Performance Metrics and Experimental Testing of Erosion-Resistant Compressor Blade Coatings

One common source of foreign object damage encountered by aircraft with gas turbine engines is sand ingestion, which often occurs when these aircraft operate in arid or desert regions. Due to the greater hardness of the sand compared to steel or titanium compressor blades, as well as the high engine inlet velocities, erosion of the compressor blades will result if exposure time and sand concentration are severe enough.

In order to mitigate sand erosion damage to fan or compressor blades, various protective coatings have been developed, tested, manufactured, and applied to blades. Currently, test methods for compressor coating evaluation using realistic conditions have been minimal. Some full-scale tests have been performed on gas turbine engines with coated blades such as Dunn et al. [1]. However, this is very expensive and impractical for most organizations to carry out. To date, the majority of testing on erosion-resistant coatings has been performed using stationary sandblasting of coated coupons or blades. Examples include: Immarigeon et al. [2] (using the ASTM G76-83 test standard, a previous version of ASTM G76-07), Gorokhovsky et al. [3] and Muboyadzhyan [4]. Linear cascade testing of 3–4 stationary coated blades has also been conducted: Klein and Simpson [5] and Tabakoff and Mason [6]. Rotating rigs were constructed by Balan and Tabakoff [7] and Ghenaiet et al. [8], but both used aluminum blades, which are not representative of gas turbine engine compressors.

The main benefit of the rotating rig over stationary coupon and linear cascade sandblasting is the presence of realistic secondary flows. An axial compressor stage is a three-dimensional machine spinning at a high rotational speed, shrouded by a nonrotating casing. The combination of the spacing between the blades and the interaction of the rotating blades and the casing gives rise to secondary flows. These flows are important because they result in performance loss in addition to that caused by profile drag and off-design inlet flow [9]. If these secondary flows become laden with sand, their characteristic vortices would lead to scrubbing erosion on the blades' surfaces. Although the pressure losses due to these secondary flows are significant, for this work, the presence, nature, and effects of the secondary flows were of most interest.

Due to the complexity of the flows in a rotating axial compressor, the shape of axial compressor blades is carefully designed in order to maximize performance. Any deviation from the design airfoil profile will result in a loss in performance. Literature [1,5–8] shows that the three major areas which sustain damage due to engine debris or sand ingestion are the LE, blade tip, and TE. Also, increased blade surface roughness has a deleterious effect on performance as shown by Leipold et al. [10] and Back et al. [11].

If a rig could be built to erode rotating compressor blades using sand under controlled conditions, then pertinent research could be conducted to determine erosion mechanisms and roles. Additionally, erosion-resistant coatings could be applied to the blades, tested and performance-evaluated before (BE) application on full-scale and proprietary gas turbine engine axial compressors.

The goal of this research was to design, construct, commission, and conduct proof-of-concept testing of a rotating erosion test rig. This rig would use a carefully selected, operationally representative sand flow rate to erode rotating representative steel compressor blades under controlled conditions. Erosion testing was conducted on bare 17-4PH steel blades as well as on blades coated with titanium nitride (TiN) or a National Research Council of Canada (NRC) designed chromium–aluminum–titanium nitride (CrAlTiN) blade coating. Erosion results of the blades were then compared. The thickness of the erosion-resistant coatings was determined based on the literature. This paper provides details on the performance metrics and erosion test results.

The Royal Military College of Canada (RMC) turbomachinery erosion rig is a new experimental test rig capable of rotating 16 blade assemblies at speeds up to 12,000 rpm. The blade assemblies are inserted into a vibrationally balanced 20.32 cm (8 in.) diameter AISI 4340 steel rotor that is driven by an aircraft gas turbine compressor (GTC) gearbox and a 20 HP electric motor. The sand is injected into the rig via a gravity-fed hopper, with a venturi generating local low pressure at the sand inlet point to the rig. A centrifugal blower supplied the air mass flow to the rig. An operator station was used for rig condition monitoring tasks, such as rotor vibrations, oil flow rate and temperatures, visual oil flow and sand flow confirmation, and rotor rpm during erosion testing. A schematic of the rig is shown in Fig. 1, and a full description can be found in Ref. [12].

In order for the RMC erosion rig to have realistic applications to in-service gas turbine engines, 17-4PH (precipitation hardened) stainless steel was used for the blades. The rotor, dovetail inserts and blades were electric discharge machined (EDM), precisely rendering the complex shapes and geometries.

The blade profile was NACA 6505 with rounded LE and TE, termed V-103. The blades were 1.27 cm (0.50 in.) tall and had a chord of 2.67 cm (1.05 in.). The dovetail inserts were 2.54 cm (1.0 in.) wide and 5.84 cm (2.30 in.) long. The inlet airflow angle was set such that the air encountered the blade LE at a 0 deg local angle of attack (AOA), negating any requirement for inlet guide vanes. To obviate the requirement for a larger electric motor, a two-blade arrangement was devised. The front (upstream) blade would act as the compressor blade and the test article. The rear (downstream) blade would simply redirect the flow such that it would exit the rotor parallel to the relative incidence flow. Therefore, the resulting static pressure increase across the rotor was practically zero, neglecting windage and friction losses.

A blade stagger angle of 27.7 deg was selected for the upstream blade, resulting in an airflow angle (α) of 47.5 deg and an axial chord of 2.36 cm (0.930 in.). Figure 2 shows a summary of the resultant velocity triangles at the inlet to the upstream blade. The blades' midspan was used for setting the rotational speed U. During erosion testing, the average achieved Mach number at the blade LE (Ma_LE) was 0.50. The rotor rpm was adjusted such that the air encountered the blade LE at a 0 deg AOA, based on the room temperature and pressure conditions during each test.

The downstream blade had to be located at least 12% of the axial chord downstream from the upstream blade in order to minimize flow disturbances near the TE of the upstream blade. This was based on Benini and Toffolo's work [13] on optimal axial distances between compressor rotor and stator stages. The resulting axial spacing between the upstream and downstream blades on the dovetail was set at 1.04 cm (0.408 in.), 43.9% of the axial chord. A maximum distance of 24% axial chord was suggested [13]; however, this was likely due to the fact that the rotor blades in a gas turbine engine's axial compressor rotate very rapidly while the stator blades are stationary. For the RMC erosion rig, the upstream and downstream blades both rotated, so it was decided that increased axial distance between the blades was acceptable and provided as much of a realistic single stage compressor flow pattern as possible for the upstream blade.

The EDM process was found to leave an oxide layer on the surface of the material, which is not ideal for erosion-resistant coating adhesion [12]. Therefore, AF EDM, the blades were carefully polished to remove this layer. The blades were then silver-soldered to the dovetails using an oxyacetylene torch. When this was performed however, oxidation re-occurred on the blades and so they were bead blasted to remove this second residual layer. The tensile strength of the silver solder was verified experimentally at expected operating forces by Massouh [14] using TiN-coated steel coupons that were silver-soldered together. Figure 3 shows a bare steel blade assembly AF silver-soldering and bead blasting. The Canadian coin shown in Fig. 3 is 23.8 mm (0.937 in.) in diameter.

Surface finish was also an important consideration. Surface finish roughness is a measure of the deviation from a perfectly flat surface. The typical first stage compressor blade has a surface finish roughness of 0.254 μm (10 μ-in.) [12], and erosion-resistant coatings are on the order of 6–30 μm (236–1181 μ-in.) [2–4]. Therefore, the blade surface finish had to be close to 0.254 μm, but could be rougher depending on the thickness of the applied erosion-resistant coatings. As long as the coating thickness was relatively greater than the surface roughness of the blade material, the coating would smooth out most of the blade surface's peaks and valleys. However, maximizing the substrate material's smoothness was still important for coating layer smoothness and adhesion.

Surface roughness tests were conducted on the blades BE and AF erosion testing using a SJ-400 profilometer, which was capable of measuring curved surfaces accurately. Three locations were examined: the SS near the LE of the downstream blade, the SS near the TE of the upstream blade, and the pressure surface (PS) near the LE of the downstream blade. AF averaging the results, the average, root mean square (RMS), max peak height, and max valley depth values for the bare blades BE erosion were: 1.47 μm, 1.86 μm, 10.75 μm, and 5.43 μm, respectively. The average and RMS values of the EDM blades were 20% smaller than previous waterjet-cut blade prototypes, and the max peak was 9% smaller. For this work, an erosion-resistant coating thickness of 16 μm (630 μ-in.) was selected, since it performed best during ASTM testing of Immarigeon et al. [2]. This selected coating thickness was sufficient to overcome the substrate material surface roughness.

BE coating, the blade assemblies, and specially designed holders were cleaned in acetone and alcohol to remove any contaminants. An arc PVD method was used to deposit the 16-μm-thick TiN or CrAlTiN erosion-resistant coatings onto the blades. The blade assemblies were mounted on a two-axis turntable during coating. Further details are provided in Ref. [12].

Garnet UT220 abrasive from Barton, designed for use in waterjet cutting machines, was found to be suitable based on being safe to handle, particle size and shape, and capability for standardization. Specifically, the average particle size of Barton 220 garnet abrasives was 82 μm, which reasonably matched the desert sand particle size range reported in Davison et al. [15]. This report was based on studies of aircraft engine sand ingestion tests in the Arizona desert (conducted by Cowherd [16]) and Afghanistan sand ground samples. The selected garnet particles were irregular-shaped with distinct edges. The specific gravity of the garnet was 3.9–4.1, and the Mohs Hardness was 7.5–8.5 [12]. For comparison, the Mohs Hardness of diamonds and 17-4PH steel are 10 and approx. 5, respectively.

A rainbow test configuration was used, which involved an alternating bare, TiN-coated and CrAlTiN-coated blade assembly installation pattern around the rotor such that all of the 16 blade assemblies (6 bare, 5 TiN-coated, and 5 CrAlTiN-coated) were tested under identical conditions. Figure 4(a) shows the rainbow test pattern installed in the rotor housing with the upstream retaining ring and screws removed. Figure 4(b) shows a top-down perspective of three test blades side by side, and Fig. 4(c) shows the full rotor assembly with all the blades installed. In Fig. 4(c), the bearing shown in Fig. 4(a) is resting flat on the table. The smallest shaft connects the main rotor shaft to the gearbox.

The garnet and air were already premixed by the time the air reached the blades, and a cone shaped structure (not shown in Fig. 4) directed the path of the garnet/air mixture into a ring shaped annulus passage 12.7 mm (0.50 in.) wide. The cone was installed in an aluminum housing (see Fig. 1 for exterior view) and held centered by three streamlined attachment posts. Immediately downstream of the cone, the annulus passage became straight for a length of 18.3 mm (0.72 in.) to allow the flow to be oriented as close to perpendicular to the blade LE as possible. Lastly, a 6.6 mm (0.26 in.) gap remained BE the garnet-laden flow arrived at the LE of the rotating blades and the blade/housing tip-gap. This gap allowed for the rotor to spin freely, taking into account the thickness of the 16 screw-heads and required washer balance weights that secured the 1.60 mm (0.063 in.) thick retaining ring to the rotor (shown in Fig. 4(c)). Evidence of erosion at the root and along the span of each blade supports the assumption that the garnet particles followed the same path as the airflow.

Using Eq. (2) for normal (90 deg) impacts, the erosion rate was expected to be 4.46 times faster at $151 m/s$ than at $84 m/s$⁠. However, since the blade passages had to be taken into account, this reduced the amount of sand that would impact each blade. Once blade size, geometry and passage spacing had been considered, it was expected to take 0.69 h (41 min) for the 16 μm-thick TiN coating to erode completely on the LE. An equivalent coating surface area on the SS or PS of the blade (assuming oblique impacts), was expected to completely wear away AF 9.1 h.

A test duration of 5 h was chosen in order to observe the progress of coating erosion on the LE and possibly 55% of coating thickness loss on the PS and SS of the blades. If, however, secondary vortex scrubbing occurred, the erosion rate on the PS and SS of the blades could be higher in some areas.

Erosion testing was conducted at an average rotation speed of 10,860 rpm at an average axial air mass flow rate of $1.11 kg/s(2.44 lbm/s)$⁠, resulting in an average Ma_LE of 0.50 and an approximate local 0 deg AOA. Oil-flow visualization testing was undertaken prior to erosion testing to determine if the blades were experiencing design incidence conditions. Erosion testing was completed in 1-h increments, and the sand hopper was refilled AF each test increment. The rig performed consistently and reliably in terms of air mass flow rate and sand flow rate. The average deviation on the local AOA during erosion testing was −0.6 deg from zero. Performance parameters for each 1-h test increment are presented in Ref. [12]. The average sand concentration used for hours 1–4 of erosion testing was $2.5 g/(m3 of air)$⁠. For hour 5, the sand concentration was increased to $4.0 g/(m3 of air)$⁠, which is considered severely limited visibility or a brown-out concentration by Davison et al. [15]. The erosion rate for both the uncoated and coated blades at a sand concentration of $4.0 g/(m3 of air)$ was determined to be approx. twice the erosion rate as that measured at a sand concentration of $2.5 g/(m3 of air)$⁠. Therefore, results were presented for 6 equivalent hours of erosion at a sand concentration of $2.5 g/(m3 of air)$⁠.

The design Ma_LE for the V103 airfoil (NACA 6505 with rounded LEs and TEs) is 0.67 according to Hilgenfield and Pfitzner [18]. The RMC erosion rig was not able to achieve this value due to the backpressure present in the rig. However, an acceptable average Ma_LE of 0.50 was reached, and impact erosion rates could be scaled up to the design Mach number using Eq. (2). This would have resulted in an erosion rate 2.0 times higher at Mach 0.67 than at Mach 0.50 for an average p of 2.475 using Eq. (2) [12]. In terms of inlet axial Mach number, an average of Mach 0.34 was reached in the RMC erosion rig. Typical axial compressor face inlet Mach numbers for aircraft gas turbine engines range between 0.4 and 0.6, the highest of which is used for engines in supersonic applications [19]. However, turboprop engines routinely operate at an inlet Mach number range of 0.3–0.6 [19]. Therefore, the RMC erosion rig results are still applicable to aircraft with turboprop engines operating at the lower end of this Mach region.

The solidity (σ) of the installed blades (ratio of the chord to the pitch spacing between blades) was 0.7, which is lower than previous experiments in the literature, such as σ = 2.0 for Balan and Tabakoff [7] and σ = 1.0 for Ghenaiet et al. [8]. It was also lower than the average σ of 1.4 for axial compressors in aircraft gas turbine engines [20]. The σ was limited by the robust rotor design and dovetail widths used in the RMC erosion rig. Therefore, the low σ meant that the likelihood of ricocheting of sand particles from one test blade to the next would be minimal. However, the secondary flow erosion effects were postulated to be better discernible. While the ricocheting damage mode is not to be neglected, its effects are well described in Refs. [7] and [8]: particle reflection off the PS and subsequent impact on the rear portion of the SS, as well as particle reflection off the SS near the LE and subsequent impact on the PS near the blade tip. Less is currently known of the erosion damage caused by secondary flows. Overall, the RMC erosion rig's configuration still operationally represented conventional fans and compressors, where erosion in the tip and hub regions was very important, and aerodynamic efficiency was not.

The flow coefficient (⁠$φ$⁠) (the ratio of axial inlet velocity to rotational velocity) of the rig was 0.94. This was just outside the range of 0.3–0.9 for most axial compressors [9]. As the flow coefficient is decreased, the compressor stage efficiency increases, except at the very low end of the spectrum [9].

The diffusion factor (DF) for the upstream (compressor) blade was calculated to be 0.85 (The DF equation can be found in Refs. [9] and [12]). This was significantly greater than the range for most axial compressor designs, which is usually limited to less than 0.6 to prevent significant rises in total pressure loss and the generation of hub-corner stall on the blade's SS [21]. However, the RMC erosion rig used a double-bladed design to have the flow enter and exit the rotor at the same angle, in order to prevent an overall static pressure rise and, consequently, the requirement for a more powerful electric motor. This meant that the outlet angle (α₂) of the upstream blade was 0 deg. Therefore, this value was a factor in raising the DF value. For a typical axial compressor rotor blade stage, this outlet angle would be greater than 0 deg. Another contributing factor to the high DF value was that the σ of 0.7 was much lower than for normal axial compressors. A σ nearer to 1, 1.5, or 2 would have reduced the DF for the upstream blade to 0.69, 0.57, or 0.51, respectively.

Overall, the RMC erosion rig met the requirements to operate at an acceptable Ma_LE and $φ$⁠. Brown-out sand concentrations typically encountered during aircraft operations in desert regions were also achieved. Internal aerodynamic characteristic comparisons with previous rotating rigs such as Balan and Tabakoff [7] and Ghenaiet et al. [8] are presented in Ref. [12]. To summarize, the RMC erosion rig was determined to be nearly as realistic as Balan and Tabakoff's [7], based on a devised realism factor [12].

In order to obtain a macroscopic erosion quantification, the 16 blade assemblies (6 bare 17-4PH steel, 5 TiN-coated, and 5 CrAlTiN-coated) were weighed BE testing, AF every hour of testing and upon test completion. Each blade assembly was weighed using a Scientech SA 210 scale, accurate to ±0.1 mg. The coated blade assemblies were only weighed AF the blades had been coated. The upstream end of each dovetail was etched with a serial number, prior to being weighed, so that each one could be tracked. The reason for not separating the blades from the dovetail during weighing AF erosion testing completion was that the removal of the blade would certainly result in extra coating or solder loss, invalidating any measurements. This meant that the actual mass-loss of the upstream test blade on its own could not be measured precisely. However, the relative difference in mass-loss between each of the blade assemblies could.

AF the mass-loss data were obtained, percent mass-loss sustained by the upstream blade of each blade assembly was estimated. It was assumed that the upstream blade experienced more erosion than the downstream blade due to increased LE exposure to impact erosion. Visual examination of the coated blades AF testing led to an estimate of 70 ± 10% of the total blade assembly erosion applied to the upstream blade.

Two mass-based relations were devised based on the mass-loss data: % erosion rate based on time (%ERT) and erosion rate based on sand (ERS). %ERT is the percent compressor blade mass-loss per hour of testing at a constant sand concentration. ERS is the amount of compressor blade mass-loss per total mass of sand impacting one blade and passing through one blade passage (which is $1/16th$ of the total area, since 16 blade assemblies were mounted in the rotor). Relative comparisons were termed RERT and RERS, where the erosion of the bare 17-4PH steel blades was defined as the base value. Therefore, the RERT and RERS were unity (1.0) for the bare blades.

To determine blade dimension changes, the chord, LE thickness, TE thickness, and blade height were measured BE and AF 5 h of erosion testing using a digital vernier caliper. The chord, LE thickness, and TE thickness measurements were taken at three locations: blade hub, midspan, and blade tip. LE and TE thickness measurements were taken no further than 2 mm from the LE or TE (7% or 93% chord), respectively. Blade height measurements were taken at the LE, midchord, and TE. Average changes were then calculated for each test blade. These average changes were converted into percentages and an overall average change for the 6 bare 17-4PH steel blades was obtained. The same process was conducted for the 5 TiN-coated and the 5 CrAlTiN-coated blades. Once these overall averages were calculated, they were scaled for the 6 equivalent hours of erosion at a sand concentration of $2.5 g/(m3 of air)$ to obtain percent change per hour. Table 1 shows the resulting performance metrics and their respective descriptions. For the relative comparison metrics (RECRR, RTETRR, RLETIR, and RHRR), the erosion of the bare 17-4PH steel blades was defined as the base value, therefore they were all equal to unity for the bare blades.

Uncertainty values for equations of the form $f=xy/z$ were calculated using the partial differential method [22]. Uncertainties on averages were calculated using the standard uncertainty method, and uncertainties on sums were calculated using the summation in quadrature method (both from Ref. [23]).

Erosion patterns were observed using both the naked eye and an optical microscope. The various lighting parameters used to take each optical microscope photo were recorded so that BE and AF photos of the same location used the same lighting parameters. Photos were taken of the following parts of the upstream (test) blades: PS rear-half, PS front-half and LE, SS front-half, SS rear-half, and blade-tip profile. For the downstream blade, less magnified photos were taken of the entire PS and SS. Optical microscope photos and visual descriptions using the naked eye were taken AF every hour of testing. Representative blade photos not shown in this paper can be found in Ref. [12].

Prior to erosion testing, oil-flow visualization was conducted to confirm the intended 0 deg local AOA at the upstream blades' LEs, as well as the characteristics of the flow conditions and paths along the blades' surfaces. This involved painting the surfaces of both blades of one bare steel blade assembly with an oil paraffin–graphite solution, then running the rig for 10 min at normal operating conditions, excluding sand, with all blade assemblies installed. The dried oil was then left on the blades of that assembly when erosion testing was conducted, in order to provide clearer visual evidence of different areas of erosion.

In addition to the optical microscope pictures, SEM EDAX was performed on one TiN-coated blade and one CrAlTiN-coated blade BE and AF erosion testing. The EDAX used a maximum power beam of 20 keV on locations on the PS and SS of each blade. As part of the analysis, a spectrum was produced, which showed the composition of elements in the area being exposed. Prior to erosion testing, the TiN-coated blade spectrum detected the presence of Ti, N, and some trace amounts of carbon and oxygen. The CrAlTiN-coated blade spectrum detected the presence of Cr, Al, Ti, N, and some trace amounts of carbon and oxygen prior to erosion testing. These spectra confirmed the coating compositions, but also that the electron beam could not penetrate the 16 μm-thick coating, since no iron was detected. Therefore, AF erosion testing, if a sufficient coating thickness was eroded, the EDAX would begin to show the presence of iron in the spectrum (since the blades were made of 17-4PH steel). Unfortunately, the threshold coating thickness at which iron would begin to appear in the spectrum was unknown. Making several different coating thickness samples in an attempt to find this threshold was cost- and time-prohibitive. However, a relative comparison could be made, based on the different spectra's iron concentrations for different locations on a blade.

The LEs of the bare, TiN-coated, and CrAlTiN-coated blades all clearly sustained impact damage. Figure 5 shows photos of the LE of bare, TiN-coated, and CrAlTiN-coated blades BE and AF 5 h of erosion testing. The severity and nature of the damage differed for all three types. The LE of the bare 17-4PH steel blade became bowed and more blunt (Fig. 5(b)). This indicated that the velocity gradient of the airflow conformed to a duct flow, where the velocity is greater in the centre and decreases toward the walls. LE blunting was commonly observed in previous compressor blade profile erosion research [5–8].

For the TiN-coated blade, the LE became polished and slightly more blunt (Fig. 5(d)). This resulted in a greater LE thickness increase than for the bare blades, likely due to the protection provided by the remaining TiN coating on the SS and PS just downstream from the LE. For the bare blades, at the same time that the LE was becoming more blunt, the polishing of the front regions of the SS and PS eroded away some of the LE thickness. This made its overall increase in bluntness less apparent. The inward bow at midspan of the TiN-coated blades (Fig. 5(d)) was less apparent than that of the bare blades. This indicated that the TiN-coating on the LE provided some protection for the steel substrate material. The inward bow at the midspan also contributed to the greatest chord length reduction occurring at the blade midspan for both the bare and coated blades.

For the CrAlTiN-coated blade, the LE sustained an irregular erosion pattern, with an overall LE radius increase (Fig. 5(f)). This resulted in almost the same LE thickness increase as that encountered on the bare blades (according to quantitative results outlined later in this paper). Erosion was greater than that experienced by the TiN-coated blade, but less than that encountered by the bare steel blade. There was no noticeable bow at the midspan. This erosion pattern indicated that the CrAlTiN-coating on the LE provided some protection for the steel substrate material, but less than that provided by the TiN coating. Chromium and aluminum have lower hardness characteristics than titanium; therefore, the irregular erosion pattern could have been due to the use of three types of metals in the coating, the distribution of which could have led to lower adhesion capabilities in different regions of the LE. An irregular erosion pattern would almost certainly increase the performance losses compared to those for a uniformly eroded blade LE. The more the LE retains its design profile, the lower the performance losses. This is consistent with previous compressor blade profile erosion research [7,8,24,25].

The SS of the bare, TiN-coated and CrAlTiN-coated blades sustained erosion damage. However, the severity and nature of the damage differed for all three blade groups. Figures 6(a)–6(c) show the SS of a bare 17-4PH steel blade AF oil-flow visualization testing (no sand used), AF 1 h of erosion testing (annotated), and AF 5 h of erosion testing, respectively. Figure 7 shows cross section sketches of the proposed flow phenomena around the blade at the annotated sections identified in Fig. 6(b).

In Fig. 6(b), Point 1 identifies a polished surface up to the maximum camber point (approx. 50% chord) of the blade. This shows that the sand was likely entrained in the boundary layer and polished the surface at shallow angles. AF the midchord point, the flow separated from the surface (Point 2), formed a separation bubble, and reattached at Point $2'$⁠, as evidenced by less residual oil downstream of Point $2'$ in the blade midspan region. Point 3 corresponds to a darker area caused by scrubbing erosion from overtip vortices which travelled from the PS over the tip to the SS of the blade. These characteristics are even clearer in Fig. 6(c) AF 5 h of erosion testing.

Sections A-A, B-B, and C-C from Fig. 6(b) are sketched in Figs. 7(a)–7(c), respectively, to describe the flow characteristics at each section. Section A-A shows evidence of a separation bubble immediately downstream of Point 2, with a turbulent flow reattachment. Section B-B shows evidence of a separation bubble immediately downstream of Point 2, followed by a turbulent flow reattachment and a second separation point immediately prior to Point 4. The sand-laden recirculation at Point 4 likely contributed to increased scrubbing erosion, shown by the darker color at this location. Section C-C shows evidence of a separation bubble immediately downstream of Point 2, which merges with the hub-corner stall. This is shown by the large amount of residual oil along the Section C-C line from Point 2 to Point 5 in Fig. 6(c).

In compressor cascades, the transition point from laminar to turbulent flow on the SS often occurs by means of a laminar separation bubble [19]. The significance of the bubble is that free shear layers, such as that over the bubble, are very unstable and become turbulent at an earlier chordwise position than would have occurred for an attached boundary layer. Downstream of the bubble, the flow in the shear layer becomes turbulent and reattaches [9,26]. Blade LE geometry changes, which resulted during erosion testing, can cause separation bubbles on the SS due to a larger acceleration around the LE, followed by a localized deceleration [25]. Based upon prior linear cascade research, the transition Reynolds number based on the chord (Re_c) is typically 250,000 [27]. Therefore, a transition from a laminar to a turbulent boundary layer likely occurred on the SS since the Re_c for the blades was approx. 260,000 (with Re = 250,000 at 97% chord). However, the upstream shift in the transition zone from 97% chord to Point $2'$ in Fig. 6(b), approx. 60–70% chord, is quite likely due to the increased flow turbulence in a rotating rig versus a linear cascade. The centrifugal blower could also have introduced additional turbulence in the flow.

Since DF was 0.85 for the upstream blades, much larger than the threshold value of 0.4 for hub-corner stall [21], such a stall likely occurred, shown by the presence of a large amount of residual oil at Point 5 in Fig. 6(b). Lei et al. [28] determined that for compressor blade cascades with low σ, a recirculation zone developed near the TE of the SS. Evidence of such a zone is shown in Ref. [28] for σ = 1 and a Re_c of 250,000, comparable to that used for this research (260,000). Since the RMC erosion rig had a blade σ = 0.7, it can be assumed that the recirculation zone similar to that shown in Ref. [21] was present. This explains the dark region at Point 4 in Fig. 6(b), which was likely caused by recirculating flow scrubbing due to a second flow separation point (Section B-B in Fig. 7) immediately upstream of the Point 4 darkened area. The overtip vortices at Point 3 and the hub-corner stall at Point 5 restricted the recirculated flow scrubbing to the midspan region of the SS near the TE (as seen in Figs. 6(b) and 6(c)). This restriction pattern was also present in Ref. [28]. For compressors with a higher σ near 1.5, resulting in a lower DF, the recirculating flow at the rear midspan region of the SS would not likely be present, as found by Lei et al. [28]. Figure 8 shows the forward and rear SS of bare and coated blades BE and AF 5 h of erosion testing.

The effects of the recirculating sand-laden flow scrubbing on the SS and the polishing on the PS combined to reduce the blades' TE thickness. The TE thickness reduction was more apparent on the bare 17-4PH steel blades than for either of the two types of coated blades. The initial tip-gap for all blades was 0.51–0.76 mm (0.02–0.03 in.), which was 1.9–2.9% of the initial blade chord (4–6% of initial blade height). This was controlled by using BONDO autobody filler material in the machined rotor housing groove. AF testing, the BONDO filler appeared to have maintained its initial thickness. This tip-gap range was similar to that studied by Tang et al. [29]. Therefore, overtip vortices traveling from the PS through the tip-gap to the SS were expected. These sand-laden vortices would cause scrubbing erosion near the TE in the tip region of the SS. The reduction in the TE thickness is evident in Fig. 9, which shows photos of the rear portion of the bare, TiN-coated and CrAlTiN-coated blade tips BE and AF 5 h of erosion testing.

For the TiN-coated blade, the TE thickness was substantially reduced (Fig. 9(d)), but not to the extent as for the bare steel blade (Fig. 9(b)). Furthermore, the tip on the PS between Points 1 and 2 in Fig. 9(d) was rounded AF erosion, denoting the presence of overtip flow erosion in this region. For the CrAlTiN-coated blade, the TE thickness had been reduced slightly (Fig. 9(f)), but not to the same degree as that of the bare steel blade nor the TiN-coated blade. For both the bare and coated blades, the rest of the blade thickness, except for the LE, remained effectively constant. The full overtip surface had a more polished appearance for all three, which indicates erosion by overtip flow.

For the coated and uncoated blades, the height was reduced the most at the TE, proposed to be due to the location of the overtip vortex in that region. However, it appears that the overtip vortex contributed more to reduction of the TE thickness than in reduction of the blade height, shown by the difference between the %TETRR and %HRR values in Table 2.

Raw mass-loss data was obtained for each blade assembly, each of which consisted of the upstream (test) blade, downstream blade and dovetail. Figure 10 shows the unprocessed blade assembly cumulative mass-loss in grams during erosion testing. Each line on the graph was generated from the average mass-loss values for the 6 bare blades, 5 TiN-coated blades, and 5 CrAlTiN-coated blades, respectively. Distinct differences are evident between the coated and uncoated blades, as well as between the two types of coated blades. The erosion rate was approx. linear, which demonstrated the repeatability of the results. The change in slope for hour 5 was due to the increase in sand concentration to the brown-out condition of $4.0 g/(m3 of air)$ recommended in Ref. [15].

Processed mass-loss data was based on the upstream blade only, which was estimated to have sustained 70 ± 10% of the total erosion. This estimation was made based on visual observation of the blade erosion patterns using both the naked eye and the optical microscope photos. In order to have a reference point to the mass of the upstream blade on its own, one spare uncoated blade was weighed and the surface area was calculated using the blade's Solidworks computer-aided design file. The spare uncoated blade weighed 3.0861 g, and the surface area covered by the coating was 7.419 cm² (1.15 in.²). Initial upstream blade weights for the coated blades were determined by adding a calculated coating mass to the nominal bare blade mass using the following information: coating thickness of 16 μm, assuming the same surface area, TiN coating density of $5.22 g/cm3$ and CrAlTiN coating density of $5.42 g/cm3$⁠. The CrAlTiN coating density was estimated based on its chemical composition as measured using energy dispersive X-ray spectroscopy at NRC [12]. Based on these estimates, the initial TiN and CrAlTiN coating masses were 0.062 g and 0.064 g, respectively, which represented approx. 2.0% of the initial coated blade mass.

Table 2 shows results for the overall quantitative performance metrics (previously identified in Table 1) for both the uncoated and coated blades. Table 3 shows results for the four geometry-based metrics per blade region for both the uncoated and coated blades. Uncertainty values were low since there were 6 equivalent hours of testing and 5 or more blades per hour for which mass-loss and dimension change data were obtained.

Table 2 shows that there was a significant difference in the percent mass-loss per hour between the bare blades and the two types of coated blades. This takes into account the estimate that the upstream blade sustained 70% of the total erosion. AF 6 equivalent hours at a sand concentration of $2.5 g/(m3 of air)$⁠, the TiN-coated upstream blades had lost 2.3% of their initial blade mass (0.072 g) on average, which was greater than the initial calculated mass of the TiN coating. It might seem that the entire coating was worn off. However, inspection of Figs. 5 and 8 shows that this was not the case. Therefore, some of the substrate steel must have been worn away at the LE and possibly the TE on the PS and SS. For the CrAlTiN-coated blades AF 6 equivalent hours at a sand concentration of $2.5 g/(m3 of air)$⁠, the upstream blade had lost 1.2% of its blade mass (0.039 g) on average, which was less than the initial calculated mass of the CrAlTiN coating. Therefore, it would seem that only a portion of the coating was worn off, substantiated by Figs. 5 and 8. Some of the substrate steel was also likely worn away at the LE and possibly near the TE on the PS and SS.

Based on the RERT metric, both the TiN-coated and CrAlTiN-coated blades, respectively, performed 82% and 90% better per hour of sand exposure than the uncoated 17-4PH steel blades. Assuming the worst case based on the uncertainty values, the coated blades were still, respectively, 57% and 67% better. This reconfirmed the capabilities of erosion-resistant coatings in minimizing blade erosion. Furthermore, this significant improvement was achieved with a 16 μm coating thickness, which only added 2.0% to the overall blade mass. The RERS and RERT results in Table 2 are very similar; however, the RERS allows for a better comparison to previous cascade and rotating rig research such as Refs. [5–8].

As shown in Table 3, the CrAlTiN-coated blades had a 44% lower midspan chord reduction compared to the bare blades, whereas the TiN-coated blades sustained very similar chord length reductions to that of the bare blades. This is likely due to the fact that most of the TiN coating had been worn off the LE AF 2 h of erosion. If the same chord reduction failure criteria of 1.5–2% used in Ref. [5] was applied, the midspan chord of the bare blades and TiN-coated blades would have reached the failure criteria within the 6 equivalent hours of erosion testing.

The reduction in TE thickness was the most significant blade dimension change. In Table 2, the bare blades' average TE thickness reduced by 6% per hour, whereas the coated blades provided substantially lower erosion rates. The significant TE thickness reductions were likely due to the combination of recirculation flow scrubbing and overtip vortex scrubbing erosion on the SS, and polishing erosion on the PS.

AF 6 equivalent hours of testing, the tip-gap increased by 0.42% of the initial blade height (or 0.40% of the initial blade chord length). Tang et al. [29] showed that small percentage increases in the tip-gap can strengthen overtip vortices. Table 3 shows that the most significant blade height reduction occurred near the TE for both the bare and coated blades. This is proposed to be due to the location of the overtip vortex in that region.

Initial average surface roughness (R_avg) values for the bare, TiN-coated and CrAlTiN-coated upstream blades were: 1.17 μm, 1.34 μm, and 1.64 μm, respectively. These R_avg values were less than 2 μm, which was considered the minimum threshold value for rough compressor blades in Ref. [11]. AF erosion testing, R_avg decreased on average by 13.7% for the bare blades, 8.3% for the TiN-coated blades and 21.7% for the CrAlTiN-coated blades. The decrease in R_avg near the TE of the SS was consistent with the causes of TE thickness reduction: vortex scrubbing on the SS and low angle impingement polishing on the PS. Scrubbing vortices also have a polishing effect, which would lead to a decrease in roughness. The largest R_avg reduction for both the bare and coated blades was the hub-corner stall region (26% average reduction overall and specifically a 41% reduction for the CrAlTiN-coated blades). The largest final R_avg value was 1.46 μm (a 14% decrease), which was measured in the recirculation scrubbing zone on the SS of the CrAlTiN-coated blade. Results from Refs. [10] and [11] showed increases in R_avg, while in this work, the opposite was determined. It is likely that the vortices were having a polishing effect and had not scrubbed the blade surface long enough to increase the surface roughness.

AF erosion testing, three areas of a TiN-coated and CrAlTiN-coated blade were analyzed with the RMC SEM using EDAX: the recirculation zone scrubbed area on the rear section of the SS (approx. 90% chord), and locations at approx. 25% chord on the SS and 75% chord on the PS. For the TiN-coated blade, a maximum of 2.8% of the spectra elements were iron at the first area. This appeared to show that scrubbing due to recirculation vortices eroded slightly more of the coating than erosion due to polishing. Figure 11 is an EDAX spectra for a TiN-coated blade, with the peaks being elements detected during the EDAX scan. The larger the peak, the higher the percentage of that element as compared to the total number of atoms analyzed. The K or L AF each element name gives information about the atoms' electron arrangement, which was not used in this analysis.

For the CrAlTiN-coated blade, a maximum of 1.4% of the analyzed spectra elements were iron at the 75% chord point on the PS. This showed an opposite trend to the TiN-coated blades.

Therefore, using EDAX, comparisons between the two types of coated blades and the erosive capabilities of vortices versus shallow angle impact polishing on each coated blade were inconclusive. However, the detection of iron AF erosion testing, as shown in Fig. 11(b), supported other evidence that the coatings on these areas of the blades were partially eroded.

The LAZ scores for the bare, TiN-coated and CrAlTiN-coated blades are presented in Table 4.

When uncertainty ranges are taken into account, the TiN-coated blades performed at least 14% better than the bare blades, and the CrAlTiN-coated blades performed at least 44% better than the bare blades. When compared to the TiN-coated blades, the CrAlTiN-coated ones performed at least 14% better. It is clear that a distinct durability performance difference was apparent in the test results. For the gas turbine engine industry where performance increases of even fractions of 1% are considered significant, these results underscore the merits of erosion-resistant coatings.

Two gas turbine erosion-resistant coatings were tested for performance in a novel rig designed and built at RMC. They were tested in a rainbow configuration against an uncoated 17-4PH steel blade baseline under identical operating conditions. Results were obtained periodically over 5 h of testing using visual inspection, measurement of blade assembly mass and geometry, blade surface roughness measurements, and SEM EDAX.

Wear patterns on the test blades included: impact erosion on the LE, polishing erosion on the PS and forward half of the SS, and scrubbing erosion on the SS near the TE. Both coated blades eroded less than the bare 17-4PH steel blades, most noticeably at the LE and TE. The TE thickness decreased while the LE thickness increased due to blunting. Of all the dimensional changes, the TE thickness reduction, on the order of 6% per hour for the bare blades, was the most prominent and likely the major source of blade mass-loss. This was almost certainly due to recirculation flow scrubbing and overtip vortex scrubbing erosion on the SS, and polishing erosion on the PS. Major chord reduction was at midspan while major height reduction was at the TE. The CrAlTiN-coated blades eroded less than the TiN-coated blades in all instances. However, the CrAlTiN-coated LE eroded in an irregular shaped pattern, whereas the TiN-coated LE became blunter but more uniformly smooth. Overall, secondary flows were determined to be a major factor in blade erosion by sand-laden air. SEM EDAX aided in detecting a reduction in erosion-resistant coating thickness in areas where the coating was not seen in photos to be fully eroded.

The CrAlTiN-coated blades' durability was superior in terms of lower mass-loss and the ability to better maintain its original blade shape, and hence, performance. This was reflected in its LAZ-16 score of 0.41 (a lower score denotes better performance), whereas the TiN-coated blades' LAZ-16 score was 0.69. The bare 17-4PH steel blades had a baseline score of 1.00. It is recommended to use ASTM standardized testing for initial coating development, and then for those coatings that perform well, assess their durability in the more turbomachinery-representative conditions of the RMC turbomachinery erosion rig.

Financial support from the Canadian Department of National Defence's Aerospace Engineering Research Advisory Committee is acknowledged, and specifically the motivation of Major D. Little. Technical support provided by Mr. R. McKellar at NRC was essential. At RMC, Dr. J. Snelgrove operated the SEM and provided the EDAX spectra, Ms. R. Massouh began and completed the design of several of the rig's components, and Dr. M. LaViolette and Dr. A. Asghar provided design and analysis advice. RMC mechanical technologists, most particularly Mr. M. Gatien, were indispensable to the success of this program.