1. Introduction
Coatings modifying the emissive properties of surfaces are gaining more and more recognition in various applications including high-temperature nuclear reactors, photovoltaic solar cells, aerospace technology, military applications, the iron and steel industry, and the (petro-)chemical industry [
1,
2,
3,
4,
5]. In order to assess the impact of coatings on the radiative heat transfer in all these different applications, an accurate description and comparison of both uncoated and coated material emissive properties is required. Ideally, the classical electromagnetic wave theory can be used to predict all radiative properties, including emissivity, absorptivity, and reflectivity, of any material. However, the model constants required in this theory are difficult to determine. Moreover, the surface conditions, such as surface roughness, oxide layers, and contaminants, which are all known to influence the radiative properties, are not taken into account in the electromagnetic wave theory [
6]. Consequently, there is a need for an extended set of experimental data providing values for the emissive properties of a wide range of materials. However, no such database is available, with information on material emissive properties being spread over different publications and textbooks [
7]. Especially for materials encountered in steam cracking furnaces, namely high temperature alloys and refractory firebricks, experimental data in scientific literature related to emissive properties are scarce.
Interest in high emissivity coatings originated in the aerospace industry. The association with spacecraft and NASA surrounded the research area with an aura of “high tech” and “mystery” [
8]. Nonetheless, the underlying physical principle behind high emissivity coatings is straightforward—
it is not rocket science, except that it is. During the atmospheric re-entry of space vehicles, friction between the space vehicle and the atmospheric gases drastically heats up the surface of the spacecraft, which is termed aero-convective heating. Since the emissive properties of a radiating surface determine the amount of heat that can be dissipated back to the surroundings, increasing the emissive properties reduces the surface temperature [
9]. Emisshield licensed the high emissivity coating technology from NASA in 2001 to include all applications except for the space industry. In the last decades, high emissivity coatings started to make their way into the chemical industry. High emissivity refractory and metal coatings promise to reduce downtime, minimize coking, increase productivity, and to reduce the firebox’s overall carbon footprint by providing a uniform temperature distribution and an enhanced radiative heat transfer [
10].
More recently, there has been a renewed interest in emissive properties related to daytime radiative cooling of surfaces, resulting in several publications in high-impact journals [
11,
12]. In this passive cooling approach, the emissive properties of a surface are tuned to allow the surface to cool below the ambient air temperature, even under direct sunlight and without the need for electricity [
4]. In both examples of aerospace and daytime radiative cooling, by tuning the emissive properties, the “hot surface” emits radiation in a specific wavelength range thereby using the cold darkness of the Universe as a renewable thermodynamic source [
13,
14].
In the last few decades, various patents have been granted to industrial players that started to explore the possibility to use high emissivity coatings in the chemical, iron, steel, and glass industry [
15,
16]. These industries heavily depend on large industrial furnaces where radiation is the dominant heat transfer mechanism. However, there is a fundamental difference in the role of coatings between, on the one hand, industrial furnaces and, on the other, aerospace applications or radiative cooling. In industrial furnaces, the temperature difference between the emitting and receiving body is less pronounced, making it harder to select a wavelength range where emission or reflection should be optimized. For daytime radiative cooling, in contrast, it is clearly beneficial to have high emissive properties in the “sky window”, between 8 and 13 µm, where the atmosphere is nearly transparent, allowing the surface to emit thermal radiation to outer space. Additionally, the surface has to reflect as much radiation as possible in the solar spectrum, situated between 0.3 and 2.5 µm, for daytime radiative cooling applications [
17]. For gas-fired furnaces, almost all radiative heat transfer occurs in the thermal infrared wavelength range, 0.8–25 µm [
18]. Hence, the industrially used high emissivity coatings maximize the emissive properties over the entire wavelength range. With this approach, it becomes more likely that the energy emitted by the flue gas is absorbed and re-emitted by the furnace wall in a clear window of the flue gas, thereby heating the furnace load.
Various researchers have studied the effect of surface emissive properties on the energy efficiency of industrial furnaces. Early research stressed the importance of both clear and absorbing wavelength bands when modelling radiative heat transfer in the presence of flue gases in order to accurately evaluate the effect of the furnace wall emissivity [
19,
20]. Ward et al. developed a spectral model to predict the transient performance of a metal reheating furnace using a zoning method. Increasing the refractory furnace wall emissivity resulted in energy savings of approximately 3% [
21,
22]. Heynderickx et al. and Stefanidis et al. performed three-dimensional Computational Fluid Dynamics simulations of a naphtha cracking furnace with varying furnace wall emissivity [
5,
23,
24,
25]. Adams et al. performed similar Computational Fluid Dynamics simulations to estimate how high emissivity coatings can reduce the fuel firing rate in combustion-heated process furnaces [
26]. Spectral radiative heat transfer simulations show that increasing the refractory wall emissivity from 0.4 to 0.9 resulted in a 1.5% increase in furnace efficiency. In the same work, energy savings observed in industrial furnaces, after applying a high emissivity coating, are reported ranging from 8% for glass melting furnaces to 5 and 6% for ethylene dichloride and steam cracking furnaces, respectively. More recently, Yi et al. developed an analytical expression to study the effect of furnace wall emissivity on the heating process of a steel slab furnace, stating that high emissivity coatings might even have a negative impact, hindering the radiative heat transfer [
27].
In parallel to these numerical studies, various experimental studies have been conducted to assess the impact of high emissivity coatings on radiative heat transfer in furnaces. Hellander et al. reported energy saving exceeding 17% for an annealing furnace [
28]. Benko et al. reported experimental energy savings in the range of 4.8–6.2% and 3.8–4.4% for natural gas- fired and oil-fired furnaces, respectively [
29]. Svantner et al. investigated the influence of furnace wall emissivity on steel charge heating, thereby stressing the importance of the emissive properties in different wavelength bands [
30]. The significant spread in reported energy savings, shown in
Table 1, makes it challenging to predict the exact effect of applying a high emissivity coating on the energy efficiency of a specific furnace.
This work aims to develop a more straightforward simulation framework to assess the effect of reactor coil and furnace wall emissive properties on the radiative heat transfer in gas-fired furnaces. Recent scientific work confirms that in some contexts it is sufficient to perform radiative heat transfer simulations of simplified geometries. Often, the numerical study of simplified geometries suffices to give insight in radiative heat transfer characteristics. When time consuming Computational Fluid Dynamics simulations of more complex geometries are avoided, more elaborate parametric studies can be performed [
22,
31,
32]. For example, Hu et al. revisited Hottel’s zoning method for the firebox modelling of a steam cracking furnace and confirmed that this computationally less expensive approach provides reliable input for the first design steps of new furnaces or for the optimization of existing furnaces [
33].
The subsequent section emphasizes the different aspects of “emissivity” before gathering experimental data describing the emissive properties of uncoated and refractory firebricks. In the next step, the effect of a refractory coating on the firing rate of a steam cracking pilot unit is tested experimentally. Additionally, a parametric study is performed to evaluate the effect of the boundary wall emissive properties on the radiative heat transfer in a so-called tube-in-box geometry.
2. Terminology
The three fundamental modes of heat transfer are conduction, convection, and radiation. Both conduction and convection require the presence of a medium in order to transfer energy. Thermal radiation is transferred by electromagnetic waves, which implies that no participating medium is required. Another distinguishing feature between conduction and convection, on the one hand, and thermal radiation, on the other, is the difference in their temperature dependency [
34]. In most applications, conductive and convective heat transfer rates are linearly proportional to a temperature difference [
34]. In contrast, the Stefan-Boltzmann equation for a blackbody radiator expresses that the total emissive power of a blackbody,
, has a biquadratic relation to the blackbody temperature
:
where
is the Stefan-Boltzmann constant. This biquadratic relation indicates that radiation will be the predominant mode of heat transfer in high-temperature applications. In reality, no material behaves as a perfect blackbody. When an electromagnetic ray strikes an opaque material surface, the ray is either reflected or absorbed. Kirchhoff’s law of thermal radiation states that a surface at thermal equilibrium emits the same amount of energy as it absorbs. This statement, just like Kirchhoff’s laws for electric circuits, is just a rephrasing of the universal law of conservation of energy. The most general definition of “emissivity” describes emissivity as the ratio of the actual material surface irradiance to the irradiance of a perfect blackbody. The emissivity, in other words, defines the deviation of surface irradiance from a perfect blackbody, resulting in a value between zero and one, with one indicating blackbody behavior. Throughout this work the term “emissivity” is avoided as much as possible, even though it is frequently used throughout scientific literature and standard textbooks. For example, in the introduction of this paper, the term “emissive properties” was used to replace the general term “emissivity”. When discussing and comparing emissive properties, it is essential to clarify which emissive property is being referred to. For this reason, the next section is dedicated to introducing the different aspects of “emissivity”.
The most fundamental radiative property for emission from a surface is the
spectral directional emissivity defined as:
which compares the actual
spectral directional irradiance,
), as shown in
Figure 1, to that of an ideal blackbody,
, at the same temperature and under the same geometrical and spectral conditions [
34].
Spectral directional emissivity does not only depend on the temperature
, wavelength
, polar angle
, and azimuth angle
, but also on the surface roughness and on the chemical composition of the surface. These last two variables are harder to quantify and are locally varying surface properties. In order to compare experimentally measured emissive properties of different samples, the National Institute of Standards and Technology (NIST) proposed to use the term
emissivity for optically smooth and homogeneous materials and to use the term
emittance for rough and contaminated surfaces [
35]. However, these terms are not generally used. In this work, the term
apparent emissivity is introduced to define the experimentally observed emissivity of real surfaces.
Intrinsic emissivity is used to refer to perfectly smooth materials [
36,
37].
The most fundamental emissive property, the
spectral directional emissivity, Equation (2), is hardly ever measured experimentally since it depends on both wavelength and polar coordinates. A lot of expensive, moving instrumental parts are required to scan the
spectral directional emissivity experimentally. The most recent experimental units measure the
spectral normal emissivity [
38,
39,
40]. In this case, the irradiance is captured by a detector right in front of the sample. In other words, the detector is placed in a way that both polar coordinates become zero, i.e., detector placed perpendicular to the surface. Accounting for Equation (2), the
spectral normal emissivity,
, can be defined as:
The experimentally measured spectral normal emissivity is thus an apparent value of an actual surface. However, the spectral normal emissivity approaches the intrinsic value for very smooth surfaces.
Integrating the
spectral normal emissivity, Equation (3), over the entire wavelength range using the blackbody irradiance,
, governed by Planck’s law, as a weight function gives the
total normal emissivity,
, at a given temperature:
Since most
spectral normal emissivity, Equation (3), measurement devices only measure part of the entire electromagnetic spectrum, the term
band normal emissivity, Equation (5), is typically introduced to emphasize this wavelength band dependency. The
band normal emissivity for a given wavelength band
at a temperature,
, is given by the following definition:
When the band normal emissivity, Equation (5), is evaluated over the entire experimental wavelength range, the concepts total normal emissivity, Equation (4), and band normal emissivity, Equation (5), are often used interchangeably. This is considered to be a valid approximation when most of the radiation is emitted in the experimentally covered wavelength range .
When the
spectral directional emissivity, Equation (2), is integrated over both the entire hemisphere and over the entire wavelength range using the blackbody irradiance as a weight function, the
total hemispherical emissivity is calculated:
The
total hemispherical emissivity, Equation (6), is the most frequently reported emissive property and can easily be measured using calorimetric emissivity measurement devices [
2,
41]. The American Society for Testing and Materials (ASTM) has even developed a standard test method, ASTM C835-06, for the
total hemispherical emissivity of surfaces up to 1673 K [
42].
4. Results and Discussion
4.1. Spectral Normal Emissivity Measurements
The spectral normal emissivity measurement results of the pilot unit refractory material before and after a high emissivity coating is applied are shown in
Figure 5. Since the spectral normal emissivity measurement apparatus at the New Technologies Research Center operates in an air atmosphere, the experimental uncertainty, also shown in
Figure 5, is higher in wavelength bands where atmospheric gases interfere along the optical path [
51]. At the shortest wavelengths, below 5 µm, the spectral normal emissivity of the high emissivity coating is significantly higher compared to the uncoated refractory results. At longer wavelengths, the uncoated sample slightly outperforms the coated sample. However, at high temperatures, e.g., in steam cracking furnaces, proportionally more energy is emitted at lower wavelengths in accordance with Planck’s law of thermal radiation. The step-like spectral normal emissivity reported for the refractory substrate matches the limited data available for silica-alumina firebricks [
52,
53].
A more quantitative comparison between both spectral normal emissivity measurements is possible by introducing the band normal emissivity. The band normal emissivity is calculated by integrating the spectral normal emissivity over the entire experimental range while accounting for the spectral density at a certain temperature specified by Planck’s law, Equation (5). Since most of the thermal radiative heat transfer takes place in the experimentally covered wavelength range, the band normal emissivity approximates the more widely reported total normal emissivity, Equation (4). Weighted integration, accounting for the wavelengths where proportionally more energy is emitted, yields a spectral normal emissivity of 0.79 for the high emissivity coating compared to 0.67 for the uncoated refractory at 873 K; the coating emissive properties exceed those of the base substrate.
If the
spectral normal firebrick emissivity measurement results, shown in
Figure 5, are assumed to be temperature independent, the
total normal emissivity can be extrapolated to a wider temperature range as shown in
Figure 6.
The total normal emissivity results confirm that the coating outperforms the base substrate at higher temperatures when proportionally more radiation is emitted in the short wavelength region where the spectral normal emissivity is relatively low for the base substrate. Due to the poor thermal conductivity of the refractory substrate only measurements up to 873 K are performed. The extrapolation assumes that the spectral normal emissivity is temperature independent.
Figure 7 shows the emissive power in accordance with Planck’s law for a perfect blackbody and both an uncoated and a coated firebrick. The emissive power spectrum is yet another representation based on the spectral normal emissivity measurement results previously shown in
Figure 7. The flue gas absorption-emission bands are introduced for the first time in this work. The band upper and lower limits for each band have been taken from Zhang et al. [
18]. The participating species, water, and carbon dioxide, present in the flue gas of gas-fired furnaces, emit and absorb radiation in these wavelength ranges. These wavelength ranges are defined by the rotational and vibrational molecular frequencies of the participating species. Based on the emissive power spectra at 873 K, shown in
Figure 7, the uncoated firebrick emits 67% of its total emitted radiation in the clear windows in contrast to the coated refractory where 74% the total emitted radiation ends up in the clear windows. At a more relevant temperature for combustion applications of 1573 K, 85% of the emitted radiation ends up in the clear widow for uncoated refractory. Applying a high emissivity coating on the firebrick increases the fraction of radiation that is emitted in the clear window to 88%. The approach of visualizing emitted radiation in specific windows highlights that there are energy savings are possible when applying a high emissivity coating. Radiation can be emitted from the furnace wall to the heat sink, i.e., reactor coil, in a clear window where it cannot be re-absorbed by the gas phase.
An alternative way of designing coatings would be aimed at maximizing the emissive properties only in the clear windows and to selectively tune the emissive properties to reflect radiation in the absorption bands. This alternative approach would work similarly to coatings used in daytime radiative cooling applications as explained in the introduction, where emissive properties are enhanced in predefined regions. This way, all radiation, 100% in theory, would be emitted in the clear windows. However, one should not forget that the combustion gases are providing the necessary energy in gas-fired furnaces, not the refractory wall. The theoretical limiting case of emitting all radiation in clear windows also implies that all radiation coming from the flue gas in absorption-emission bands will be reflected in wavelength regions where it can easily be re-absorbed by the flue gases. To avoid this phenomenon, high emissivity coatings typically increase the emissive properties over the entire wavelength range, maximizing the possibility to absorb radiation in the absorption bands and re-emit the energy in a clear window. Here the role of “emissivity” is not only regarded as the raw emissive power of a body at a specific temperature, but also as a way to absorb and re-emit radiation.
In order to properly evaluate the effect of coating the refractory wall on the energy re-distribution in a gas-fired furnace, pilot unit experiments are performed and discussed in the following section. In a next step, radiative heat transfer simulations are performed in a tube-in-box geometry. In contrast to the percentages derived based on
Figure 7, not only emission but also reflection and absorption followed by re-emission of the participating surfaces will be modelled accounting for the surrounding gas. Initially, a non-gray gas model will be used to capture the gas phase absorption-emission characteristics. Next, a simplified gas model will be implemented. The subsequent discussion focuses on why only a non-gray gas implementation is able to quantify the effect of applying a high emissivity coating on the refractory wall and what this tells us about the role of high emissivity coatings on radiative heat transfer.
4.2. Pilot Scale Experiments
With the spectral normal emissivity results of (un)coated refractory, shown in
Figure 5, in mind, the effect of the same coating on the firing rate of a pilot steam cracker will be evaluated experimentally. Four experiments, two reference experiments and two experiments after applying a high emissivity coating on the refractory wall of the cracking cells, are performed with the same operating conditions and cracking severity as specified in
Table 2. The fuel, i.e., natural gas, consumption of the steam cracking pilot unit before and after the high emissivity coating is applied, is shown in
Figure 8. In all experiments, the natural gas consumption decreases over time to reach a steady-state after almost 6 h of cracking. The decrease can be attributed to the fact that the entire furnace initially needs to heat-up and find an equilibrium with the surrounding environment.
The fuel consumption at steady state is summarized in
Table 3. The application of the selected high emissivity coating results in a decrease in fuel firing rate of 4.8%, based on averaged values for the uncoated and coated runs when the system has reached a steady state.
Figure 9 shows the total flow rate of premixed burner gas, that is natural gas mixed with combustion air, going towards cells 3 and 4 of the pilot unit, where the cracking reactions take place in the suspended coil. No real energy savings are observed here. This implies that the reduced firing is mostly limited to the initial two preheating cells. However, there is no coating present in the initial two preheating cells and the process gas enters the cracking cells with the exact same temperature, indicating that the same amount of energy has been absorbed by the process gas in the preheating cells.
In contrast to the premixed burner gas flow meters, which are only installed in the cracking cells, the premixed burner gas pressure is measured for every cell. The premixed burner gas pressure is correlated to the premixed burner gas flow rate by Bernoulli’s equation and provides qualitative information about the distribution of premixed burner gas to the different cells, even for the preheating cells where no flow meters are present. The correlation between premixed burner gas flow rate and premixed burner gas pressure is visualized in
Figure 10 for cells 3 and 4 individually.
The premixed burner gas pressure for each experiment at the end of the cracking cycle, when the system has reached steady state, is shown in
Figure 11, for all four cells. The averaged burner gas pressure, and thus the premixed burner gas flow, increases in cell 3 and decreases in cell 2 after the high emissivity coating is applied. This suggests that for the coated experiments, the energy from the third cell is partly used to heat the adjacent preheating cell 2. This phenomenon can be explained by the geometric nature of the experimental unit. The cells are located directly next to each other which makes that a cell can overcompensate the burner gas flow rate for one of its adjacent cells. This explains why energy savings are visible when looking at the total natural gas flow rate,
Figure 8, but not apparent when comparing the premixed burner gas flow rate to the cracking cells before and after applying a high emissivity coating,
Figure 9. Based on the premixed burner gas pressure at steady state, shown in
Figure 11, the second preheating cell, cell 2, is underperforming. After the coating is applied in cells 3 and 4, the process control loop of the pilot plant operates in a suboptimal region where cell 3 is overcompensating for cell 2.
The pilot scale experiments show a fuel efficiency increase of 4.8%, based on the firing rate for the uncoated and coated runs when the system has reached a steady state. The behavior and control of the unit changes noticeably after applying a high emissivity coating with an increased premixed burner gas flow rate to the first cracking cell, cell 3, and reduced firing in the second cracking cell, cell 4. However, the total fuel consumed in the cracking cells remains constant before and after a high emissivity coating is applied. The main difference is the reduced firing in the last preheating cell, cell 2. Even on a pilot scale the application of a high emissivity coating results in fundamental changes to the process control. There results confirm that in order to properly evaluate energy savings, the entire process has to be examined, not limiting the observation to the cracking cells.
4.3. Tube-in-Box Radiative Heat Transfer Simulations
Based on spectral normal emissivity measurements, applying a high emissivity coating offers a way to increase the surface emissive properties. Detailed process simulations from scientific literature combined with the experimental results of the previous section suggest that the emissive properties have an impact on the firebox efficiency and the fuel firing rate. However, is it possible to quantitatively predict this effect without the need for an extensive three-dimensional Computational Fluid Dynamics simulation? The remaining part of this work will focus on the role of heat sink (reactor coil) emissive properties and enclosing adiabatic surface (furnace refractory) emissive properties when modelling radiative heat transfer.
Based on the pilot unit experiments, a reactor tube wall temperature,
, of 1173 K will be used as heat sink boundary condition for the tube-in-box model. For the initial simulations, conditions as shown in
Figure 4a are used. A constant gas phase temperature,
, of 1373 K is assumed and simulations are performed using different emissive properties
and
. In the next simulations, conditions are as shown in
Figure 4b. The gas phase temperature,
, is increased to realize the same net heat flux,
, to the inner tube, again using different boundary emissive properties
and
.
4.3.1. Gray-Plus-Clear Gas Simulation Results
In order to solve the tube-in-box radiative heat transfer problem, the total exchange area approach can be applied. The reactor tube emissivity,
, is varied from 0.7 to 1.0 (x-axis) and the refractory emissivity,
, is varied from 0.4 to 1.0 (y-axis) to give an almost complete overview of the effect of the emissive properties of the boundary surfaces on the radiative heat transfer inside the geometry. Calculated refractory wall temperatures are shown in
Figure 12 for both the constant gas phase temperature simulations,
Figure 12a, and for the constant heat flux simulations,
Figure 12c, using a gray-plus-clear gas model. The heat maps showing the net radiative heat flux for a constant gas phase temperature,
Figure 12b, and the gas phase temperature required to get a constant net heat flux,
Figure 12d, complete
Figure 12.
Keeping the gas phase temperature constant and changing the boundary emissive properties appears to have a notable effect on the net radiative heat flux towards the heat sink, shown in
Figure 12b. A higher tube wall emissivity,
, implies that more energy will be absorbed. This is clearly visualized by the last resistance before the heat sink in the “electric circuit analogy”, discussed in the
supporting information. A higher boundary wall emissivity,
, also results in an increase in net radiative heat flux towards the heat sink. An uncoated furnace wall with a higher probability of reflecting the incoming radiation, tends to reflect the radiation in an absorption band where it is re-absorbed by the gas, shown in
Figure 7. A higher emissivity material, e.g., a high emissivity coating, will increase the probability that the incoming energy is re-emitted in a transparent window. Radiation will be more likely reach the heat sink in a transparent window and thus heat the load. The effect on the refractory wall temperature is negligible when keeping the gas phase temperature constant, as shown in
Figure 12a. However, the refractory wall temperature,
Figure 12c, increases noticeably when increasing the gas phase temperature to get the same net heat flux towards the process gas. It is important to emphasize the biquadratic relation between temperature and emitted radiation. As can be seen when increasing the gas phase temperature in
Figure 12d, only a relatively minor increase is required to compensate for the offset caused when the boundary emissive properties are changed. Additionally, a change in gas phase temperature does not immediately translate into a major shift in firing rate. The gas phase temperature has to be increased with 95.7 K to offset a 32.2% decrease in net heat flux when lowering both the refractory wall emissivity,
, from 1.0 to 0.7, and the radiant coil emissivity,
, from 1.0 to 0.4 (difference between best case, high heat sink emissivity, and high adiabatic wall emissivity, and worst case-scenario, low heat sink emissivity and low adiabatic wall emissivity).
To correlate the gas phase temperature, shown in
Figure 12d when a constant heat flux is imposed, with an energy efficiency or firing rate, a complete furnace modelling approach including a combustion model would be required. In other words, a complete Computational Fluid Dynamics framework would be necessary to translate gas phase temperatures into a firing rate. This would considerably increase the computational cost. For the two-dimensional tube-in-box model the energy available to heat up the furnace load is determined by the enthalpy of combustion of methane minus the heat required to heat up the combustion gases to the equilibrium temperature. When the flue gas temperature is higher, less heat of combustion is available to the heat sink.
Figure 13 shows the relative amount of energy available to heat the heat sink for the gray-plus-clear gas simulations with a constant heat flux, by translating the gas phase temperatures mentioned in
Figure 12d to an energy efficiency using the previously described approach.
The main drawback when using a general weighted sum of gray gases approach, or the derived gray-plus-clear model, is that the band assumption is implicit. The radiative transfer equations are solved with a specific portion of the radiation transferred in the absorption bands. This portion,
, is calculated based on the gas phase temperature. In reality, radiative heat transfer is a bit more complicated as the fraction of total radiation transferred in a specific band also depends on the heat sink temperature and the surrounding adiabatic surface temperature. For this reason, the case study could be repeated with an exponential wide band modelling approach for the radiative transfer equation. The radiative transfer equation would in this case be solved in four absorption bands and five clear bands [
5,
24,
25,
53]. Since the exponential wide band model provides explicit upper and lower band limits, it would be possible to visualize the exact amount of radiation emitted and reflected by a surface in a specific band. These results visualize that in case a high emissivity coating is applied on the refractory wall, more radiation gets redistributed into the clear bands. As a result, the net heat flux to the coil increases when applying a high emissivity coating due to this redistribution of radiation into the clear bands.
4.3.2. Gray Gas Simulation Results
Figure 14 shows the results for the tube-in-box heat transfer simulations using a gray gas model. This time the adiabatic surface emissivity,
, plays no role in the radiative heat transfer. On the other hand, the effect of the heat sink emissivity,
, remains visible when using a gray gas model. Again, a higher tube wall emissivity,
, implies that more energy will be absorbed by the heat sink. Analogously, if the enclosing surface is not completely adiabatic but accounts for some heat losses, it is undesired to have a high furnace wall emissivity,
, when using a gray gas model. In this case, the resistance in the “electric circuit analogy” should be maximized, in other words the reflectivity should be maximized to prevent heat losses through the refractory wall to the surroundings. However, in the case of a perfectly insulated enclosing surface, the gray gas model solutions are independent of the adiabatic wall emissivity,
, as shown in Equation (10). With the gray gas approximation, all radiation striking the enclosing adiabatic surface will either be reflected or absorbed and re-emitted. Since there is no clear distinction between the reflected radiation, on the one hand, and the absorbed and re-emitted radiation, on the other hand, the enclosing surface can be regarded as a blackbody, whatever its emissivity is. The enclosing surface redirects all radiation towards the heat sink, which is by definition the fundamental property of a blackbody. A non-gray gas model is needed to distinguish between reflected and re-emitted radiation [
27].
Overall, radiative transfer characteristics of an opaque wall can often be described with sufficient accuracy by gray emission, absorption, and reflection. However, the radiative properties of a molecular gas vary strongly and rapidly across the wavelength spectrum. As a consequence, the gray gas assumption almost never suffices [
34,
54]. A non-gray gas modelling technique is required to accurately capture the radiative heat transfer characteristics of the gas. Even though the emissive properties of a high emissivity coating can often be described as a gray surface, a non-gray modelling approach is necessary to simulate the effect of the enclosing wall as a redistributor of radiation.
5. Conclusions
There is a clear linear correlation between incident or outgoing radiation and surface emissivity. The origin of high emissivity coatings is based on this fundamental principle, which allows space vehicles to dissipate more radiation to the surroundings by increasing the surface emissive properties. Analogously, high emissivity coatings have also been making their way into the chemical industry for emitting surfaces. For a heat sink, namely a surface envisioned to absorb radiation, increasing the emissive properties allows for energy savings. This way, the incident radiation can be lowered to maintain the same net heat flux to the heat sink, e.g., lower firing rate when increasing the steam cracking coil emissivity. However, this argument has long been used against the application of high emissivity coatings on refractory walls. Refractory material is by nature reflective in the infrared wavelength region, so it is not inclined to absorb radiation which would result in undesirable heat losses to the outside surroundings [
20]. However, when accounting for the absorption-emission spectra of flue gases, the opportunity of high emissivity coatings to absorb radiation with the possibility of re-emitting in a clear, non-absorbing window, makes research in the application of high emissivity coatings on refractory worthwhile. With a total normal emissivity of 0.79 for the high emissivity coating compared to 0.67 for the uncoated refractory, at 873 K, applying a high emissivity coating is found to significantly improve the emissive properties of a refractory surface. Steam cracking pilot unit experiments report a 5% reduction in firing rate after applying a high emissivity coating on the refractory wall. Radiative heat transfer simulations using a tube-in-box model show that the emissive properties of an adiabatic enclosing surface do not influence the temperature distribution in the presence of a gray gas. All energy is redirected towards the heat sink (tube) with the surrounding surface (box) acting as a blackbody. However, the gray-plus-clear gas model confirms that the furnace can be operated at a lower temperature when the emissive properties of an adiabatic surface are improved. Energy radiated by the hot flue gas is more likely to be absorbed by a surface and consequently re-emitted by that surface in a clear gas window when its emissive properties are maximized, resulting in energy savings.
The radiative heat transfer simulations show that the possible gains in terms of efficiency are a matter of percentages when changing the boundary wall emissive properties. This makes it statistically challenging to experimentally evaluate the impact on a pilot or industrial scale. In that case, Computational Fluid Dynamics simulations, preferably with a spectral radiation model, are the preferred option to accurately evaluate possible changes to the process after coating the reactor coil or the furnace wall. Changing the radiant coil emissive properties is expected to have the most impact on the overall process since proportionally more heat is absorbed. In this case, changing the emissive properties has a direct effect, not a secondary effect of redistributing radiation, which is the case when coating the refractory wall. However, similar to the lack of experimental data related to the emissive properties of refractory materials, more information should be gathered to quantify the difference in emissive properties between uncoated and coated high-temperature alloys.