Experimental Study of Impact of In-Service Deterioration on Aerodynamic Performance of High-Pressure Nozzle Guide Vanes

High-pressure (HP) nozzle guide vanes (NGVs) deteriorate with service time due to (see, for example [1]) oxidation, deposition of airborne contaminants, thermal fatigue cracking (e.g., of the thermal barrier coat (TBC) layer), and erosion due to impact-abrasion [2]. These effects lead to both local changes in surface roughness, and global changes in geometry. Thermal barrier coating spallation and cracking is common, as is the partial collapse of cooling holes (smaller effective size), or the blockage or geometry change of holes due to particulate deposition. Eroded trailing edges (TE), or TE burn-back due to overheating are also common. These effects generally reduced the aero-thermal performance of the parts. The purpose of this paper is to perform accurate aerodynamic characterization of engine-run parts. We first review the literature related to the impact of surface roughness and TE damage on aerodynamic performance.

There have been several cascade experiments to assess the impact of surface roughness on the aerodynamic performance of turbine components. Most of these studies have used simple simulated roughness elements (sandgrains, hemispheres, and truncated cones) [3–5] or surface finishes arising from manufacturing processes (e.g., machining processes, polishing processes, and coating types) [6–11], with only one study (to the authors knowledge) attempting to simulate engine-representative roughened surfaces on an airfoil [12]. A comprehensive review of literature (up to 2010) on effects due to roughness was provided by Bons [13]. As expected, the primary effects are earlier laminar-turbulent boundary layer transition, and an increased rate of thickening of the boundary layer (in both states), leading to higher aerodynamic loss. One exception is the case of low Reynolds number (Re) operation (low pressure turbines for instance) where an earlier roughness-induced transition has been shown to prevent a laminar separation on the suction surface (SS) of the airfoil, leading to lower losses than a smooth (separated) airfoil [10,13,14]. This general effect is—of course—well known.

We can differentiate between studies in which the surface roughness of the entire airfoil is varied evenly to represent possible manufacturing routes [3,4,6–10], for example, metal printing techniques, and those in which the roughness on only part of the surface is varied [5,11,12]. These latter studies are more representative of in-service roughening which appears to preferentially affect the pressure surface (PS) and early SS.

Kind et al. [5] performed aerodynamic surveys 0.4 axial chords (C_ax) downstream of a low-speed linear cascade of turbine blades to study the impact of roughness on profile loss. The inlet Re_C (based on tangential chord) was 3.0 × 10⁵. They tested 18 cascade configurations: one smooth cascade configuration in which the airfoil surface (PS and SS) and endwalls are reported to be aerodynamically smooth (roughness height not reported) and 17 other configurations in which spanwise-oriented bands of roughness (application of sand grains) were introduced at different locations around the airfoil. The roughness distribution most relevant to the current study was a roughened band covering approximately 30‒100% of the PS with the rest of the airfoil being smooth. For this distribution, two sandgrain roughness heights (k) were used: k/C = 4.5 × 10⁻³ and 6.3 × 10⁻³, where C is the tangential chord. They reported that at the design incidence angle the profile loss increased from the smooth configuration by approximately 30% and 40% for the two roughness heights, respectively.

Matsuda et al. [11] performed aerodynamic surveys 0.165 tangential chords downstream of a linear cascade of vanes to study the impact of airfoil and endwall roughness on profile loss and net secondary loss. Net secondary loss was defined as the difference between the total pressure loss measured in the secondary flow region, and the estimated profile loss (based on mid-span measurements). The exit Re_C (based on tangential chord) range was 0.30 × 10⁶ < Re_C < 1.0 × 10⁶ and inlet turbulence intensity (Tu) was 0.43%. They tested two configurations: a smooth cascade configuration in which the airfoil surface (PS and SS) and endwalls had a normalized maximum peak-to-valley roughness height R_z/C of 5.2 × 10⁻⁵ (typical of a milled surface of a new steam turbine vane); and a rough cascade configuration in which the PS and endwalls had R_z/C equal to 84.0 × 10⁻⁵ (approximately 16 times larger than the smooth configuration, and typical of a shot blasted surface), and with the SS essentially unchanged from the smooth case with R_z/C equal to 6.0 × 10⁻⁵. At Re_C = 0.87 × 10⁶, they found that the profile loss and net secondary loss for the rough configuration increased by approximately 36% and 80%, respectively, compared to the smooth configuration. The measured increase in loss was similar in magnitude to the study of Kind et al. [5], despite an almost order-of-magnitude difference in reported surface roughness height. This apparently surprising result may be explained by a combination of two factors: the peak-to-valley roughness height, R_z, and sandgrain roughness height, k, both fail to take account of shape and density of the roughness elements in the profile; the roughness types are different, one surface being artificially roughened (sand grains) and the other being the result of a manufacturing process. Differences in roughness type and quantification make the comparison between studies difficult.

Erickson et al. [12] studied the impact of in-service roughening on the aerodynamic performance of turbine vanes in a low-speed linear cascade over a wide range of Tu and Re. Because of the similarity of their study to the current experiment (in terms of roughness type, distribution, modeling method, height, and operating conditions of Re and Tu) we consider the Erickson et al. [12] study an important reference point when it comes to the aerodynamic impact of roughness. At the closest condition to our experiment an increase in midspan total pressure loss coefficient of 12% was measured. We devote a later section of this paper to a more detailed comparison of our work and that of Erickson et al. [12]. The purpose is to allow detailed comparison of very similar studies, in the hope of cross-validation of the results, and in the hope of developing a benchmark loss-enhancement factor for typical engine geometries under engine realistic conditions. Comparison to existing correlations would be desirable, but is problematic, because, as demonstrated in the extensive review of Bons [13], variation in roughness type (simulated, manufacturing-related, and service-related) appears to have a significant effect on the correlation, and therefore a single global correlation remains elusive. It may be the case that in the foreseeable future it is necessary to perform high-fidelity studies of the type presented in this paper, in the hope of direct rules-of-thumb for particular combinations of part-type and deterioration characteristic.

We now review the literature on the impact of TE damage on the aerodynamic characteristics of turbine components. Sjolander et al. [15] studied the impact of a semi-circular cutout in the TE of a turbine blade (to simulate damage caused by burn-back or erosion) on loss and flow turning angle. Using a low-speed linear cascade, they did aerodynamic surveys downstream of an undamaged blade and two blades with a semi-circular cutout in their TE with diameters equivalent to 15% and 25% of the blade tangential chord. Results showed that the wake downstream of the cutout was significantly modified, with local under-turning of the flow (up to 4.2 deg yaw angle deficit compared to a new blade) in the cutout region. They also observed the formation of counter-rotating streamwise vortices at the edges of the cutout, driven by the sharp gradients (with radius) in yaw angle. Similar flow structures have been observed in computational fluid dynamics (CFD) studies of vanes with TE damage by Di Mare et al. [16] and Meyer et al. [17]. Sjolander et al. reported that the mass-flux-average profile total pressure loss coefficient increased by 56% (compared to undamaged blade) for the deepest cutout.

Bouchard et al. [18] did aerodynamic surveys downstream of full rings of uncooled new and in-service deteriorated engine vanes in a transonic annular cascade. The vanes were those of a Rolls-Royce A-250 engine. The TE of the deteriorated vanes had small chordwise cracks, and some TE were significantly thermally distorted. No surface roughness measurements were reported in the paper. Profile loss measurements were taken one half of an axial chord downstream of the vanes at exit Mach number (M) M = 1.2. They reported an increase in profile loss of 51% for the deteriorated vane ring.

Although there is a small body of literature which presents results of studies of simulated deterioration features in isolation, there are very few studies indeed [18] which characterize components with in-service deterioration. Although the overall impact of deterioration could reasonably be expected to be well-modeled by superposition of the individual deterioration effects, for the purpose of building whole-life turbine performance models it is important to characterize real engine parts at various stages through the life cycle. There is very little data available for this task. The purpose of this paper—and the main distinction from previous studies—is to characterize the range of aerodynamic performance that exists for real-engine HP NGVs subject to in-service deterioration. The data provide a reference point for whole-life performance modeling. To this end, for 15 engine-run parts, we present detailed measurements of: surface roughness; coolant capacity; and aerodynamic performance. Measurements were conducted for parts operating under engine-representative conditions of exit M and exit Re. The data are to be used to benchmark whole-life modeling codes.

In this study we experimentally characterize a total of 17 real-engine vanes. The vane design was that from a modern large civil jet engine. The vanes were divided in two sets, referred to as set 1 and set 2. Set 1 was composed of one new vane and 12 mid-life vanes and set 2 was composed of one new vane and three end-of-life vanes. The new vane within each set has the same as-new geometry as the deteriorated vanes in the same set and acts as a baseline for that set. The deteriorated vanes within each set are from the same engine, and therefore were operated in the same environment and for the same number of cycles. The two sets operated in a non-sandy, yet not identical, environment. A summary of the operation history of both sets of vanes is given in Table 1.

Although the vanes in set 1 and set 2 are from the same engine-class, they are from different generations of the same engine and have slightly different geometries. Thus, comparisons between sets should be avoided. All vanes were fully featured, including film cooling holes, TE slots, and TBC. We identify individual vanes using their set number (S1 or S2) and vane number within the set (V1, V2, etc.).

Although experimental data were acquired for all vanes in both sets, in this paper we focus on data from four vanes for each set: one new vane and three deteriorated vanes. For the focused study, we chose the vanes with deterioration patterns representative of the full population of vanes. We refer to selected set 1 vanes as: S1-V1 (new); and S1-V2, S1-V3, and S1-V4 (mid-life). The corresponding nomenclature for set 2 is: S2-V1 (new); and S2-V2, S2-V3, and S2-V4 (end-of-life). Although we present detailed measurements for only this sample, we present statistics on the parent set of parts where space allows.

We now characterize the deterioration of the tested vanes. First we consider TE burn-back and TBC spallation, then we present surface roughness measurements.

Figure 1 shows a schematic of the vanes in set 1 (S1) and set 2 (S2). Film cooling holes and the TE slots are omitted for clarity. The vane outline is shown by a solid outline. Solid and dashed lines on the vane surface demarcate regions of TE burn-back and TBC spallation, respectively.

We consider first the selection of set 1 parts: one reference vane and three mid-life vanes. Vanes S1-V1 to S1-V4 are illustrated in Fig. 1(a). We see TE burn-back for vanes S1-V2 and S1-V3, with notches at approximately mid-span that extend to a depth equal to approximately 5% of the tangential chord and up to 18% of the span of the vane. There is a small region of TBC spallation at the leading edges (LE) close to the stagnation point for vanes S1-V3 and S1-V4. These regions are narrow and cover less than 10% of the span of the vanes. The TE burn-back and TBC spallation on vanes S1-V2 to S1-V4 are representative of the damage of the other deteriorated vanes in set 1.

Now we consider the set 2 parts: one reference vane and three end-of-life vanes. With reference to Fig. 1(b), we see both significant TE burn-back and significant TBC spallation. Looking first at the TE damage, vane S2-V2 was burned back up to 14.2% of the tangential chord and the damage covers 80% of the span of the vane. The TE of vane S2-V3 is burned back up to 9.6% of the tangential chord at midspan and the damage covers 40% of the span of the vane. The TE of vane S2-V4 is slightly cracked and distorted towards the SS, with TE TBC spallation on both the PS and SS extending up to 7% of the tangential chord over 35% of the span of the vane. It is suspected that the TE damage is the result of local overheating rather than erosion (by, for example, sand) because approximately half of the vanes from the same engine had relatively undeteriorated TE.

Now consider the LE region damage for the set 2 (end-of-life) vanes. The LE of all vanes in this set were severely damaged with TBC spallation covering up to 65% of the span of the vanes, and extending as far as 30% of the tangential chord on the PS of the vanes (see Fig. 1(b)). The spallation was so severe in some places that the metal was left unprotected. This is thought to be due to a combination of thermal fatigue cracking and oxidation. In the regions of most severe deterioration, primarily in the showerhead region of the parts, there was significant collapse (shrinkage of diameter, and damage to form) of the cooling holes, and cracking between holes.

A descriptive summary of the deterioration characteristics of both sets of vanes is given in Table 2.

In this section we present surface roughness measurements for the tested vanes. Profilometry measurements were performed using an Alicona InfiniteFocus profilometer (non-contact optical three-dimensional surface measurement system). Measurements were taken at three span fractions and three streamwise positions on the PS, and at three span fractions and four streamwise positions on the SS. These 21 locations are indicated in Fig. 2. Each profilometry measurement covered a 500 µm × 500 µm area and consisted of 250,000 (500 × 500) discrete points. The measurement resolution in the wall-normal direction (z) direction was 0.1 µm. The macroscopic surface shape was evaluated by cubic surface fitting and was removed from the measurement to only keep the surface profile. Figure 3 shows examples of the measured and processed surface profiles. The positive streamwise direction is along the x axis. The data show good measurement resolution. The new TBC surface profile (Fig. 3(a)) was measured at 50% span at position PS2 (Fig. 2) on vane S1-V1 and is characterized with a high density of shallow and blunt valleys. The deteriorated TBC surface profile (Fig. 3(b)) was measured at 50% span at position PS2 on vane S1-V2 and is characterized by sharper high peaks formed by deposits. The shape of the peaks appears to be independent of direction (streamwise or spanwise).

We represent roughness at the measurement locations with an equivalent sandgrain roughness height (k_s) using the methodology proposed by Bons [19]. We present k_s normalized by the tangential chord C of the vanes. Figure 4 shows k_s/C measured on vanes S1-V1 to S1-V4 for set 1 and S2-V1 to S2-V4 for set 2 (entire set).

Plots (a) and (b) show data for the PS of the vanes in set 1 and set 2, respectively. Plots (c) and (d) show data for the SS of the vanes in set 1 and set 2, respectively. Position labels indicate: the side of the vane (PS or SS); streamwise position (1–3 on PS and 1–4 on SS, as shown in Fig. 2); and span fraction (20%, 50%, and 80%). Each bar at a given position corresponds to a particular vane measurement.

Consider first the vanes in set 1 (mid-life) part data of Figs. 4(a) and 4(c). On the PS and early SS (SS/1/XX locations, where XX = 20, 50, or 80) we see a significant increase in k_s/C of the deteriorated vanes compared to the new vane. The mean increase in these 12 locations was 441%, with a maximum increase of 1056%. On the remainder of the SS (locations SS/2–4/XX) the deteriorated and new vanes are more similar, with a mean increase in roughness of 53.3%, and a maximum increase of 192%. This is in accord with the result of Bons et al. [20], who documented the surface roughness on nearly 100 in-service-deteriorated turbine components and noted that the PS and early SS suffer from in-service roughening much more than the late SS.

Now consider the set 2 (end-of-life) part data of Figs. 4(b) and 4(d). The early PS (locations PS/1/XX) and early SS (locations SS/1/XX) show a significant increase in roughness compared to the new vane, with a mean increase in roughness of 104% and a maximum increase of 723%. In all other locations on the vane the increase in roughness is smaller, with a mean increase of 49.5%, and a maximum increase of 201%.

Taking all vanes within set 1 (mid-life) (entire set of 12 deteriorated vanes and one new vane), the mean values of k_s/C across all measurement points were 0.109 × 10⁻³ for the new vane and 0.399 × 10⁻³ for the deteriorated vanes (average of 12 vanes). The average increase in roughness (compared to the new vanes) was 266%. This was strongly biased towards the PS and the early SS (locations SS/1/XX).

Taking the vanes within set 2 (end-of-life) as a whole (three deteriorated vanes in total and one new vane), the mean values of k_s/C across all measurement points were 0.103 × 10⁻³ for the new vane and 0.185 × 10⁻³ for the deteriorated vanes (average of three vanes). The average increase in roughness (compared the new vanes) was 79.6%. Again, this was strongly biased towards the PS and the early SS (locations SS/1/XX).

Although in this paper we focus on the airfoil surface, for completeness we note that the hub and casing platforms of the deteriorated vanes of both set 1 and set 2 were qualitatively rougher than new platforms, but without major geometrical deviations.

Aerodynamic measurements were conducted in the Engine Component AeroThermal (ECAT) facility at the University of Oxford [21]. This is a high technology readiness level (TRL) blowdown facility capable of testing an annular cascade of HP NGVs from operating engines at engine-representative conditions of Mach number, Reynolds number, and coolant-to-mainstream pressure ratio. A schematic cross section of the facility working section with the traverse system installed is presented in Fig. 5.

During a test, high-pressure air (initially stored in large tanks) is discharged through the cascade of vanes. Steady operating conditions can be achieved for approximately 60 s by the action of a pressure regulator upstream of the cascade. During this time aerodynamic traverse measurements are taken. The vanes are supplied with coolant from the hub and case (see Fig. 5), feeding the vane film cooling holes and the TE slot. Further details on the operation of the ECAT facility can be found in Refs. [21–23]. The operating conditions used in this study are summarized in Table 3.

Mainstream total pressure and total temperature measurements at the cascade inlet (p₀₁ and T₀₁) were performed using probe rakes (approximate axial location shown in Fig. 5) at four circumferential positions around the annulus. Hub and case coolant cavity total pressure and total temperature (p_0c and T_0c) were also measured at four circumferential positions. The vane exit static pressure (p_2,hub and p_2,case) was measured on six vanes (from a total of 40) with 28 tappings per vane (84 tappings in total) located on the hub and case platform overhangs approximately 5 mm downstream of the cascade TE plane (see Fig. 5).

Area surveys were conducted in an axial plane one-quarter of an axial chord downstream of the HP NGV TE using an automated radial-circumferential hub-mounted traverse gear. The location and approximate geometry of the traverse are shown in Fig. 5. The probe tip is approximately aligned with the mean flow angle, to ensure measurements were in the range of minimum error for the probe system. Aerodynamic measurements were taken using a five-hole probe (tip diameter 2.8 mm) calibrated over a wide Mach number range (0.3 ≤ M ≤ 1.4). The probe was calibrated to allow determination of total and static pressures, Mach number, and flow angles. Data on a full area-traverse plane are built from six discrete blocks (separate runs) which are then stitched together. A full-area traverse map contains approximately 60 traverse passes with a radial step size of approximately 1 mm. Better than 1 mm effective resolution is obtained in the circumferential direction. More details on the instrumentation and measurement processing for the ECAT facility can be found in Refs. [21,23].

Ex-situ capacity measurements were performed individually on all 17 vanes (2 new vanes and 15 deteriorated vanes) used in this study. It is necessary to perform ex-situ measurements on individual parts to separate effects on a per-vane basis.

It is—of course—impossible in practice to measure the change in in-situ coolant capacity on a vane-by-vane basis because there is a common feed to the entire vane ring. It is possible however, to accurately infer the per-vane in-situ capacity from the per-vane ex-situ capacity measurements. The following process was used:

• Ex-situ capacity measurements were performed on individual deteriorated vanes and non-deteriorated vanes. In these experiments the TE slot capacity characteristic and the capacity characteristic for all film cooling rows were separately determined. These experiments define the absolute ex-situ capacity characteristics of the individual vanes.

• The ratios between individual film row capacity characteristics were taken from the design intent (validated against earlier experiments). These ratios were assumed to be unchanged with deterioration. The inferred overall in-situ cooling capacity can be shown to be relatively independent of these ratios over a relatively wide range of variation.

• A low-order flow-capacity model for the in-situ part was developed in which—for each film row—an average film-row exit pressure at a particular vane operating point (p₂/p₀₁) was defined from CFD simulations at that operating point. Absolute capacity (all films and TE slot) come from the ex-situ experiments; ratios between rows come from the design intent. This low-order model allows prediction of individual capacity values for rows operating at particular local pressure ratios p_0c/p_x, where p_x is the local static pressure defined by the vane pressure ratio and the vane aerodynamics. This allows the in-situ capacity on a per-vane basis to be calculated as a function of vane operating point (p₂/p₀₁) and coolant-to-mainstream pressure ratio (p_0c/p₀₁). By summation the overall in-situ coolant capacity can be calculated as a function of vane operating point and coolant-to-mainstream pressure ratio.

• Separate in-situ experiments of overall coolant capacity were performed and compared to the predicted value of in-situ coolant capacity from the low-order model (based on the ex-situ measurements and CFD boundary conditions). These values agreed to within 1.1% giving confidence in this check-sum validation of the low-order model.

• Once the check-sum was performed, the low-order model could be used to predict overall film capacity and TE slot capacity on a per-vane basis. This can be done for any particular combination of p₂/p₀₁ and p_0c/p₀₁. This enables us to predict per-vane variation in film and TE-slot capacity for the individual deteriorated vanes of this study.

This is a simple process, relatively accurate to first-order. In a more sophisticated analysis a fully-coupled internal-external network loss model would be used, with individual deteriorated parts would be characterized on a per-film-row basis in the ex-situ experiments. This would enable extremely accurate predictions for the in-situ environment. Such an approach is discussed in Ref. [24]. The improved process is arduous in practice, however, and for the purpose of studying the overall change in in-situ coolant capacity from the situation for undamaged vane, we believe the simple process to be at least adequate.

Bench-test measurements were taken in an experiment in which the NGV was fed from both the hub and case using custom-made (plastic printed) feed plena sealed to the vane with wax sealant (to ensure complete removal). The feed plena were pressure-balanced using variable metered feeds fed from a constant pressure (regulated) supply. Total temperature and total pressure were measured upstream of the NGV coolant inlet using measurement rakes. The flow discharged directly to atmosphere through the film cooling holes and the TE slot. Test data were taken at five coolant-to-mainstream pressure ratios, p_0c/p_atm = 1.1, 1.2, 1.4, 1.6, and 1.8.

At the nominal coolant-to-mainstream pressure ratio and nominal vane pressure ratio (Table 3) the coolant-to-overall (mainstream plus coolant) mass flowrate ratio was approximately 10%, of which approximately one-tenth was the TE slot flow. Using the ex-situ capacity measurements we express proportional changes (from the undamaged vane) in ex-situ per-row coolant capacity (e.g., Γ_c,S1–VX − Γ_c,S1–V1)/Γ_c,S1–V1, where X = 2, 3, or 4). That is, changes between the individual deteriorated and undeteriorated parts. As these results are approximately independent of pressure ratio, we present average values (from all tested pressure ratios). For unchanged in-situ aerodynamics the proportional change in per-row in-situ capacity is the same as the proportional change in the per-row ex-situ capacity. The same statement is equally true for any group of cooling rows, and independently for the TE slot. In our particular tests we consider two groups: the TE slot; and the entire flow from 11 film cooling rows. These results are presented in a later section.

The experimental results are now presented. We will consider: coolant capacity measurements; local kinetic energy (KE) loss coefficient distributions; plane-average and mixed-out KE loss coefficients; comparisons to open-literature data of loss enhancement due to increased surface roughness; circumferential profiles of local KE loss coefficient; and radial distributions of flow angle.

In this section we consider the measurements of ex-situ coolant capacity which—we have explained—is a reasonable proxy for in-situ coolant capacity in our experiments. The proportional changes in coolant capacity (Γ_c,S1–VX − Γ_c,S1–V1)/Γ_c,S1–V1 for set 1 and (Γ_c,S2–VX − Γ_c,S2–V1)/Γ_c,S2–V1 for set 2) are presented in Table 4. The measurement process allows separation of the TE slot flow and the film flow. Proportional changes are relative to the reference (new) vane in each set.

Looking first at the set 1 data, we see a reduction in coolant capacity for all the mid-life deteriorated vanes for both film cooling holes and TE slot. For the film cooling flow the reduction in capacity is between ‒6.21% and ‒3.87% with an average (across three parts) of ‒5.15%, and for the TE slot the reduction is between −15.0% and −2.54% with an average of −10.3% (across three parts). The decrease in film coolant capacity appears to be caused by partial blockage of the holes due to particle deposition. This is marked A in Fig. 6(b). The condition of an undamaged film cooling hole is shown in Fig. 6(a). Close inspection of the film cooling holes of the mid-life vanes shows some level of deposition in all holes. The decrease in TE slots coolant capacity appears to be due to partial collapse of the TE slot, due to overheating and thermal distortion. This is shown in Fig. 6(c), marked B. Most mid-life vanes showed partial collapse in some areas, even on vanes without TE burn-back, but the effect was more advanced on vanes which also showed TE burn-back (e.g., S1-V2 and S1-V3). Taking the average across all vanes in set 1 (total of 12 mid-life vanes; not individually reported in Table 4) we find a mean reduction in film cooling flow capacity of ‒2.20% and a mean reduction in TE flow capacity of ‒4.87%. These average values are presented in Table 4. The difference in the mean reduction in film hole coolant capacity between the three selected vanes of set 1 (‒5.15%) and the parent set of 12 vanes (‒2.20%) is attributed to the random nature of the deposition process (even within the same engine), and the small sample size. Likewise the mean reduction in TE slot coolant capacity between the three selected vanes (‒10.3%) and the parent set of 12 vanes (‒4.87%) is explained by the fact that only three vanes from the parent set have TE burn-back: i.e., in selecting the vanes to be studied in detail, vanes with more significant TE burn-back were preferred, and these had greater associated collapse of the TE slot than average across the parent set.

Looking now at the data for set 2 (end-of-life parts) we see a reduction in coolant capacity between ‒6.27% and ‒5.43% for the film cooling holes (mean of ‒5.77% across three parts) and between ‒24.7% and ‒9.79% for the TE slot (mean of ‒16.8% across three parts). In contrast to the set 1 data, the decrease in coolant capacity of the film cooling holes appears to be primarily due to shrinkage of holes in the region of spalled TBC (spallation region shown in Fig. 1(b)), due to local over-heating of the part (similar effect to the TE slot collapse) This is shown in Fig. 6(d), marked C (LE of vane S2-V4). Interestingly, the effect is partially offset by enchained cracking between holes (marked D in Fig. 6(d)), leading to weeping flow through the cracks. As for set 1 parts, the decrease in TE flow for the set 2 parts is due to partial collapse of the TE slot. The radial extent of this was, on average, greater for the set 2 parts than the set 1 parts, on account of more severe TE overheating. This is consistent with greater radial extent of TE burn-back for the set 2 parts. Extreme burn-back (greater than the parts shown in the present study) would potentially lead to a reversing of this effect, due to an increase in slot width in the direction towards the LE of the vane. The impact of a change in film cooling flow capacity and TE flow capacity on aerodynamic loss will be considered in the next section.

Figure 7 shows the local distributions of ζ′(r, θ) downstream of the mid-life vanes S1-V1 to S1-V4 (top row, frames a–d respectively) and the end-of-life vanes S2-V1 to S2-V4 (bottom row, frames e–h respectively). Flow turning is clockwise when viewed from downstream, and data are viewed from downstream. The distribution of ζ′(r, θ) is normalized by the maximum value of ζ′(r, θ) measured across all the data presented.

Consider first the results for the new vanes, S1-V1 and S2-V1. There is a well-defined wake profile, which has a curve due to the compound lean in the aft region of the vane. The wake has a number of distinct maxima associated with regions of TE coolant ejection (for detailed analysis see [23]), which are intermittent on account of internal webs. The maximum value of ζ′(r, θ) in the wake is near the hub of the vane, where the vane exit Mach number is the highest. Compound lean and sweep in the vane geometry mean the traverse plane is closest to the TE at the hub, and furthest from the vane at mid-span. This is in accord with the result that the wake towards the hub is the least mixed (a second reason for the greatest peak value in this region), and the wake in the mid-span region is most mixed.

Now consider the set 1 (mid-life) deteriorated vanes. Vanes S1-V2 and S1-V3 have only minor TE burn-back (see Fig. 1(a)), but the distortion to the downstream wake is significant. In the region of the TE damage, the wake is under-turned with respect to the undamaged vane, leading to turning of the passage flow away from the SS and towards the PS. The regions of under-turned flow (co-incident with the TE burn-back regions of Fig. 1(a)) are marked A and B in Figs. 7(b) and 7(c) respectively. Vane S1-V4 (Fig. 7(d)) did not exhibit TE burn-back and the wake is deeper than that of the reference vane (Fig. 7(a))—due to greater surface roughness—but of relatively similar form. All of the mid-life vanes assessed had greater profile loss than the undamaged vane, which was attributed to the increased surface roughness causing greater boundary layer loss (combination of earlier transition, but also greater growth rate within the turbulent boundary layer).

The deteriorated set 2 (end-of-life) vanes have more significant TE burn-back (for extents see Fig. 1(b)), leading to gross disruption of the wake due to severe under-turning, and—in the case of S2-V2 and S2-V3—the formation of counter-rotating vortices (marked C and D in Fig. 7) either side of the more severe burnt-back regions (caused by strong radial gradient of whirl angle). Similar effects have been observed in Refs. [15–17]. Between counter-rotating vortices, there are patches of low loss between the vortices, as the lossy flow is rolled-up into the vortex. The small cracks and bends at the TE of vane S2-V4 have caused a significant local broadening of the wake (marked E in Fig. 7). In the profile loss regions undisturbed by TE damage, the wake is thicker and deeper than the reference vane due to increased surface roughness of the part.

We now examine the secondary flow regions in more detail. Local distributions of ζ′(r, θ) in the near-endwall regions are shown in Fig. 8. Secondary flow loss cores in the casing region (Figs. 8(a)–8(h)) are difficult to identify for two reasons: they are less intense than at the hub due to the lower Mach number; there is downwash on the vane surface, spreading the boundary layer fluid throughout the wake (see e.g., [23]) and causing coalescence of the flow structures. In contrast to the casing data, in the near-hub region (Figs. 8(i)–8(p)) the secondary loss cores are well-defined: for the reference vanes S1-V1 (Fig. 8(i)) and S2-V1 (Fig. 8(m)) the SS corner vortex and passage vortex are marked A and B respectively. For the corner vortex and passage vortex, the peak measured values of local normalized KE loss coefficient were approximately 80% and 66%, respectively, for both reference vanes. The radial locations of the peaks were 2% and 10% span fraction respectively. For all deteriorated vanes the secondary loss cores are both more intense and greater in size than for their corresponding reference vane.

In this section we consider both plane-average KE loss coefficients, and mixed-out KE loss coefficients for the vanes. The plane-average coefficient may be taken to represent the loss already manifested at the measurement plane. The mixed-out KE loss coefficient can be taken to represent the sum of the loss already manifested, and the unavoidable loss as the result of the mixing out of gradients already present within the flow (and unlikely to contribute usefully to work): i.e., the secondary kinetic energy (SKE) within the flow.

Consider first the data for set 1 (mid-life). For the reference vane, the normalized mixed-out KE loss coefficient is by definition unity. The normalized plane-average KE loss coefficient and SKE are 0.65 and 0.35 (sum to unity). The relatively large SKE (0.35 of the mixed-out loss) is accounted for by significant unmixed wake (traverse plane relatively close to TE). Now consider the set 1 deteriorated vanes. The average changes in $ζ″″$⁠, $ζ″″′$⁠, and SKE for vanes S1-V(2–4) were 14.1%, 8.72%, and ‒2.13%. Corresponding values for the full set of 12 vanes—vanes S1-V(2–12)—are extremely similar: 13.6%, 7.81%, and ‒2.57%, respectively. The vane-to-vane spread in data was very low, suggesting high similarity between parts. The increase in plane-average KE loss coefficient (average of 13.6% across all parts) is thought to be due simply to an increase in roughness, and associated boundary layer thickening. The SKE remains almost constant (average change of only ‒2.57%), which is explained by the fact that the flow structure is essentially unchanged (slight change in wake thickness). The corresponding average change in mixed-out loss is 7.81%.

Looking at the data for set 2 (end-of-life), the reference vane, has normalized plane-average KE loss coefficient and SKE values of 0.65 and 0.35. These are identical to the values for set 1 parts, suggesting high consistency between sets. Looking at the set 2 deteriorated vanes (vanes S2-V(2–4)) we find the average increase (across three vanes) in $ζ″″$⁠, $ζ″″′$⁠, and SKE to be 22.7%, 18.8%, and 14.9%, respectively. The loss coefficients changes are significantly greater for set 2 than for set 1 both because of the flow-structure changes caused by severe TE burn-back, and the effects of increased surface roughness. The vane-to-vane spread in data is also higher that for set 1, on account of significant differences in the particular TE damage (see Fig. 1(b)). The increase in plane-average KE loss coefficient (average of 22.7% across all parts; approximately double that for set 1) is associated with both an increase in roughness and an increase in—already manifested—secondary flow loss at the mixing plane. In contrast to the set 1 data, the SKE change is significant (average change of 14.9%), and is caused by significant structural changes in the flow (see discussion above, noting—in particular—streamwise vortices marked C and D in Fig. 7) with the average dominated by the changes for vanes S2-V2 and S2-V3 (these have the most significant TE burn-back; see Fig. 1(b)), 34.3% and 14.3% respectively. The corresponding average change in mixed-out loss is 18.8%, approximately double that for set 1.

So far as whole-life turbine performance modeling is concerned, we suggest preliminary (limited data sets) enhancement factors for mixed-out row loss of approximately 8% for mid-life parts, and 19% for end-of-life parts. Taking a typical mixed-out row total pressure loss of 6% [23] for new parts, these would translate to approximately −0.49% and −1.19% on row efficiency, respectively. For stage loss, an additional enhancement factor might be expected for the case of end-of-life parts (but not mid-life parts) on account of the extreme yaw angle variation in the vicinity of the TE burn-back, which would be expected to cause additional losses due to incidence variation in the rotor frame.

We now consider the individual contributions of the change in film cooling flow capacity and TE slot flow capacity (due to in-service deterioration) on the overall change in aerodynamic loss. For this purpose we used the Hartsel [27] film cooling flow loss model and a TE slot flow loss model which we develop for our specific vane geometry and operating conditions.

We first consider the film cooling loss model. This model is based on that of Hartsel [27], in which the flow is considered to be comprised of a mixing layer and mainstream layer, interacting at constant static pressure in the surface normal direction, but with the possibility of acceleration in the streamwise direction. Relative row capacities were calculated using nominal hole diameters, with a scaling factor (equivalent to a common discharge coefficient) chosen to match the overall coolant capacity of the model to that measured in the experiment. In the model individual rows exhaust to particular static pressures determined using CFD (see Burdett et al. [28] for more details). Changes in film cooling flow capacity due to deterioration were assumed to affect all film cooling holes uniformly. Using the model, we can calculate the predicted change in mixed-out KE loss coefficient (normalized by $ζS1−V1″″′$⁠) as a function of a change in film cooling flow capacity: a so-called exchange rate between the two parameters. This function is shown in Fig. 9. We observe that a reduction of film cooling flow from the design condition leads to a reduction in overall loss. This trend was expected since in the Hartsel [27] model, the mixing loss (between the mainstream and the ejected coolant) is proportional to the coolant-to-mainstream mass flow ratio, which has reduced due to deterioration. If deterioration acts to reduce film capacity and increase roughness and TE damage, the reduced film loss acts to mitigate the increase in aerodynamic loss due to the increase surface roughness and TE burn-back.

For the set 1 parts, the mean change in film cooling flow capacity was −2.20% (see Table 4), leading to a predicted change in mixed-out KE loss coefficient $ζ″″′$ of −0.91%. For the set 2 parts the mean change in film cooling flow capacity was −5.77% (see Table 4), leading to a predicted change in mixed-out KE loss coefficient $ζ″″′$ of ‒2.36%. It should be noted that the vanes in set 2 have heavily damaged LE with distorted film cooling holes and cracks (see Fig. 6(d)), and changes in the aerodynamics of the mixing process are beyond the scope of the model. We must regard the results for set 2 parts as therefore only indicative of likely trends rather than quantitatively robust.

We now consider the TE slot flow loss model. The model is that presented in Burdett and Povey [29], based on earlier work by Stewart [30] and then Deckers and Denton [31]. The model uses a mass-momentum control volume method to determine the mixed-out loss coefficient for a particular TE geometry and TE slot mass flowrate. We assume that the change in TE slot flow capacity due to deterioration is uniform across the entire span of the slot, and represented by the measured change in TE slot flow capacity. The predicted change in mixed-out KE loss coefficient (normalized by $ζS1−V1″″′$⁠) as a function of the change in TE slot flow capacity is shown in Fig. 9. We observe that a reduction of TE slot flow from the design condition leads to an increase in overall loss. Deckers and Denton [31] showed that at a given coolant ejection pressure ratio, an increase in coolant TE slot flow capacity increases the base region pressure of an airfoil, and thus reduces the mixed-out KE loss coefficient (lower pressure drag) of the vane. The mean changes in TE slot flow capacity for set 1 and 2 parts (see Table 4) were −4.87% and −16.8% respectively, leading to predicted increases in mixed-out KE loss coefficient $(ζ″″′)$ of 2.18% and 7.62%, respectively. For our deteriorated parts both the film capacity and TE slot flow capacity are reduced, but the modeled changes in KE loss coefficient are in opposite directions.

Taking the results of these models in combination with the measured overall change in mixed-out KE loss coefficient with deterioration (Table 5), we can estimate the change in mixed-out loss attributable to aerodynamic changes not included in the loss models. That is, effects such as: boundary layer changes caused by surface roughness and spalling; changes in secondary flow behavior caused by TE burn-back; and additional loss introduced at the point of coolant injection not captured in the basic model. The estimated values for additional aerodynamic loss were 6.54% and 13.5% for set 1 and set 2 parts, respectively. The measured overall loss (Table 5) and the predicted (loss models) and inferred (by difference) contributions are represented diagrammatically in Fig. 10.

We now consider the circumferential profiles of local KE loss coefficient, ζ′(θ), downstream of the vanes. Figures 11 and 12 show the circumferential profiles of ζ′(θ) at 10%, 50%, and 90% span downstream of vanes of set 1 (S1-V1 to S1-V4) and set 2 (S2-V1 to S2-V4), respectively. Each profile is radially averaged over ±5% at the relevant span location, and is circumferentially centered at its peak value. The distributions ζ′(θ) are normalized by the maximum value of ζ′(r, θ) measured across all the data presented.

First consider the reference vanes: S1-V1 in Fig. 11, and S2-V1 in Fig. 12. As expected, the wake profiles have a quasi-normal distribution. The wake profiles at 10% span are narrower and have a greater peak value of ζ′(θ) than at 50% and 90% span. This is primarily due to spanwise variation of the TE position with respect to the traverse plane (see earlier discussion). The gradient of ζ′(θ) is slightly greater moving away from the SS than the PS. This is thought to be due to differences in the circumferential distributions of the separation loss contribution to overall loss, arising due to greater TE thickness on the PS than on the SS (design is of SS TE overhang style; see, for example, [29]). This effect opposes, and is thought to be more dominant than, the significantly larger boundary layer momentum thickness on the SS at the TE (for further discussion see [23,29]).

Values for these metrics evaluated at 50% span are presented in Table 6, both for the three deteriorated vanes of Fig. 11, and for the parent set of 12 vanes. Taking an average across all 12 vanes, the mean changes in peak height, peak width, and integrated loss were 23.9%, ‒1.67%, and 29.8% respectively. That is, there is a substantial increase in overall loss, but relatively little thickening of the wake. This is because the wake width at the traverse plane is dominated by the width of the separated TE region (or base region), as opposed to the boundary layer thickness (more details in Ref. [23]), into which both the boundary layer loss and so-called pressure drag of the base region are subsumed.

Based on this statistically meaningful set of vanes (12 in total) we conclude that mid-life vanes can be characterized by a loss increase of approximately 30%, with no substantial change in the wake width (rotor forcing implications) or profile shape. These numbers could be used in preliminary design for whole-life modeling.

Now consider the wake profiles for the deteriorated set 2 (end-of-life) vanes. For this set, the wake profile shapes for the deteriorated vanes differ from the reference vane by more than for the set 1 comparisons, on account of significant vortex activity associated with strong radial whirl angle variation in the regions of the TE burn-back. At the traverse plane, these secondary flows affect all span locations, but with a significant impact at midspan (where the burn-back is most severe). This leads to more local variation in the wake depth and width than for parts without TE burn-back. For all the three tested deteriorated vanes at all span locations, there was an increase in peak height, with the greatest increases in peak height being substantially larger than for the set 1 parts. That is, the peak loss was greater, but there was more randomness in the profile due to significant secondary flow activity. The mean changes (across three vanes) in peak height, peak width, and integrated loss were 46.7%, ‒1.99%, and 24.5% respectively. These are summarized in Table 6.

Based on the relatively small sample (three parts) we conclude that end-of-life vanes can be characterized by a loss increases in the range ‒20.4–61.8% (relatively low confidence due to small statistical set). There is significant part-to-part variability in the integrated loss, but also in the radial distribution of loss, the wake shape, and the associated secondary flow, leading to greater variability in the downstream rotor inlet flow.

In this section we perform a detailed comparison of the measured changes in midspan profile total pressure loss—with increased surface roughness—to the results of study of Erickson et al. [12]. We consider this useful because of the high degree of similarity between the experiments. Erickson et al. tested two configurations: a reference cascade configuration in which the airfoil surface (PS and SS) has an aerodynamically smooth surface (roughness height not reported); and a rough cascade configuration in which the entire PS and the first 10% of the SS had a surface profile designed to simulate in-service roughening (the remaining SS had the same surface finish as the reference case). The reported k_s/C for the roughened surface was 0.98 × 10⁻³, using the same reporting methodology as the present study [19].

For the comparison we consider only two vanes: the reference vane for set 1 (S1-V1); and vane S1-V10, which has no TE burn-back, and minimal changes in film cooling flow capacity and TE slot flow capacity (‒0.31% and ‒1.21% respectively). These vanes were chosen to isolate the effect of roughness without conflating it with effects due to other deterioration features. For part S1-V10, average values of roughness were k_s/C = 1.02 × 10⁻³ for the PS and early SS, and k_s/C = 0.29 × 10⁻³ for the rest of the SS. These values were very similar to those in the Erickson et al. [12] study. Table 7 compares the two studies in terms of: experimental facility; operating conditions; details of tested components including roughness. The primary difference between the studies was unmatched exit Mach number (0.05–0.20 for Erickson et al. versus 0.92 for the present study).

Figure 13 shows a comparison between the change in Y_p (due to increased surface roughness) reported by Erickson et al. and reported in the present study. In the present study an increase in Y_p (from reference vane S1-V1) of 12.2% was measured, for Tu = 12%. Linearly interpolating the Erickson et al. data (between Tu = 8.50% and Tu = 13.5%, and between Re = 1.0 × 10⁶ and Re = 2.0 × 10⁶) gives an increase in Y_p of 15.2%.

We take these results to be in good agreement, with the discrepancy potentially arising due to higher non-dimensional roughness for our reference (k_s/C = 0.11 × 10⁻³) than the vane used in the Erickson et al. study (reported to be aerodynamically smooth, defined by k_s/C < 100/Re_C, i.e., k_s/C < 0.05 × 10⁻³). That is, we would expect our reference vane loss to be slightly higher than in the Erickson et al. study. We conclude the following: agreement between the results can be taken as evidence of cross-validation; from a whole-life modeling perspective, we propose profile loss enhancement factors in the range 12.2% to 15.2% for mid-life nozzle guide vanes with k_s/C ∼ 1.00 × 10⁻³ on the PS and early SS, where the lower end of the range can be taken for parts with TBC coating, and the upper end for parts which are initially aerodynamically smooth.

We now consider radial flow-angle distributions downstream of the vanes. Pitch and yaw angles (α₂ and β₂, respectively) are defined in Fig. 14.

Spanwise distributions of circumferentially mass-flux-average flow angle (pitch angle, $α¯2$⁠, and yaw angle, $β¯2$⁠) downstream of vanes in set 1 (S1-V1 to S1-V4) and set 2 (S2-V1 to S2-V4) are presented in Figs. 15 and 16 respectively.

First consider the pitch angle results for the set 1 (mid-life) vanes (Fig. 15). Radial distributions of pitch angle, $α¯2$⁠, are shown in Fig. 15(a). For the reference vane (S1-V1) pitch angle is small and positive and decreases slowly with increasing radius. This is thought to be due to the more prominent annulus hade at the hub than the case. That is, an overall annulus contraction, dominated by the hub annulus line hade (see Fig. 2). The pitch angle trend is approximately linear with radius, varying between 1.9 deg < $α¯2$ < 5.3 deg. The mid-life vanes (S1-V2 to S1-V4) have the following characteristic differences from the reference vane: slightly greater mean pitch angle (by 0.2 deg on average); slightly lower spanwise variation (by ‒1.1 deg per unit of span on average); and slightly greater local variability from a linear trend (mean RMS variation of 0.5 deg from a linear trend compared to 0.4 deg for the reference vane). The small increase in both mean pitch angle and radial variation of pitch angle are not thought to arise from deterioration effects, but rather are thought to be due to quasi-random variation in the manufacturing of the vanes. The increased local variation is thought to arise from increased secondary flow activity, associated with surface defects, increased boundary layer thickness (and associated intensification of vortex features), and TE notching. Vane S1-V3 has the largest region of TE burn-back (see Fig. 1(a)), and this is clearly associated with a local deviation in $α¯2$⁠. This is marked A in Fig. 15(a). The estimated value of Δ$α¯2$ (deviation from a locally linear trend for the particular vane) is Δ$α¯2$ = ‒0.7 deg.

Now consider radial yaw angle distribution for the mid-life vanes (Fig. 15(b)). Looking at the reference vane, yaw angle increases with radius according to the particular forced vortex design of the vane, varying approximately linearly between 74.1 deg $<β¯2<$ 80.3 deg. The effect of deterioration (S1-V2 to S1-V4) is threefold: to very slightly decrease the mean yaw angle (by 0.5 deg on average); to cause slightly more local variation in hub and case secondary flow regions (2–10% span; 90–98% span); and to cause strong local deviations in regions of TE burn-back. An example of this last effect is marked B in Fig. 15(b), for which the deviation from the locally linear trend for the particular vane was approximately $Δβ¯2$ = ‒1.9 deg.

Now consider the corresponding results for the set 2 (end-of-life) vanes. The flow angle distributions downstream of the baseline vane (S2-V2) have similar trends to the baseline vane of set 1 (S1-V1). The aerodynamic vane standard was essentially identical so this was expected. For the deteriorated vanes, in regions of severe TE burn-back (see Fig. 1(b)) there are significant variations in both pitch and yaw angles. The yaw angle variation arises because of under turning in regions of burn-back (see, for example, features E and F in Fig. 16(b)). Pitch angle variation arises because of significant secondary flow associated with the strong shear (and associated secondary flow formation) at the edges of the burn-back regions (see, for example, features C and D in Fig. 16(a)). For the most severe burn-back (vane S2-V2) the local under turning (deviation from local trend) was estimated to be $Δβ¯2$ = ‒6.1 deg. The corresponding variation in local pitch angle was estimated to be $Δα¯2$ = ‒1.8 deg.

The trend was constrained to go through the origin for obvious physical reasons. A slightly better fit of set 1 data than set 2 data, and a slightly higher gradient for set 2 data than set 1 data (when individually fitted, not shown) may be explained by the greater radial extent of set 2 burn-back features, leading to less diminution of the yaw angle deviation effect in the intermediate mixing process (between vane TE and measurement plane). The empirical correlation (9) is proposed for preliminary whole-life performance assessment. The correlation would be improved with further experimental data taking separate account of both burn-back depth and radial extent.

In this paper, 15 real-engine in-service-deteriorated HP NGVs with a broad variety of deterioration features were studied for the purpose of describing and quantifying their aerodynamic performance, as a step towards whole-life engine modeling. The deterioration features included increased surface roughness; thermal barrier coating spallation; damaged film cooling holes; and trailing edge burn-back. The study was split into six areas, from which the following conclusions were drawn:

1. Surface roughness. In-service roughening was found to primarily affect the PS and early SS, but not the late SS. For vanes initially coated with TBC, typical mean equivalent sandgrain roughness height was found to increase on the PS and early SS from k_s/C = 0.109 × 10⁻³ for new vanes, to k_s/C = 0.590 × 10⁻³ and k_s/C = 0.222 × 10⁻³ for mid-life and end-of-life vanes, respectively. These are increases of 441% and 104%—on average—for mid-life and end-of-life vanes in this study.

2. Coolant flow capacity. Coolant flow capacity decreased with service time both due to accumulation of deposits in the film cooling holes, and partial collapse of film cooling holes and TE slots due to overheating. For mid-life vanes, the mean change in coolant capacity was ‒2.20% for the film cooling holes and ‒4.87% for the TE slots. Corresponding values for end-of-life vanes were ‒5.77% and ‒16.8% respectively.

3. Overall aerodynamic loss. In-service deterioration was found to increase aerodynamic loss. The overall mean enhancement for mixed-out row KE loss was found to be approximately 8% for mid-life vanes, and 19% for end-of-life vanes. We propose these enhancement factors as rules of thumb for whole-life engine modeling. Mechanisms for increased mixed-out loss include: boundary layer thickening on the vane surface and endwalls due to increased surface roughness, causing change in transition point, and change in rate of growth of the boundary layer; streamwise vortex generation (and associated higher residual SKE) in the downstream wake caused by TE burn-back leading to high gradients of yaw angle; reduction in TE slot flow leading to increased base loss. In addition to these effects it is believed that the reduced cooling flow reduces loss, partly mitigating these increases (but leading to excessive part temperature).

4. Separated loss components. Using film cooling and TE slot loss models we show that of the 7.81% and 18.8% loss enhancements associated with mid-life and end-of-life deterioration, changes of ‒0.91% and ‒2.36% can be attributed to reduced film cooling flow (reduced loss), and 2.18% and 7.62% can be attributed to reduced TE flow (increased loss). The remaining 6.54% and 13.5%, respectively, are attributable to aerodynamic changes not included in the loss models, for example: boundary layer changes caused by surface roughness and spalling; changes in secondary flow behavior caused by TE burn-back; and additional loss introduced at the point of coolant injection not captured in the basic model.

5. Profile loss for undamaged mid-life parts. Taking the results of the present study, and the study of Erickson et al. [12] together, we conclude that in-service deterioration increase profile loss of mid-life undamaged parts with k_s/C ∼ 1.00 × 10⁻³ on their PS and early SS by between 12.2% and 15.2%, where the lower value corresponds to parts initially TBC coated, and the upper value for parts which were initially aerodynamically smooth.

6. Downstream flow angles: TE burn-back was found to locally reduce downstream pitch and yaw flow angles. A simple empirical correlation between maximum yaw angle deviation and maximum TE burn-back depth was presented, and is proposed for preliminary whole-life performance assessment (stage modeling).

This paper addresses a gap in literature, by providing the first detailed analysis of the overall impact of combined in-service deterioration features on the overall aerodynamic performance of HP NGVs. It is hoped that the results from this paper will be a useful step towards whole-life engine performance modeling and assessment.

The financial support of Rolls-Royce plc is gratefully acknowledged. Nafiz Chowdhury and Daniel Burdett are thanked for their support with the experiments. David Newman is thanked for his support with the coolant capacity measurements and technical advice. The Laboratory for In-situ Microscopy and Analysis (LIMA) at the University of Oxford is also thanked for support of the surface roughness measurements.

There are no conflicts of interest.

The authors attest that all data for this study are included in the paper.

ax =

axial position, m

C =

tangential chord, m

C_ax =

axial chord, m

D =

trailing edge damage depth, m

D_max =

maximum trailing edge damage depth, m

IL =

integrated loss, –

k =

sandgrain roughness height, m

k_s =

equivalent sandgrain roughness height, m

$m˙$ =

mass flowrate, kg/s

M =

Mach number, –

p =

static pressure, Pa

p₀ =

total pressure, Pa

$p¯$ =

plane-average static pressure, Pa

$p0¯$ =

plane-average total pressure, Pa

$p¯′$ =

mixed-out static pressure, Pa

$p0¯′$ =

mixed-out total pressure, Pa

r =

radial position, m

Re_C =

exit Reynolds number based on tangential chord, –

R_z =

maximum peak-to-valley roughness height, m

SKE =

secondary kinetic energy, –

T =

static temperature, K

T₀ =

total temperature, K

Tu =

inlet turbulence intensity, –

x =

streamwise distance, m

y =

spanwise distance, m

Y_p =

plane-average total pressure loss coefficient, –

z =

wall-normal distance, m

α =

pitch angle, °

$α¯$ =

circumferentially mass-flux-average pitch angle, °

β =

yaw angle, °

$β¯$ =

circumferentially mass-flux-average yaw angle, °

γ =

ratio of specific heat capacities, –

Γ =

capacity, kg/s K^1/2/Pa

ζ =

kinetic energy loss coefficient, –

ζ′ =

local kinetic energy loss coefficient, –

ζ″″ =

plane-average kinetic energy loss coefficient, –

ζ″″ =

mixed-out kinetic energy loss coefficient, –

θ =

normalized (by vane pitch) circumferential position, –

χ =

(γ − 1)/γ, –

1 =

upstream of HP NGVs

2 =

downstream of HP NGVs

atm =

atmospheric

c =

coolant

is =

isentropic

ref =

reference

CFD =

computational fluid dynamics

ECAT =

Engine Component AeroThermal (facility)

EXP =

experiment

HP =

high pressure

KE =

kinetic energy

LE =

leading edge

NGV =

nozzle guide vane

PS =

pressure surface

SS =

suction surface

TBC =

thermal barrier coating

TE =

trailing edge

TRL =

technology readiness level