm) К-type thermocouples were selected to measure the surface temperature. To mount these, the thermocouple junctions were curved upward before sliding them into the grooves cut into the balsa louver. The thermocouple end sections were then taped to the wood, resulting in the junctions curving upward and standing some 2 mm above the louver surface. This was done to ensure a good contact between the junctions and the metal copper foil (0.25 mm thick) which was placed
FIGURE 1.2
Cross section of the instrumented louver [9].
on top to provide electrical heating. Once the metal foil was glued to the balsa wood, the electrical connections were added on one side. Current was sent through the foil to set the heat flux, which is calculated based on the measured foil resistivity. To verify the heat flux was uniform, an IR image was recorded of the louver while being heated, which showed an even temperature profile apart from near the electrical connections at the edge of the louvers.
The heat balance over the louver bank was verified using temperature measurements; both upstream (4) and downstream (12-16) of the test setup, К-type thermocouples were spaced evenly across the channel cross section. The bulk averaged temperatures were calculated using the measured velocity profiles (hot-wire) as weighing factor. This provided sufficient data for a closure of the heat balance to within 5%. All thermocouples were mounted with their junctions pointing to the direction of the flow; no impingement heating occurred due to the low air velocity present in the wind tunnel, even at the highest Reynolds number. The thermocouples had to be supported by a metal grid structure to ensure they did not vibrate at certain velocities, improving data quality.
At the lowest air velocities, the temperature data recorded downstream of the test section showed higher temperatures at the top of the channel compared to the bottom. This showed how buoyancy started to affect the results, now operating in the mixed convective regime rather than forced convection, [10]. By scaling the heat flux down, this effect could be dampened; however, this cannot be completely mitigated. It is important to consider the implications of natural convection on test design and execution as it may drive the need to lower the surface heat flux significantly at lower Reynolds number, which affects the feasible temperature differences and thus the relative accuracy of the results. This is often overlooked early in the design phase of experiments.
Expanding Gas Jet—Impingement Cooling on Pipe Walls
This case study looks at impingement cooling of expanding gas flows (e.g., downstream a valve) in natural gas facilities. Due to Joule-Thomson cooling, when natural gas (mainly methane) expands from a high pressure, significant cooling will occur. These low temperatures get transferred to the pipe wall, which can then result in local embrittlement of the metal. This is a safety concern, as a brittle fracture will occur suddenly and result in a catastrophic loss of containment releasing gas into the environment. An experiment was set up to perform detailed measurements of the velocity and temperature field (planar in both cases) through particle image velocimetry (PIV) with smart tracers coated with a phosphor material. A dual-laser system was used, where one of the lasers provided the PIV data through local illumination of the target area (as per standard PIV practice taking two images in fast succession) and the other laser was attuned to the excitation frequency of the phosphor. By then taking high-speed images following the excitation laser pulse, the decay of the phosphorescence could be measured, which is a measure for the particle temperature. Through this dual-laser approach, both local temperature and velocity can be measured throughout the plane of interest. More details on the experimental setup can be found in [11]. Phosphor-based temperature measurements inside fluid flow is a developing field ([12]), and it offers a way to look closer at fluid and temperature dynamics in complex (reacting) flows. The experimental setup, however, becomes sensitive and complex, requiring optical access, controlled particle seeding, laser timing control, and well-defined environmental ambient/background conditions to allow for post-processing.
The test setup (Figure 1.3) was designed to mimic natural gas flow conditions and achieve fully turbulent flow inside the channel downstream of the gas expansion, while still providing a sizeable interrogation area for the PIV to record at high frequency (3 kHz). In addition, sufficient expansion cooling was required across the expansion to ensure the uncertainty on the temperature data was sufficiently low to calculate the local wall heat transfer coefficient. Various fluids were screened, and argon (stored in 300 bar cylinder) was selected as the working medium, expanding from 130 bar to atmosphere across a small round orifice opening (1.5 mm). The tube section was made square (50 mm) from aluminum, which allowed for optical access from three sides as required to generate the laser sheet and record the flow images. However, the square tube complicates the flow patterns, but this choice was accepted due to the challenge of light reflections/scatter through a circular glass wall which would have been the geometrically similar configuration. The experiment was conducted in transient mode, running for a
FIGURE 1.3
Jet impingement test setup [11].
period of 1-3 min triggering wall cooling. The transient mode was selected due to the seeding particles depositing inside the channel and causing high local brightness on the recorded data—hampering data quality. This effect was originally overseen in the design of the test setup requiring further modification. The working pressure level of 120 bar upstream of the orifice was established as a trade-off between experimental duration (1-3 min) and having sufficient cooling while avoiding depletion of multiple high-pressure gas bottles in a single run.
Analytical assessment of the gas expansion inside the cylinder revealed that as the gas expanded from 300 to 120 bar, it also cools down inside the cylinder. The expanding gas absorbs heat from the cylinder wall and the laboratory surroundings, which would result in temperatures of around -17°C upstream of the orifice for most of the duration of the experiment. However, the measurement setup and seeding equipment with associated valves (items 2-6 in Figure 1.3) are a large metal mass storing energy. This combined would result in a large thermal transient at the inlet of the orifice as the argon flow is cooling down in essence the rig, and this time is then "lost" due to highly non-uniform conditions. To counter this transient, the entire upstream end of the test setup was placed into a large refrigerator and kept at a temperature of—17°C and all piping connecting the channel was insulated. T-type thermocouples were used throughout this setup, measuring the temperature just upstream of the orifice as well as at various locations on the channel wall. Preconditioning of the unit in this manner enabled controlling the transients in short-duration experiments (<15 s to reach a near steady state) and provided a more constant inlet temperature as boundary condition. A fast-acting pressure regulator was selected to provide control of the mass flow entering the channel. No direct mass flow measurement was added to the system as the critical flow through the orifice provided a means to determine the rate based on the upstream pressure and temperature conditions which were measured.
To determine the local heat coefficient, the local surface temperatures were measured on the inner wall using a phosphorescent dye, which showed a uniform temperature distribution. To avoid the impact of axial conduction in the aluminum wall, the section where the inner wall temperature was measured consisted of a polymer insert which was insulated at the back. On the outer channel wall, T-type thermocouples were placed every few centimeters to track the wall cooling through the experiment (and indirectly provide an estimate of the through-wall heat flux). These thermocouples were connected using copper-based tape to ensure a good contact between the junction and the wall, and there were clear differences in the time response of these probes when comparing the polymer and aluminum wall section. Numerical work had shown that axial conduction would smoothen the local wall cooling. A surface averaged heat transfer coefficient was determined based on the rate of temperature decrease measured on the inner wall, and it was found to be significantly higher than that predicted by standard turbulent flow correlations due to the local impingement. Flow visualization including schlieren images further revealed that a strong back-flow area appears around the jet, which results in locally strong variations of the boundary layer and heat transfer, as illustrated in [11].
The thermal response of the seeding particles was calibrated through a dedicated setup, where small seeding rates were applied in a gas stream which impinged on a thin metal surface. The carrier gas was nitrogen, a stream taken from stored bottles and mixed with a second stream of boiling liquid nitrogen to provide a set temperature ranging from -120°C to 0°C. An exponential calibration curve was then established based on the measured phosphorescent response, as this was the physically expected behavior. The temperature of the back of the metal surface which was in the flow was measured through a thermocouple as direct input into the particle calibration curve.
Through careful design and calibration, it was possible to generate high- quality data on the jet flow in terms of both velocity and local temperature, which have been compared to numerical predictions. The closure of the heat balance was achieved to within 10% in this case through enthalpy-based calculations using the upstream conditions and transient heat lost by the wall during the experiment.
As the setup cooled down, condensation droplets started to appear on the optical access windows, which resulted in loss of signal. To resolve this, a continuous N2 purging was set up on the windows to keep it locally clear of droplets. This was originally not foreseen in the design and is an example of a challenge with optical access. At high temperatures, other challenges appear, such as local soot deposition/blackening. Seed particle deposition in walls and crevices resulted in high signal back to the detector; to reduce this, firstly all windows were installed flush mounted with the walls, avoiding a backward facing step which would otherwise accumulate particles, and secondly, after every run, the test setup was carefully cleaned with demineralized water and ethanol to reduce the tendency for particles to deposit. Metallic surfaces which could cause a local glare in the field of view were taped off with a matt black tape. This laboratory practice resulted also in less risk of scattering laser beams as sharp edges were taped off, which improved overall laboratory safety.
