Influence of AM Generated Burner Surface Roughness on NO x Emissions and Operability of Hydrogen-Rich Fuels

The additive manufacturing (AM) technique enables the fabrication of advanced burner components to enhance the hydrogen capability of the existing gas turbines (GTs) and reduce the carbon footprints of the power generation sector. This technique produces rough surfaces that may require post-processing to maintain the desired functionality of a burner, particularly for hydrogen fuel with unique thermo-physical properties. This study, therefore, compared the stability of three swir-lers of variable surface roughness manufactured using AM and traditional machining methods, with one of the AM swirler post-processed by grit-blasting. The comparison included a conventional benchmark (100% CH 4 ), low carbon (23% vol CH 4 /77% vol H 2 ) and zero carbon (100% H 2 ) fuels across a range of equivalence ratios. Additionally, the study quantified the flame topology and emissions performance of the fuel blends for each swirler using high-speed OH* chemiluminescence and exhaust gas emissions measurements, respectively. The experimental investigation concluded that the AM-generated surface roughness within the considered range does not detrimentally impact NO X emissions and the stability of the fuel mix. However, the flame location was observed to be influenced by surface roughness and shifted more toward the vertical centerline of the burner with increased roughness. From the practical perspective, the results showed that post-manufacturing surface finishing offers negligible performance advantages, indicating potential cost reductions. It is recommended that further studies should investigate the influence of increased surface roughness on burner performance, as well as numerical modeling techniques which could provide an insight into when AM surfaces are likely to be more influential.


Introduction
Additive Manufacturing (AM), also known as Additive Layer Manufacturing (ALM) or 3D printing, has been introduced as a pioneering manufacturing technique over the past 10 years.It has the potential to revolutionize industries with the power generation and biomedical sectors being two notable examples.This innovative manufacturing technique enables the fabrication of enhanced structures of complex geometry, previously not possible through subtractive methods.Additionally, it offers several advantages, including multiple-component integration, rapid prototyping, freedom of design, minimization of material waste, and lead time, as well as multifunction components of novel internal structures, making AM superior to conventional manufacturing techniques (Ngo et al. 2018).In the biomedical sector, these advantages promote the fabrication of customized end-products, through more efficient and simplified supply chain systems of minimal waste, greatly improving quality of life and sustainability index (Velu et al. 2020).In the context of the gas turbine (GT) industry, AM also has the potential to address challenges linked to material properties and combustion inefficiency, associated with flashback, unburnt fuel, elevated emissions, and combustion instabilities (ETN Global 2020).Therefore, AM could greatly contribute to sustainability by promoting the utilization of low and zero-carbon fuels (e.g., H 2 , NH 3 , NH 3 /H 2 /CH 4 blends) to enable a smoother transition from fossil-fuel-based heat and power toward a more renewable supply, with lower harmful greenhouse gas emissions and other gaseous pollutants, such as NO x emissions (ETN Global 2020, National Academies of Sciences, Engineering, and Medicine 2020).
Hydrogen has been gaining attraction as an alternative to fossil fuel use in automotive and GT systems.It can be produced from various renewable or nonrenewable feedstocks and chemical processes and serve as an energy storage medium for carbon-free power generation.Hydrogen supports the electrification of transportation through Fuel Cell Electric Vehicles (FCEVs) and power-to-gas economy through "Green" hydrogen.It can be used to generate electricity or it can be blended with other carbon sources to produce fuels such as syngas or methane (Goldmeer 2019;Jayakumar et al. 2022).For the given potential of AM applications to hydrogen utilization, GT market-leading Original Equipment Manufacturers (OEMs) have already invested in utilizing metallic AM technology for the development and seamless manufacturing of novel GT components.This secures the role of GTs in energy transition and renewable energy installations (ETN Global 2020National Academies of Sciences, Engineering, and Medicine 2020; Runyon et al. 2021).Numerous examples of AM development have accommodated hydrogen utilization both in academia (Fan et al. 2021;An et al. 2021) and in industry (Larfeldt et al. 2017;Patel 2018;Prandi 2019;Walton 2021).Projects such as those by the US Department of Energy and ongoing European projects, aim to demonstrate the performance of AM combustion systems and components in high H 2 /syngas (York et al. 2015), and zero-carbon H 2 /NH 3 GT (UK Research and Innovation 2020).Other examples include AM-based burner tip temperature improvements for high H 2 operation and innovative AM nozzle designs and fuel injector patents, regarded for hydrogen combustion applications (Runyon et al. 2021).
Furthermore, a unique advantage that AM offers is the fabrication of specific, predefined surface roughness, depending on the selected build parameters employed during manufacture.As rough surfaces can influence boundary layer fluid flow and hence the operation and the efficiency of the GT systems significantly, the investigation of surface texture, as a function of building parameters and resultant surface roughness, has become a focused area of research (Bons et al. 2008;Mumtaz and Hopkinson 2010).The alteration of the roughness texture of GT blades and stator due to prolonged operation, fuel deposits, corrosion, erosion, and thermal barrier coating constitute some of the most common issues associated with surface roughness and GTs.A review of these and related issues may be found in Bons (Bons 2010).Consequently, due to the importance of surface roughness and its synergy with GT operation, the influence of the former on the latter has recently gained scientific interest (Al-Fahham, Bigot, and Valera Medina 2016;Crayford et al. 2019;Hatem et al. 2017;Runyon et al. 2019).
The investigation of the effects of surface roughness on fluid flows started over a century ago (Darcy 1857;Fanning 1886).Since then, experimental and numerical studies of the effects of surface roughness on heat transfer, as well as on isothermal and reacting flows, have been conducted, as reviewed thoroughly by Kadivar et al. (Kadivar, Tormey, and McGranaghan 2021).In summary, when the roughness height is above an "admissible" value, it interacts with the boundary layer modifying its characteristics.Thus, the transition of the boundary layer from laminar to turbulent, as well as the separation susceptibility of the flow, are influenced, ultimately affecting the overall form drag (Schlichting and Gersten 2014).Furthermore, surface roughness often increases skin-friction (Moody 1944;Nikuradse 1950), consequently increasing the overall pressure drop, which is a function of both skin-friction-related drag and form-related drag (Schlichting and Gersten 2014).However, it has been proposed that surface roughness potentially has a positive impact on GT performance.Several studies have shown that "manufacturable roughness" can potentially improve the performance of GT components by enhancing its heat transfer and aerodynamic characteristics (Dean and Bhushan 2010;Domel et al. 2018;Li, Guo, and Huang 2020;Liu et al. 2020).Analogous to the dimpled surface of golf-balls, biomimetic textures, such as shark-skin, have been found to reduce aerodynamic drag whilst enhancing heat transfer due to promoted turbulent mixing near the wall (Al-Fahham et al. 2017;Dean and Bhushan 2010;Hatem et al. 2017).In the context of combustion instabilities, the introduction of such a geometry in a generic swirl burner resulted in reduced boundary layer flashback propensity without modifying the bulk geometric characteristics of the burner (Al-Fahham et al. 2017;Hatem et al. 2017).
As for conventionally manufactured GT parts, AM-manufactured ones that fail to meet the desired functional criteria upon printing, often subject to post-processing, such as gritblasting, to reduce the surface roughness height and ultimately improve surface quality (Lu et al. 2023;Sinha et al. 2022).However, since this activity increases total lead-time and cost of production, both profitability and the necessity of such action is an active area of interest (Liu et al. 2023).Of particular interest to the present study is the experimental campaign carried out by Runyon et al. (2019), in which two additively manufactured swirlers (of which, one was post-processed via grit-blasting, and one was left raw) of different average surface roughness height were compared against a conventionally manufactured "smooth" swirler.The two AM swirlers, which were tested under 100% CH 4 combustion and atmospheric pressure conditions, resulted in improved combustion performance, with respect to NO x emissions, whilst also being found to affect flame stabilization location.However, although there is scattered evidence of the potential of "manufacturable roughness" derived from AM, the investigation of the influence of surface roughness on specific combustion phenomena is still overlooked.
Motivated by the aforementioned challenges, the current status of synergy between AM and GT, and a previous experimental campaign (Runyon et al. 2019), the present study aims to gain an empirical understanding of the influence of surface roughness and its postprocessing requirements on emissions characteristics and combustion performance of a generic AM swirl burner fueled on blends ranging from conventional methane to pure hydrogen across a range of equivalence ratios.

Methodology
The experimental program aimed to investigate the effect of surface roughness on leanpremixed swirling flames under atmospheric pressure and elevated inlet temperature conditions.A first database of well-controlled experiments was generated at Cardiff University's Gas Turbine Research Centre (GTRC).
The study investigated three different fuel blends, including conventional (100% CH 4 ), low carbon (77% vol H 2 /23% vol CH 4 ) and zero carbon (100% H 2 ) fuels.Hydrogen-rich fuel blends were selected to analyze the interaction of surface roughness given premixed hydrogen flames are prone to intrinsic thermo-diffusive instabilities (Berger, Attili, and Pitsch 2022), impacting flame shape, heat release, and flame speed.In this study, the CH 4 /H 2 ratio in the low carbon fuel blend was determined by considering equal contribution of each fuel to the total thermal power.
To avoid geometric effects on emissions and combustion characteristics of the fuel blends, nominally identical three-dimensional generic swirl burners of different surface roughness were manufactured (Figure 1).Two AM swirlers were deployed, one "gritblasted" (AM-G) and one "raw" (AM-R), with the average surface roughness height roughly 5 μm and 9 μm, respectively, together with a traditionally manufactured "smooth" swirler (Machined), with an analogous value of 1 μm.Further details on the surface roughness measuring set-up and geometry of the swirlers are found elsewhere (Runyon et al. 2019).
The experimental investigation of the conventional methane flame was used as a benchmark case of the test matrix for the comparative appraisal of low-carbon/hydrogenrich fuels.In all cases, the thermal power output was kept constant at 25 kW.Thus, only airflow was adjusted to produce necessary changes in the equivalence ratio.Additionally, the airflow determines the bulk flow velocity of the mixture and provides a broad range of Reynolds numbers to examine the influence of surface roughness on flow characteristics.Surface roughness effects were examined through a combined analysis of data involving the operational flame stability map, exhaust emissions, and temperatures.The flame stability map (Lean Blow Off (LBO) and FlashBack (FB) limits) were established over a range of equivalence ratios at inlet temperatures of 150 ± 5°C.The burner geometry limited optical excess to the flame root in the nozzle section, hence the flashback limits were evaluated based on the measurement of the nozzle surface temperature and visual observation of flame stability over the nozzle.To obtain independent evidence of surface roughness effects, the study employed integrated diagnostic tools, including continuous gaseous emission sampling (NO x and O 2 ), systematic measurements of single-point gas temperatures and high-speed OH* chemiluminescence.Flame shape and location were visualized through high-speed imaging of the OH* radical.Excess O 2 measurements were utilized to confirm the equivalence ratio for the three swirlers.Further details of the diagnostic tools and measuring equipment are presented in Section 3.2.The repeatability of the test points for each swirler was evaluated and reported within.

Atmospheric pressure generic swirl burner (APGSB)
The Atmospheric Pressure Generic Swirl Burner (APGSB) utilized in this study, depicted in Figure 2, is a direct geometrical replica of the High-Pressure Generic Swirl Burner (HPGSB) Mk.II, which has been extensively characterized previously (e.g., Pugh et al. 2018;Runyon 2017;Runyon et al. 2015Runyon et al. , 2017)), enabling comparison of results.Schematics of the crosssection and side view of the HPGSB Mk.II and APGSB are presented in Figure 2(a,b), respectively.Further details regarding the burner design, dimensions, and features can be found elsewhere (Runyon 2017).The flame was contained within a quartz tube flame holder of 100 mm ID with a length of approximately 407 mm.A top hat was constructed of a custom Nimonic 80a alloy structure and placed at the exit of the quartz confinement, serving as a support structure for the emissions probe of the gas analyzer.The air and fuel delivery to the APGSB was achieved through a dedicated supply system, allowing the delivery of air and fuel at controlled temperatures and flow rates.For the fuel supply, Bronkhorst miniCORI-FLOW M14V11l mass flow controllers were used, capable of supplying flows up to 8 g/s, with an uncertainty of ± 0.5% of full-scale.Regarding air delivery, an IN-FLOW F-203Al industrial-style thermal mass flow controller was utilized, capable of delivering up to 25 g/s, with an uncertainty of ± 0.5% of full-scale.For the specialized cases where more air flow rate was required, two mass flow controllers were used in parallel.The path of inlet air and fuel is presented in Figure 2(b).

High-speed OH* chemiluminescence system
The high-speed OH* chemiluminescence system used in the study is shown in Figure 3

Exhaust gas analysis
An industry standard exhaust gas emission system, provided by Signal Gas Analysers Ltd., was operated for sampling the exhaust gases.The sample handling system comprises a multi-point equal area probe connected to a water-cooled tube in tube heat exchanger used to condition the sample to 160°C, before subsequent transfer, using a heated diaphragm pump, through heated lines and filter units to a distribution oven that delivered hot wet sample gas to the NO x CLD and to a chiller prior to the dry O 2 paramagnetic sensor.
The temperature was maintained constant at 160°C throughout the sample handling system.NO x concentrations were quantified by a heated vacuum chemiluminescence analyzer (Signal Instruments 4000VM).O 2 concentrations were detected in the dry sample using a paramagnetic analyzer (Signal 9000MGA).All NO x concentrations captured were normalized to equivalent dry conditions (NO x, dry ), according to Eqn. ( 9) ISO-11,042 (British Standard ISO 11,042-1:1996), before being further normalized to equivalent 15% O 2 according to Eqn. ( 10) ISO-11,042 (British Standard ISO 11,042-1:1996).Concerning the uncertainty of the measurement, it was calculated at approximately 2 ppmv, accounting for analyzer specifications, linearization, accuracy in span gas specifications and the drift in the measurements.

Thermocouples
To appraise the combustion performance of the three swirlers of different surface roughness, while monitoring the system and ensuring safe and reliable operation, several precalibrated K-type thermocouples were instrumented around the APGSB, as shown in Figure 3(b).Other than TC 1, which was an exposed junction thermocouple monitoring the water-cooled emissions probe, the thermocouples were 310 stainless steel sheathed K-type .These instruments were suitable for continuous exposure up to +1100°C, with a maximum temperature rating of ≈ 1350°C.Regarding the measurement uncertainty in temperature, it was estimated as ± 2.2°C as per the manufacturers' specifications.The data was collected and logged in real-time at 10 Hz over 60 s.Three thermocouples were placed on the quartz glass confinement (TC3, TC4, and TC5) for surface temperature monitoring, while another (TC2) was placed in the exhaust section of the burner, aligned with its centerline, to capture the transient behavior of the exhaust gas stream temperature.Across all the test points investigated, the maximum standard deviation of the exhaust gas temperature was circa 1%.

Stability maps
The present study investigated the influence of surface roughness on LBO and FB limits for the two AM swirlers (AM-G & AM-R), in comparison to the traditionally manufactured "Machined" benchmark.Additionally, the study compared the operational stability of pure H 2 and CH 4 /H 2 flames against pure CH 4 flames to explore differences resulting from the thermo-diffusive properties of hydrogen.For the CH 4 case, FB could not be achieved at 25 kW for the current burner configuration, as the bulk flow velocity of the mixture remained higher than the flame-speed, even at near-stoichiometric conditions, where the flame-speed is estimated to reach its maximum.A "technical flashback point" (TFP) was, therefore, chosen as the upper limit of the stable operating curve (φ = 1.05).Figure 4 presents the burner stability envelopes of the three swirlers for 100% CH 4 flames as a function of Reynolds number and equivalence ratio.The horizontal error bars in the figure relate to the uncertainty of the mass flow rate, and hence equivalence ratio variations.It is observed that uncertainties remained sufficiently small, around 1% in the overall test points.
It is observed that within the relative uncertainty the three swirlers of different surface roughness demonstrated the same flame stability behavior, with almost identical TFP and LBO limits at φ = 1.05 ± 1% and φ = 0.558 ± 1%, respectively.This trend indicates a negligible effect of surface roughness on the CH 4 flames under the test conditions investigated, in agreement with previous investigations (Runyon et al. 2019).
Figures 5 and 6 demonstrate the burner stability map for the 100% H 2 and 23% vol CH 4 /77%v ol H 2 fuel blends, respectively.In contrast to the 100% CH 4 flames, flashback events were observed for the 100% H 2 and 23% vol CH 4 /77% vol H 2 flames.For the pure H 2 flames, the flashback points of the three swirlers overlapped at φ = 0.311 ± 1% regardless of the surface roughness (see Figure 5).For the 23% vol CH 4 /77% vol H 2 blend, the flashback propensity slightly shifted from φ = 0.458 ± 1% (Machined) to φ = 0.450 ± 1% (AM-R) with increasing surface roughness (see Figure 6), though it is noted this difference is within the stated uncertainty of the mass flow controller.The variation in the FB limit owing to surface roughness was expected to be small due to the small range of surface roughness (Runyon et al. 2019).The LBO limits for the pure H 2 case and the CH 4 /H 2 blend were detected at φ = 0.244 ± 1% and φ = 0.327 ± 1%, correspondingly.For both cases, surface roughness was found to have a minimal impact on LBO instability since for the three swirlers the LBO limits were identified under nominally similar equivalence ratios and were characterized by the same instability mechanism.This is likely due to the minor effect of the surface roughness height on modifying the bulk flow velocity against the flame speed.Regardless of the surface roughness, the H 2 content in the fuel mix has a significant effect on modifying the burner operability regime.Maintaining the same thermal power output, switching from 100% CH 4 through 23% vol CH 4 /77% vol H 2 to 100% H 2 modifies the amount of air flow to be introduced into the system to balance the flame speed.Therefore, the stable operating curve spread across different equivalence ratios and Reynolds numbers based on fuel compositions.As seen in Figures 4-6, the burner stability envelope was shifted toward leaner equivalence ratios with increased H 2 percentage in the fuel mix due to the higher reaction rate, burning velocity, and diffusivity associated with H 2 combustion.Consequently, the lean flammability limits were extended, and the LBO occurred at leaner conditions for the higher H 2 content of fuel mixes.This is consistent with several previous experimental studies (Kim, Arghode, and Gupta 2009;Liu et al. 2021;Schefer 2003;Tuncer, Acharya, and Uhm 2009;Kim et al. 2009).Specifically, the LBO limit was lowered by around 41% for 77% vol H 2 (φ = 0.327 ± 1%) enrichment in CH 4 and almost 60% for 100% H 2 (φ = 0.244 ± 1%), compared to baseline 100% CH 4 (φ = 0.558 ± 1%).The enhanced flammability limits by H 2 addition could be related to the thermo-diffusive properties of H 2 under lean turbulent conditions, in which the flame stretch accelerates the flame, resulting in higher burning velocities (Lapalme, Lemaire, and Seers 2017), consequently improving the resistance of the flame to combustion instabilities (Fairweather et al. 2009;Lapalme, Lemaire, and Seers 2017).
An increased H 2 content in the fuel mix significantly narrowed the operability range of the burners.For the 100% CH 4 flames, the stable operating curves of the "Machined" swirler, ranged from 0.558 < φ < 1.05 (Figure 4), whereas the stability limits of 23% vol CH 4 /77% vol H 2 and 100% H 2 , were reduced to a narrower equivalence range 0.327 < φ < 0.458 (Figure 5) and 0.244 < φ < 0.312 (Figure 6), respectively.This is qualitatively consistent with previous experimental studies utilizing a range of H 2 fuel blends (Runyon 2017;Syred et al. 2012).The narrow operability limits of H 2 -enriched flames impose significant technical challenges in burning hydrogen fuels in large-scale power plants, as even minor variations in operating conditions (inlet temperature, air/fuel mixture concentration, and pressure) could potentially result in blow-off or flashback phenomena, risking the power plant's operation and durability.This challenge was also issued in identifying the FB limits for the H 2 enriched flames, in which the flashback limits determined were within the quoted error bars.represent the standard deviations of the repeated data sets.The data was dispersed around 5-10% of the average mean values (≈2-4 ppmv) for φ > 0.66, complying with the uncertainty requirement of the BS-EN standard 14,792:2017.The NO x emissions showed substantial variations with the surface roughness toward the fuel-rich conditions, particularly at φ = 0.9 and φ = 1.1, in which the AM-R and AM-G exhibited an observable difference from the "Machined" swirler, though being within the standard measuring error of the experiments.

NO x emissions
Figures 8 and 9 show the variation of the normalized NO X emissions (dry, 15% O 2 ) across the equivalence ratio for the pure H 2 and 77% vol H 2 /23% vol CH 4 flames, respectively.However, the analysis did not give any solid evidence to draw a statistically significant conclusion regarding the surface roughness effect, as it is noted the error bars overlap with differences in measured NO X emissions of less than 2 ppmv.However, the lean premixed  H 2 flames under the considered test conditions showed significantly lower NO x emissions across their flammability limits when compared with the pure CH 4 flames.This is due to the suppression of the thermal NO x contribution to the total NO x formation due to the lower relative reactivity of hydrogen flames at very lean conditions.These observations are consistent with previous experimental studies investigating the effect of H 2 enrichment in CH 4 with respect to NO x emissions (Griebel, Boschek, and Jansohn 2007;Tuncer, Acharya, and Uhm 2009).

Exhaust temperatures
Figures 10-12 present the gas temperature variations across equivalence ratios for the three different fuel blends and swirlers.In contrast to the NO x emissions, the three swirlers of the different surface roughness (1-9 μm) yielded statistically validated differences in the exhaust temperatures at the center of the exit plane.The AM-R swirler was found to systematically produce higher exhaust temperatures, followed by the AM-G and "Machined" swirler, which showed similar readings.The difference in exhaust temperature between the AM-R and the "Machined" swirlers was approximately ≈ 40°C for the pure CH 4 case, which reduced to ≈ 29°C and ≈ 25°C for the blend and pure H 2 cases, respectively.The analogous difference between the "Machined" and the AM-G swirlers was substantially lower and equal to ≈ 5°C for most equivalence ratios across the range of fuels.Compared to the 100% CH 4 case, the resultant exhaust temperatures were also significantly lower for the two alternative fuel cases.This is consistent with the lower NO x emissions recorded under pure H 2 and CH 4 /H 2 combustion.Since all profiles were measured using the same pre-calibrated K-type thermocouple mounted at the same location, the differences in exhaust temperature should not stem from inconsistencies of the measurement setup.Moreover, since the inlet plenum temperature was within the acceptable level of deviation of 150 ± 5°C, these differences are not likely owing to the physical properties of the mixture.This was further confirmed by the NO x emissions, which were within similar levels, which confirmed approximately equal flame temperatures.Therefore, it is proposed that the observed temperature differences in the exhaust stem from variations in the flow field aerodynamics and subsequently, the stabilization location of the flame resultant from changes in surface roughness.To investigate the validity of this hypothesis, further experimental investigations were undertaken utilizing high-speed OH* chemiluminescence measurements of the flame, to confirm the flame location under stable operation.

Flame locations
Figure 13 shows examples of Abel deconvoluted OH* chemiluminescence images for 100% CH 4 (φ = 0.80), 77% vol H 2 /23% vol CH 4 (φ = 0.40), and 100% H 2 (φ = 0.285) flames, including the weighted centroid of the flame area as represented by a black "dot" on each image.The weighted centroid was constructed on the Abel deconvoluted images through the standard MATLAB operators.Regardless of the fuel type, the weighted centroid of the flame area slightly moved toward the centerline of the burner (r = 0) with increasing surface roughness.Apart from the visual observations, the shifting was quantified in the cartesian plane, across all equivalence ratios, shown in Figures 14-16, for 100% CH 4 , 77% vol H 2 /23% vol CH 4 and 100% H 2 , respectively.For all the fuel types tested, noticeable changes in the x-coordinate of the flame centroid were noted for the swirlers of different surface roughness, with the stable flame location appearing closer to the centerline of the burner for relatively rougher surfaces, which is in qualitative agreement with other studies (Runyon et al. 2019).For the 100% CH 4 and the 77% vol H 2 /23% vol H 2 cases, this difference of the x-coordinate of the flame centroid between  the "Machined" and the AM-R swirlers was equal to ≈4 mm, whilst for the 100% H 2 case, the analogous change was ≈6 mm.The most considerable difference between the "Machined" and the AM-R swirler was found for each fuel tested at its corresponding LBO limit.This marginal difference can be explained by the fact that the Reynolds number was at its maximum, and so the surface roughness for this case was expected to have a more significant impact due to the increased surface roughness/boundary layer thickness ratio (Kadivar, Tormey, and McGranaghan 2021).As discussed, for the increased surface roughness cases, the flame stabilized closer to the centerline of the burner corresponding with the aforementioned higher temperatures recorded by the exhaust thermocouple, which was located on the centerline.It is noted that the increase in temperature corresponded only to a single point and should not be confused with a spatially averaged temperature value across the diameter of the exhaust, which would explain the resultant similar NO x emissions measured.
Regardless of the surface roughness and the fuel type, along the stability curve (LBO to FB), the y-coordinate of the flame centroids decreases.This decrease was ≈25 mm for the pure CH 4 and the CH 4 /H 2 blend cases, and ≈15 mm for the pure H 2 case where the flame had already more compact shape due to the thermodynamic properties of hydrogen and equivalence ratio variation.This behavior can be explained by the modification of bulk flow velocity and, thus, flame location with the equivalence ratio.When airflow decreases at a constant fuel flow, the equivalence ratio increases toward the stoichiometry, which in turn decreases bulk flow velocity and increases the flame speed (Law 2006).This change in the flame speed with respect to bulk flow velocity causes the flame location to shift.As a result, the flame retreats closer to the nozzle exit.

Conclusions
Additive Manufacturing not only enables the control of the resultant surface roughness during the fabrication stage of the component but also potentially minimizes the need for post-processing, leading into potential cost and lead-time reductions.In this context, empirical and image processing methods were developed and applied to evaluate the influence of surface roughness on burner characteristics and combustion performance for CH 4 , CH 4 /H 2 blended and H 2 fuels.Image processing enabled the flame centroids to be located, which were found to shift closer to the centerline of the burner as the swirlers' surface roughness increases.Variation of surface roughness also altered the aerodynamic flow field in a such way that the flame was stabilized and aligned closer to the burner axis, where highest temperature readings were recorded for the AM-R swirler.Despite the altered flame location, the surface roughness heights selected did not significantly influence the burner stability envelopes and NO x emissions performance of the swirl burner, the latter considered due to temperature spatial-averaging across the exhaust.For all fuel types tested, the swirlers manufactured using additive layer (AM-R and AM-G) resulted in combustion performance comparable to the traditionally manufactured Machined swirler, indicating from a practical perspective for manufacturers of additive layer parts, that there are negligible performance advantages in post-manufacture surface finishing, which could be an important consideration concerning production cost reductions.
(a).The system comprises a high-speed camera relay lens and image intensifier, UV lens (Ricoh FL-GC7838-VGUV, f/16), and 310 nm narrow bandpass filter.The high-speed camera is a monochromatic Vision Research Phantom v1212 (12-bit 12,000 frames/second at full 1280 × 800 resolution) controlled using Vision Research PCC 2.8 and Specialised Imaging Limited SILControl2 software.For each test point, 2000 images were acquired at the frame rate of 4000 Hz, corresponding to a recording time of 0.5 s, with an intensifier gate time of 10 μs utilized at constant gain.The resultant field of view was approximately 157 mm by 162 mm in the radial (x) and axial (y) direction at the camera resolution of 544 × 648 pixels, providing a resolution of approximately 3.46 pixels/mm.Several imaging techniques were subsequently used to post-process the recorded images before the reconstruction of Abeldeconvoluted OH* chemiluminescence images, involving noise filtering through a 3 × 3 pixel median filter, background intensity subtraction, and temporal averaging.An opensource MATLAB code developed by Killer(Killer 2016) based on the Abel transformation method reported by Pretzel(Pretzel 1991) was applied to identify the flame structure.

Figure 3 .
Figure 3. (a) Experimental set-up view with the high-speed camera (Yellow rectangular section) and intensifier (red rectangular section) aligned in front of the APGSB burner.(b) Combustion and exhaust section of APGSB, indicating the metallic sampling probe support structure (yellow dotted rectangular section), the water-cooled probe (green dotted oval section) and the various K-type thermocouples (TCs).

Figure 7 Figure 7 .
Figure 7 presents the results of the investigation on the normalized NO x emissions (dry, 15% O 2 ) for the three swirl burners for CH 4 /air mixtures.The error bars in the figure

Figure 8 .
Figure 8.Average NO x emissions indicating repeatability for H 2 /air mixtures.
Stable operating trends for 100% H 2 with machined, am-G and AM-R swirlers.Stable operating trends for 23% vol CH 4 /77% vol H 2 with machined, am-G and AM-R swirlers.
Average NO x emissions indicating repeatability for CH 4 /H 2 fuel mixtures.