Particle deposition patterns on high-pressure turbine vanes with aggressive inlet swirl

Xing YANG, Zihan HAO, Zhenping FENG

Shaanxi Engineering Laboratory of Turbomachinery and Power Equipment, Institute of Turbomachinery, School of Energy and Power Engineering, Xi’an Jiaotong University, Xi’an 710049, China

KEYWORDS Combustor-turbine interaction;Gas turbine;Inlet swirl;Numerical simulation;Particle deposition

Abstract Characteristics of particle migration and deposition were numerically investigated in presence of aggressive swirl at the turbine inlet.The isolated effects of the inlet swirl were considered in detail by shifting the circumferential position of the swirl and by implementing positive and negative swirling directions. Particles were released from the turbine inlet and the resulting deposition on the vanes was determined by using the critical velocity model in a range of particle diameters from 1 to 25 μm.Results show that the particles are more likely to move outwards to the boundary walls of the passage by the action of the swirling flow.However,this could be relieved by increasing the particle size. An imbalance problem of the deposition is found between the adjacent vanes,which could introduce additional inlet non-uniformities towards the downstream rotor and thus accelerate performance degradation of the turbine stage.Overall,the negative swirl case has higher overall capture efficiency within the entire turbine than the positive swirl case for larger particles,and when the inlet swirl is shifted to the mid-passage of the turbine, more deposits could be produced in comparison with the case in which the swirl aims at the vane leading edge.

1. Introduction

Deposition of particles including sand, volcanic ash, and sea salt has been an ever-increasingly critical issue to the operating safety of gas turbines because aircraft engines are exposed to environment with particle concentrations and also because alternative synthetic gas (gasified coal, for instance) is used to fuel the land-based gas turbine.Small-size particles are ingested into the gas turbine, accelerated and heated by the hot gases, and then deposited on the turbine surfaces. It has been verified that particle deposits and its build-up are detrimental to the reliability and durability of the turbine hardware by increasing heat transfer levels and surface friction losses due to increased roughness,and even reducing film cooling effectiveness due to blockages of film holes.Among the turbine hot-components, the High-Pressure (HP) turbine directly downstream of the combustor is a region where the gas temperature is the highest and a considerable number of film holes are present. Therefore, the aero-thermal performance of the Nozzle Guide Vane (NGV) is more sensitive to deposition.

Nomenclature Pt,intotal pressure at combustor inlet Tt,inair temperature at turbine inlet Tuturbulence intensity Λturbulence integral length Ps,exstatic pressure at turbine outlet P-t;inarea-averaged total pressure at combustor inlet Caxaxial chord length Sspan height at combustor exit ρpparticle density μviscosity of air Uinmagnitude of velocity at turbine inlet ggravity vpparticle Poisson ratio vsPoisson ratio of vane surface EpYoung’s modulus of particle EsYoung’s modulus of vane surface

With an attempt to understand the deposition effects and to alleviate the problem in gas turbine,many fundamental studies have been conducted to deal with this topic. Tabakoff and Husseincould be the pioneers to work on this topic.Although they mainly focused on erosion patterns of the larger-size particles onto the blade surfaces, the results of the particle trajectories within the turbine were valuable. Particles with larger diameters had higher Stokes numbers, resulting in their deviation from the flow path of the gas, and these particles were more likely to move outwards to the blade tip due to higher centrifugal force.The findings were confirmed by a later work of Barker et al.Wenglarz and Foxexplored the effects of gas temperature and found that increasing the gas temperature significantly increased the deposition rates. In addition,the work by Bonsand Crosbyet al. showed that the metal temperature was also an important factor affecting the deposition behaviors and a reduced formation of deposits was observed in cooler areas. Recently, rotating effects were numerically documented by Zagnoliand Bowenet al. in rotor passages. In comparison with the deposition patterns in stator passages, additional physical mechanisms were discovered due to the rotating conditions. Besides, because the upstream stator vane captured most of the larger particles,the particles entering the rotor were much smaller.

Deposition has effects on film cooling performance,and in turn, coolant injection can also influence the extent of deposition. On a film-cooled flat plate, coolant injection was found to inhibit the deposition and higher blowing ratios reduced capture efficiency of the particles.In enginerepresentative components, further studies by Bonilla,Borello,and Prenteret al. showed that deposition rates and distribution patterns were dramatically changed by film cooling and higher cooling effectiveness decreased the deposition levels. Additionally, slot film cooling was found to be more effective in reducing particle deposition relative to discrete coolant injection.

With respect to a HP turbine, interactions between an upstream combustor and the turbine generally dominate flow and thermal fields in the subsequential turbine passage, and thus influence turbine deposition effects.Common interactions from the combustor are the flow and temperature distortions entering the turbine. The effects of temperature distortions,i.e.hot streaks,on the turbine aero-thermal performance have been extensively conducted in past literature,such as Refs.However,there are very few studies regarding the deposition in the inclusion of hot streaks. Casadayand Prenteret al.imposed hot streaks at the inlets of a single turbine vane and a turbine stage,respectively.Since the deposit rates were found to be linked to the gas temperatures, the presence of the hot streaks further increased the deposition on the turbine surfaces and most importantly, altered the distribution patterns of the deposits as well. Later, Prenter et al.continued their work by integrating the hot streaks with film cooling and concluded that despite the hot streaks, film injection dominated the deposit rates and patterns because the particles were cooled by wall-nearby coolant prior to impacting the turbine surfaces.

As highlighted from above discussion, although the effects of hot streaks (temperature distortion) have been documented experimentally and numerically, there is lack of work on the impacts of inlet swirl (flow distortions) on turbine particle deposition in the open literature.Martine et al.imposed flow conditions exiting a combustor onto a turbine inlet where particles were fed and Jiang et al.performed a combustorturbine integrated model to investigate ash concentration in the hot-components.However,in their work,on the one hand,the inlet swirl and the hot streak were coupled with one other;on the other hand, they both examined the migration or impact patterns of the particles within the turbine only, without considering deposition effects on the turbine surfaces. In this paper,the isolated effects of turbine inlet swirl on particle deposition on NGV surfaces are numerically investigated in a wide range of particle dimeters. This is the first work, to the authors’ knowledge, on the detailed migration behaviors and deposition patterns of particles in a HP turbine vane passage by considering the oncoming swirl flow in isolation. In particular, the circumferential position effects of the inlet swirl relative to the NGV and the swirling direction are considered as well.It is vital to obtain such additional knowledge of the particle migration and impacting behaviors and the resulting deposition patterns in HP turbine passages with inlet swirl,which further provides guidelines to designers in turbine design and end-users in operation and maintenance management.

2. Methodology

Deposition of contaminant particles was numerically investigated in an NGV passage of an advanced aircraft engine using the FLUENT code and a particle-wall interaction model that was applied into the code as a User-Defined Function (UDF). The simulation strategy adopted in this study is different from that reported by Martine et al.,who conducted a separate inlet boundary condition simulation and extracted out the thermal and flow conditions at the combustor exit and then imposed onto the inlet of an isolated stator vane. In this case, the coupling effects between the combustor and the turbine were not considered when particles were seeded into the air flows. For the current work, in order to consider the combustor-turbine coupling effects in the presence of particles and the clocking positions between the inlet swirl and the turbine, the simulations were conducted in a combustor-turbine integrated model by imposing a periodic repeat boundary condition at the component interface, allowing the coupling and the circumferential position effects to be resolved. In addition, with an attempt to analyze the isolated effects of the inlet swirl, the particles were not tracked through the combustor but were released from the interface into the turbine passage.

2.1. Generation of inlet swirl

It has been revealed that the downstream turbine could also influence the flow structures within the upstream combustor. Therefore, an integrated model of a combustor and a HP turbine vane is chosen to generate engine-like inlet swirl.Fig. 1 shows the entire domain of the integrated model which is a periodic domain of one annular combustor sector to two turbine vane passages. The combustor is a typical modern‘‘lean-burn” combustor design that uses aggressive swirler to strengthen mixing in the combustion. To avoid the hot streak factor, effusion holes for liner cooling are removed from the liner walls but the other geometrical features, including the swirler and the convergent hot gas path,are retained to closely mimic flow patterns due to the combustor-turbine interactions.An exemplary result of flow structures at the interface when the swirler is aligned to the mid-passage of the turbine is shown in Fig.2.A typical velocity profile is well defined and the swirl angle and pitch angle fields clearly indicate a strong inlet swirling flow exiting from the combustor, representing a real engine condition.

A more particular focus in this study is the circumferential position effects between the inlet swirl and the turbine.Two typical clocking positions of the swirler are considered,as shown in Fig.3.The case when the swirler is aligned to the mid-passage between Vane 1 and Vane 2 is referred to as a‘‘MP”case,and another configuration where the swirler is directly located at the upstream of the leading edge of Vane 2 is referred to as a‘‘LE” case. Furthermore, the effects of positive and negative swirl are investigated as well.The positive swirl is when the swirling flow has the same rotating direction as the downstream rotor,i.e.clockwise inlet swirl from the upstream view,and vice versa for the negative swirl. Therefore, totally four inlet swirl conditions are generated for comparative studies: MP-Pos,MP-Rev,LE-Pos,and LE-Rev cases(see Fig.3).

2.2. Turbine geometries and boundary conditions

The NGV was a HP turbine stator vane from an aircraft engine. The stator consisted of 34 vanes and had a contoured flow path along the axial direction. The averaged span height of the vane was 53.25 mm and the true chord length at the mid-span was approximately 56.33 mm. Fig. 1 shows the whole computational domain consisting of the combustor and the turbine. The domain inlet was in front of the swirler with a distance of twice the dimeter of the swirler apart from the swirler inlet and the outlet was at the NGV exit, which was two chord length from the vane trailing edge. The combustor-turbine interface was positioned at 1.37Cupstream of the passage inlet.For the purpose of investigating the effects of inlet swirl only, the nonuniformity of inlet temperature was not included in this study,but instead,a uniform temperature with a mass-averaged value of T=2140 K and a total pressure of 4058.54 kPa were specified at the domain inlet and an averaged static pressure of 1392.51 kPa at the domain outlet. The inlet turbulence intensity and integral length were 9.8% and 10 mm, respectively. In addition to the vane surfaces having a constant temperature of 1500 K,which was approximately 0.7Tas reported in Ref.,the other surfaces were set as non-slip, adiabatic wall conditions.In fact, a considerable number of film holes were present on the NGV surfaces to provide film coverage for the vane metal,but the film holes were removed for future studies of the effects of coolant injection, where the results obtained in this work would be used for valuable comparisons.

Fig. 1 Periodic domain of a 1/17 sector of combustor and nozzle guide vane.

Fig. 2 Distributions of inlet swirl and pitch angles at turbine inlet for positive swirl at mid-passage.

Fig.3 Schematics of swirling direction and clocking positions of swirler relative to turbine vane.

2.3. Mesh generation and turbulence model validation

Meshes for the combustor and the turbine vane passage domain were fully unstructured with refined prism mesh within the boundary layer regions near wall surfaces, guaranteeing the non-dimensional distance, y, lower than 1.0. Besides,areas with large flow gradients,such as the downstream region of the swirler and the leading and trailing edges of the vane,were also refined. To determine the final grid size used for the simulations, a detailed grid-independent analysis was performed by generating coarse, medium, fine, and refined meshes. The total number of the grid sizes for the four sets of meshes are 10.24 million, 21.89 million, 43.10 million, and 65.31 million cell elements,respectively.To guarantee the combustor and the turbine domains both achieve the gridindependent solutions, axial velocity (u) profiles along the spanwise (radial) direction at the mid-passage of the combustor exit (Fig. 4(a)) and static pressure coefficients (C) along the surfaces of Vane 2 at the mid-span (Fig. 4(b)) were compared among the four meshes, as shown in Fig. 4 that takes the MP-Pos case as an example. It is clear that no oscillations are found when increasing the number of the cell elements.When the grid size has 43.10 million cell elements, further increasing the grid nodes has slight effects on the calculated results both within the combustor and turbine domains,showing that the fine mesh achieved a good compromise between numerical accuracy and computational sources. Therefore,the fine mesh, which totally had 43.10 million cell elements(25.20 million cell elements in the combustor domain and 17.90 million elements in the turbine domain),was determined for all simulations in this study.

Fig. 4 Velocity profiles and static pressure coefficients calculated by four sets of meshes.

Fig. 5 Comparison of predicted results and measurements along mid-span of vane surface.

The accurate predictions of the particle migration and deposition within the turbine vane passage rely on the accurate solution of the air flow and the particle–wall interaction model.The particle–wall interaction model will be presented in a later section. The flow fields were obtained by solving the steady Reynolds-averaged Navier-Stokes equations of continuity,momentum, and energy using the FLUENT code. Prior to all simulations, a separate turbulence model validation was conducted to evaluate the solution of the flow via comparison to available experimental data. A category of k-ω turbulence models, which has been extensively used in turbomachinery simulations,was chosen as a closure for the equations and the validation simulation was performed on a rotor blade that was originally downstream of the stator vane investigated in this study. Fig. 5(a) plots wall temperatures (T) along the blade surfaces at the midspan from the k-ω turbulence models against in-house experimental data.The data were obtained by running experiments at high temperature conditions, in which 24 thermocouples were embedded into the blade to measure the wall temperatures. The blade wall temperatures calculated by the three k-ω type turbulence models share a very similar tendency but the levels vary with each other. Both the Shear Stress Transport (SST) k-ω turbulence models with and without the γ-θ transition model under-predict the wall temperatures. As to the standard k-ω model, except for a limited region on the pressure surface near the trailing edge,the calculated wall temperatures agree well with the experimental data.Furthermore,measured pressure coefficients,C,are also used to validate the reliability of the standard k-ω model in predicting flow fields in the turbine passage, showing a good agreement with one another (see Fig. 5(b)). Therefore, it is convincing that the standard k-ω turbulence model can generate reliable flow fields in which the particles will be tracked.

2.4. Particle injection

In experimental and numerical modeling, an accelerated deposition method by seeding particles with higher fraction is generally used to generate deposition that is found in service for thousands of hours in several hours.In this study,the particle concentration at the inlet was set as 9.285 × 10kg/s(which was 300 parts per million by weight of the mainstream mass flow rate) but the volume fraction was approximately 3.95%, far lower than 10%, which met with the requirements of the usage of the Discrete Phase Model (DPM) in the FLUENT. Since the particle concentration was relatively higher than that reported in previous literature, the effects of particles on the flow were taken into consideration by using the ‘‘interaction with continuous phase” model (a two-way method) in the code. In the simulations of this study, the first step was to solve the flow in the absence of particles, and the second step was to release particles from the combustorturbine interface into the turbine vane passages in which the flow was previously solved in the first step. The particles were injected at each grid nodes of the interface, resulting in more than 37000 particle sources at the turbine inlet. In the second step, after the particle–fluid interaction model was activated,the simulations continued until the flow convergence was achieved. Once the particles were seeded at the turbine inlet,their temperatures and velocities were initialized using a UDF to be equal to the air flow at the inlet. Moreover, the stochastic tracking method was used to consider the effects of turbulent dispersion on the particles and ten particles were thereby released at each particle source at one time.

The density of the particles is 2320 kg/mand the specific heat capacity was 984 J/(kg∙K). Five various diameters, d,of the particles are considered in this study: 1, 3, 5, 12, and 25 μm. According to the definition of Stokes number,

the resulting Stokes numbers are from 0.00541 to 3.38, which are in the typical range of a real engine. In the case of the Stokes number lower than unity, a particle is likely to follow the flow path of the mainstream flow and for a case where the Stokes number is higher than unity, a particle tends to remain its initial trajectory.

2.5. Particle tacking

The continuous air flow is solved in the Eulerian frame while the DPM calculates the trajectories of the discrete particles in the Lagrangian frame according to the balance between forces acting on the particle and its inertia.

where a,a,and aare constant.A particle in the engine combustor is generally heated to a molten state and then is cooled,which could make the shape of the particle to be spherical due to the acceleration of the air flow. Therefore, the assumption of the spheres was reasonable and adopted in most of past studies.

2.6. Particle-wall interactions

From Eqs.(6)to(9),the critical velocity,V,is determined by both properties of the particle and flow.A particle that has a normal impact velocity,V,higher than the Vwill rebound and continue its trajectory,and that with a Vvalue lower than the Vwill deposit. However, the critical velocity model of Brach and Dunnonly considers a pure sticking interaction between the particle and the wall but does not include the detachment process, in which the shear stress of the air flow applied to the deposited particles should be taken into consideration.After a particle is stuck onto the vane surface,the van der Waals force is the main sticking force.According to the balance between the moment applied to the particle by the shear stress of the fluid and the particle–wall van der Waals force, a critical wall shear velocity, u, can be calculated aswhere mis the particle mass; cis the specific heat capacity;Ais the external surface area of the particle; h is the convection heat transfer coefficient around the particle; Tis the local temperature of the air flow.

3. Results and discussion

3.1. Flow and thermal fields

The flow and thermal fields within the passage with positive and negative inlet swirl are shown in Fig. 6. The threedimensional streamlines issued from a circumferential curve at the midspan of the vane inlet are used to illustrate the skew of the flows due to the inlet swirl. In the MP-Pos case,although the swirler geometrically lines up with the midpassage of the turbine vane, the inlet swirl is slightly deflected towards Vane 1(refer to Fig.1)when approaching the turbine because of the circumferential pressure gradient.This results in that the flow around Vane 1 is significantly skewed while that around Vane 2 is much smooth throughout the passage.When the swirler is circumferentially shifted towards the Leading Edge (LE) of Vane 2, the perturbation is visible across the whole passage between Vane 1 and Vane 2. Moreover, as the deflection of the inlet swirl towards Vane 1, the flow around Vane 1 is still disturbed despite the alignment of the swirler with the LE of Vane 2. Compared with the two positive swirl cases (MP-Pos and LE-Pos), the application of the negative inlet swirl generates a much more distorted flow pattern within the passage. It is found that the inlet swirl is more likely to impinge onto Vane 2 regardless of the clocking position of the swirler. In addition, the four cases share a similar pattern that the inlet swirl tends to move outwards to the shroud due to the contraction of the flow path along the axial direction.

As mentioned previously, the Young’s modulus of the particle is linked to the thermal states of the vane surfaces. Fig. 6 also plots the distributions of heat flux into the vanes ( ˙Q).Since the vane LE directly faces the oncoming flow,high thermal loads are generated in that region, but it is because of the distortion of the flow, the high thermal loads are found to be discontinuous along the span at the LE,which is different from the pattern for a uniform oncoming flow case. On the vane Pressure Surfaces (PS), the footprints of the disturbed flow can be also seen from the non-uniform distributions of the thermal loads. It appears that higher heat rates will be transferred into the vane where the inlet swirl impacts. Moreover,the high thermal loads on the vanes indicate high heat transfer levels there, and hence, the particles impacting onto the high thermal loads regions could be heated to higher temperatures,which would increase the particle deposit rates.

As the particle is carried by the air flow,the flow patterns of the air are the most important factor to determine the migration and impact characteristics of the particles within the passage.The flow fields at the mid-span due to the aggressive inlet swirl are further displayed in Fig. 7, where the dash line with an arrow indicates the deflection of the swirl and the dash line without an arrow indicates the hypothetical path of the swirl’migration in inviscid flows at the mid-span.In Fig.6,the threedimensional streamlines present the origin of the inlet swirl at the vane inlet and the two-dimensional streamlines and Mach number on the mid-span plane in Fig.7 can be used to indicate the migration of the inlet swirl and the variation of the velocity across the passage. In Fig. 7, it is much clear to observe the position of the inlet swirl relative to the vane when it proceeds into the passage. In both positive swirl cases (Fig. 7(a) and 7(b)), the inlet swirl progresses following the circumferential pressure gradient and impacts the vane at the fore part of the PS of Vane 1 and Vane 2, respectively. However, for the negative swirl cases (Fig. 7(c) and 7(d)), the inlet swirl is capable to circumferentially migrate against the pressure gradient.As the inlet swirl is directed towards the pressure side of Vane 1 in the case that the positive inlet swirl aims at the midpassage of the turbine vane, the flow in the passage between Vane 1 and Vane 2, which is aligned to the swirler, is slightly affected.Globally,the flow in the positive swirl cases is less disturbed compared with that in the negative swirl cases. Additionally, the MP-Pos case has relatively higher Mach numbers within and at the exit of the passage than the other cases, because the size of the swirling flow in the MP-Pos case is smaller and thus has less blockages to the flow passage.

Fig. 6 Streamlines issued from mid-span at turbine inlet and thermal loads on vane surfaces.

Fig. 7 Distributions of Mach number at mid-span, superpositioned with streamlines.

3.2. Migration of particles

Fig. 8 Particle concentration on three cut planes at X/Cax = - 0.13, 0.34, and 0.81 (dp = 12 μm).

A basic understanding of the flow inside the turbine passage with aggressive inlet swirl has been obtained from the previous analysis, and it is helpful to analyze the migration patterns of the particles within the air flow in this section.Fig.8 shows the particle concentration on three cut planes along the axial direction. The particles are sampled at X/C= - 0.13, 0.34, and 0.81, respectively, and the diameter of the particles is 12 μm.As the flow has a great effect on the particle migration, the footprints of the inlet swirl that progresses within the passage can be visible from the particle concentration patterns on the three planes, where the absence of the particles indicates the possible disturbance of the swirling flow. At the passage inlet of X/C= -0.13,the positive inlet swirl sweeps the particles out of its way,resulting in an absence of the particles in a large portion of that plane but a concentration of the particles in the front of the LE of Vane 2 (MP-Pos case) or Vane 1 (LE-Pos case).A similar pattern of the particle concentration is also visible for the negative inlet swirl case at the passage inlet. However,as the negative swirl tends to move towards Vane 2,more particles are found in the region upstream of the LE of Vane 1 regardless of the circumferential location of the swirler. Inside the passage, in spite of the circumferential pressure gradient,the particles are more likely to gather along the PS, particularly in the passage from the PS of Vane 1 to the Suction Surface (SS) of Vane 2 where the flow is less distorted except for the MP-Pos case(see Fig.7).This is because the particles with a dimeter of 12 μm have a high Stokes number value of 0.779,which makes the particles to be more likely to follow their own initial flow paths.

The effects of particle diameter on their trajectories within the air flow are displayed in Fig. 9, where the positive inlet swirl aims at the mid-passage and the diameters of the particle are 1,5,and 25 μm.To clearly show the trajectories of the particles in the air flow, the particles injected from one source at the cross-point of the mid-passage and mid-span are tracked.Fig. 9(a) presents the trajectories of the particles with a diameter of 1 μm.It is clear that the particles follow the streamlines of the air flow and have a swirling flow pattern similar to the air flow structures in Fig.6(a).The trajectories of the particles with a larger dimeter of 5 μm are shown in Fig. 9(b). As the inertia of the larger particles is higher, the swirling motion of the particles generates the trajectories with larger radii across the passage. A further increase of the particle diameter up to 25 μm makes the air flow to impose little action on the particle motion (see Fig. 9(c)). The particles migrate along their initial injection trajectories but are directed by the circumferential pressure gradient. It can be thereby concluded from Figs. 8 and 9 that the particle trajectories rely on not only the air flow fields but also the particle size.

3.3. Positive swirl effects

Fig. 9 Trajectories of particles issued from a source at mid-span and mid-passage for MP-Pos case.

Fig. 10 Particle deposits on turbine vane for MP-Pos case.

Fig. 11 Particle deposits on turbine vane for LE-Pos case.

Particle deposit patterns on the turbine surfaces for the MPPos and LE-Pos cases are shown in Figs. 10 and 11. Also included in Figs.10 and 11 are the effects of the particle diameter since the above discussion has shown that the particles with various diameters have dramatical differences in their trajectories. In the MP-Pos case, the particles with a diameter of 1 μm follow the air flow and impact on both PS and SS of the vanes.As the positive inlet swirl mainly distorts the flow in the passage from the PS of Vane 1 to the SS of Vane 2(see Fig. 7(a)),the deposits on the PS of Vane 1 and the SS of Vane 2 are lower than those on the other sides of Vane 1 and Vane 2(Fig. 10(a)). Another difference of the deposition patterns between the two adjacent vanes is in the leading-edge region.At the LE of Vane 2,the distribution of the deposits is continuous along the spanwise direction. By increasing the particle diameter, the deposits are gradually invisible on the vane SS because the larger particles are more likely to follow its original injection trajectories and thus tend to directly impact the PS. This causes the deposit rates on the PS are significantly increased.As the inlet swirl could sweep the particles outwards to the boundary walls of the domain, it is more possible to observe high particle concentration on the vane surfaces near the hub and shroud walls because the particles that rebound from the hub and shroud walls could impact the vane surfaces again. This pattern is also applied to the negative swirl cases that will be presented in a later section.

The variation of the deposition patterns on the vanes with the shift of the inlet swirl are found by comparing the MP-Pos and the LE-Pos cases (Fig. 10 vs. Fig. 11). At d= 1 μm for the LE-Pos case, less particles are seen to deposit on the vane SS and the deposits on the SS of Vane 2 become higher than those on the SS of Vane 1, which is contrary to the MP-Pos case. When the particle diameter rises, in addition to the changes of the deposit distribution patterns, it appears that the deposit rates on the surfaces of the two vanes are decreased by moving the inlet swirl to the LE of Vane 2. Moreover, in comparison with the MP-Pos case, the LE-Pos case has more particle deposits at the LE of Vane 1 but less deposits on the LE of Vane 2.

To quantitatively compare the deposit rates on the two vanes for the positive inlet swirl cases with different particle sizes, the deposit rates are spanwise averaged and their distributions as a function of the axial direction are plotted in Fig.12.Overall,the LE region and the aft part of the PS have the highest possibilities to be the regions where the deposits are high,regardless of the position of the inlet swirl and the particle size. As to the smaller particles with dimeters of 1 μm and 5 μm, the deposits on the SS of Vane 1 are generally higher than those on the SS of Vane 2 and the opposite patterns are seen on the PS of the two vanes for the MP-Pos case.However, for the LE-Pos case the deposit rates on both PS and SS of Vane 2 are higher than those on Vane 1.It has been shown in Fig. 9(c) that the particles with a diameter of 25 μm are slightly affected by the distorted flow but are directed towards Vane 1 by the circumferential pressure gradient. The deposits on Vane 1 are thereby found to be much higher than those on Vane 2 (Fig. 12(c)). Moreover, the shift of the inlet swirl towards the LE of Vane 2 has more prominent effects on the deposits on Vane 1.

3.4. Negative swirl effects

The effects of the negative inlet swirl at two circumferential positions on the particle deposits are shown in Figs. 13 and 14 for various particle diameters. Similar to the positive inlet swirl cases,the deposits on the SS of the vanes are only visible for smaller particle sizes and, most of the particles are deposited on the PS of the vanes because the PS is windward to the impacting particles. A general pattern from Figs. 13 and 14 is that the distribution patterns of the deposits on the walls of the vanes could be reversed due to the change of the swirling direction.In other words,the region having high deposit rates in the positive swirl cases in Figs. 10 and 11 could become a region with low deposit rates in the negative swirl cases.Besides the reversed deposit locations on the vane walls,changing the swirling direction also alters the levels of the deposit rates, which is linked to the changes of the air flow and the heat transfer levels near the vane walls. Compared with the LE-Pos case in Fig. 11, the deposits on Vane 2 in the LE-Rev case are increased for large particle sizes.Although increasing the particle size yields to different deposition patterns on the vanes,the deposit rates are continuously increased for all cases.

Fig. 12 Spanwise-averaged deposit rates along axial direction for MP-Pos (top) and LE-Pos (down) cases.

Fig. 13 Particle deposits on turbine vane for MP-Rev case.

To demonstrate the possibility of the particle deposits on each surface of the vane, the percentage of the deposits on the PS and SS for each vane are summarized in Table 1,showing the tendency of the likelihood of the particle deposits on the vane surfaces. It is much clear that the PS on both Vane 1 and Vane 2 is always the high deposit rate regions and the percentage of the deposits on the PS is rapidly increased with the increase of the particle size. Overall, in the negative swirl cases, the percentage of the deposits on the SS of Vane 2 is higher than that on the SS of Vane 1. In fact, the deposition patterns in Figs. 10-14 have shown that most areas of the SS have no deposits. The deposits on the SS are attributed to those in the LE region of the SS. Therefore, the percentage of the deposits on the SS is strongly linked to the deposit rates at the vane LE that is dependent on the coupling effects of the swirling flows and the particle size. In Table 1, the lowest percentage of the deposits on the SS occurs on Vane 1 in the MPRev case with a particle diameter of 5 μm and the highest percentage of the deposits on the SS is found on Vane 2 in the LERev case with a particle diameter of 1 μm.

Fig. 14 Particle deposits on turbine vane for LE-Rev case.

Table 1 Percentage of deposits on pressure and suction surfaces of Vane 1 and Vane 2 for negative swirl cases.

3.5. Capture efficiency

In this section,capture efficiency,which is the mass ratio of the deposits to the total injected particles, on Vane 1 and Vane 2 are discussed. Fig. 15 shows the variations of the capture efficiency(η)with Stokes numbers and the pie charts in the plots indicate the percentage of the deposits on each vane to those on the entire turbines. The increased deposit rates on Vane 1, as shown in Figs. 10–11 and Figs. 13–14, generate the increasing trends of the capture efficiency as the Stokes number of the particle rises.This is consistent with the results from the uniform inlet case in Ref.However, because of the inlet swirl,the capture efficiency on Vane 2 has a drop for the moderate range of particle diameters in the LE-Pos case as well as for the large particle sizes in the MP-Pos and MP-Rev cases.Globally,the capture efficiency on Vane 2 is higher for smaller particles yet lower for larger particles compared to that on Vane 1. Furthermore, it is noted that the LE-Pos case has the highest percentage of the deposits on Vane 1 for a Stokes number of 7.79 × 10, and also has the highest percentage of the deposits on Vane 2 for a Stokes number of 5.41×10.

The overall capture efficiency on the whole turbine vanes is plotted in Fig. 16. Although increasing the particle size could lead to the decrease of the capture efficiency on Vane 2, the capture efficiency on Vane 1 always rises quickly with the particle size. It thereby generates a general trend of an increasing overall capture efficiency within the entire turbine passages by increasing the particle size, as shown in Fig. 16. A closer look at the bar chart shows that for the smaller particles,the overall capture efficiency for the four inlet swirl cases are comparative to one another. However, the gaps of the overall capture efficiency among the different inlet swirl cases become larger for the larger particles. In general, the two MP cases have higher overall capture efficiency relative to the two LE cases. In the MP cases, the positive swirl generates higher overall capture efficiency than the negative swirl in the range of d= 1–5 μm while in the LE cases, the negative swirl produces higher capture efficiency than the positive swirl for the particle diameters of 5–25 μm.

Fig.15 Capture efficiency of particles on Vanes 1 and 2 for various Stokes numbers(pie charts in plots indicate percentage of deposits on each vane to those on entire turbines).

Fig.16 Overall capture efficiency within entire turbine passages.

4. Conclusions

Particle migration and deposition patterns in a high-pressure turbine were numerically investigated in presence of aggressive inlet swirl.An integrated model of a combustor simulator and a turbine was used to replicate the combustor-turbine interaction, providing a more engine-representative inlet swirl condition for the particle-laden flow in the turbine nozzle guide vane.To address the effects of the inlet swirl in detail,two typical circumferential positions relative to the vane and positive and negative swirling directions were imposed on the inlet swirl.In addition to particle migration and deposition patterns for various particle sizes,flow and thermal fields in the absence of the particles were also shown to support the observations of the particles’ behaviors. The key conclusions obtained in this study are summarized as follows.

(1)The flow within the vane passage is strongly affected by the clocking position and the swirling direction of the inlet swirl. Generally, a positive swirl is more likely to move along the direction of the circumferential pressure gradient while a negative swirl has the capability to migrate against the pressure gradient. Moreover, the distorted flow due to the inlet swirl makes the heat flux rate on the vane to present a much nonuniform distribution pattern, which further aggravates the non-uniformities of the particle deposition on the vane surfaces.

(2) The migration of the particles in the air flow is dominated by the distorted flow, which allows to observe the footprints of the inlet swirl inside the passage from the particle concentration distributions. Therefore, most of the particles are swept towards the boundary walls of the vane passage following the pathlines of the swirling flow. However, increasing the particle size could alleviate this problem and make the trajectories of the particles relatively smoother.

(3) The presence of the inlet swirl generates a much nonuniform deposition patterns on the vane surfaces and most importantly, the deposition between the adjacent vanes is no longer balanced, which could add additional inlet flow nonuniformities to the downstream rotor and thus accelerate the aero-thermal performance degradation of the turbine stage.As established in previous literature,increasing particle size increases the deposits on the vane,but in the swirling flow,larger particles also significantly alter the deposit distribution patterns,indicating a strong coupling effect between the inlet swirl and the particle size.Overall,the mid-passage swirl cases could have higher deposit rates relative to the leading-edge swirl cases and the negative inlet swirl produces more deposition than the positive inlet swirl for larger particles (d> 12 μm).

This study fills the gap in the literature where the isolated effects of inlet swirl have previously not been documented.The fundamental principles obtained in this work are expected to enrich the knowledge of the deposition in the high-pressure turbine under more realistic inlet flow conditions and to provide guidance regarding better operation and maintenance for a gas turbine engine.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This study was co-supported by the National Nature Science Foundation of China (No. 51906185) and the National Postdoctoral Program for Innovative Talents of China (No.BX20180248).