Particle erosion of gas turbine thermal barrier coating

In the experimental study, tests were conducted in the erosion wind tunnel facility at University of Cincinnati for TBC coated and uncoated blade materials to determine the erosion rates

Particle Erosion of Gas Turbine Thermal Barrier Coating

A thesis submitted to the University of Cincinnati

In partial fulfillment of the requirements for the degree of

Committee Chair: Dr Awatef Hamed

Dr Widen Tabakoff

Dr Robert A Miller

Abstract

The purpose of this research is to examine and understand the complex phenomenon associated with the particle impacts on turbine blades and the associated erosion of Thermal Barrier Coated (TBC) turbine vane and blade surfaces by ingested solid particle impacts Both experimental and computational techniques were used to find out the parameters relevant to rebound characteristics of particles and erosion rate of TBC coatings

In the experimental study, tests were conducted in the erosion wind tunnel facility at University of Cincinnati for TBC coated and uncoated blade materials to determine the erosion rates and particle restitution characteristics under different impact conditions Particle Image Displacement Velocimetry (PIDV) technique was used to determine particle rebound characteristics for different impact conditions From the experimental results, empirical erosion rate models and restitution coefficient models for alumina particles impacting on TBC coated blade surface are developed using non-linear regression analysis technique to predict the erosion rate and restitution coefficients for various impact conditions

In the computational analysis, numerical simulations were conducted for the three-dimensional flow field and particle trajectories through a high pressure single stage

axial gas turbine The solution to the Reynolds Averaged Navier Stokes (RANS) equations for turbulent compressible flow were obtained numerically using ANSYS CFX solver for unsteady N-S equations in their conservation form In gas turbine applications generally the particle loadings that are encountered are sufficiently low hence a one-way gas-particle interaction model was used to simulate the particle dynamics involved This does not take into consideration the effects of dispersed particles’ momentum exchange with the gas flow field

The experimentally based particle surface restitution models were incorporated in the simulations to determine particle rebound conditions after each surface impact The computed particle surface impact statistics were combined with experimentally based erosion models to predict the stator vanes and rotor blades coated surface erosion pattern and intensity

The experimental results reveal that the erosion rate increases with increase in impingement angle, impact velocity, and temperature The trajectories are determined for

26 μm and 500 μm alumina particles The simulation results show that the particle velocity in the stator is reduced by the surface impacts, which causes the particles to enter the rotor with negative incidence compared to the flow The rotor impacts reduce the particle velocities in the rotating frame, but could increase their absolute velocity It was

after rebounding from the rotor leading edge The simulation results predicted the intensity and pattern of TBC erosion over the stator and rotor blade surfaces and the variation in the overall blade surfaces erosion with ingestion velocity

Acknowledgements

I would like to express my gratitude to all those who helped me in completing this thesis This thesis would not have been possible without the help and support of a great number of people

The individual to whom I am most indebted is my advisor and mentor, Dr Awatef Hamed, for her constant support, guidance and encouragement during the course of my

MS

In particular, I would like to express my most sincere gratitude to Dr Widen Tabakoff for his important contribution to this work I deeply appreciate his expertise, knowledge, valuable suggestions and support for the completion of my thesis

I would also like to thank Dr Robert A Miller from NASA GRC for supporting

my research and for obliging to serve in my thesis committee I wish to thank Rob Ogden and Russ DiMicco for their technical expertise and support in the lab I am grateful to all

of my friends and fellow graduate students in Cincinnati for their help during my study

Lastly, I would like to give my love and thanks to my parents and my sister for their sacrifice and encouragement during this journey I dedicate this work to them

To my family,

Table of Contents

List of Tables ……… viii

List of Figures ……… ix

1 Introduction 1.1 Literature Review ……….…… …1

1.2 Motivation for research ……… 4

1.3 Need for Experimental study……… 5

1.4 Need for Computational study……….6

1.5 Objective of the present work……… 7

1.6 Present Research……… 7

2 Experimental Setup 2.1 Air Supply System……… 9

2.2 Particulate flow wind tunnel……… 10

2.3 Test Section……… 11

2.4 High Speed Photography……… 12

3 Measurement Methodology

3.1 TBC Erosion Rate Measurements……… 14

3.2 Data reduction from High Speed photography ……….15

3.3 Statistical Analysis of High Speed Photography Data……… 17

4 Computational Analysis 4.1 Turbine Geometry and modeling ……… 21

4.2 Grid generation ……… ……… 21

4.3 Governing Equations……… ……… 22

4.4 Computational Details……… ………24

4.5 Particle Trajectory Simulations……… ……… 25

5 Experimental Results and Discussion 5.1 Erosion test Results from TBC coated samples ………26

5.2 Results from High Speed Photography ……… 28

6 Empirical Models

6.1 Empirical Model for TBC erosion rate……… 29

6.2 Empirical Model for particle rebound characteristics………30

7 Computational Results and Discussion 7.1 Particle trajectory simulation results……… 31

7.2 TBC coated turbine blade erosion results……… 32

7.3 Parametric Study……… 33

8 Conclusions……… …35

Bibliography ……… … 38

Tables……… ……… 43

Figures ……… ……… 46

List of Tables 1 Particle Type and Particle Size……… 43

2 Error estimates of restitution ratios based on t-distribution method……… 43

3 Compressor Drive HP Turbine geometry……… 44

4 Turbine operating conditions……… 44

5 Erosion Test Conditions……… 45

6 Rebound Test Conditions……… 45

List of Figures

1 Particulate Flow Wind Tunnel……… ………46

2 High Temperature Erosion Wind Tunnel……… …47

3a Test Section in the Particulate flow Wind Tunnel……….48

3b Schematic of Holder with the target sample ……….48

4 High Speed Camera……… 49

5 High Speed Photography……… 50

6 Schematic of Sample Particle Trajectory……… 51

7 Solid Model of the Turbine blade……… 52

8 Computational Grid……… 52

9 Measured erosion rates of coated and uncoated samples……….… 53

10 Nominally 26 µm alumina particles……… 53

11 TBC Erosion Test Results (10 mils coating, T = 1,800o F, Qp = 5 gm) …………54

12 TBC Erosion Test Results (10 mils coating, T = 1,600o F, Qp = 5gm) …………54

13 Comparison of Erosion Test Results for 5 & 10 mil coated samples………55

14 5 mil TBC Erosion Life Test Results (T = 2,000 oF, Qp = 5 gm) ……….55

15 Variation of Velocity coefficient of restitution (ev) with impact angle β1……….56

16 Variation of Direction coefficient (eβ) with Impact angle β1……….…56

17 Pressure Contours at blade mid-span……… 57

18 Mach Contours at blade mid-span……… 57

20 Particle Trajectory for 500 μm alumina Particles………….……… 59

21 Schematic of particle size affect on entrance velocity relative to the rotor…… 60

22 Erosion pattern for 26 μm alumina particle impacts……….…….61

23 Erosion pattern for 500 μm alumina particle impacts……….… 62

24 Erosion Rate variation with injection velocity……… 63

1 Introduction

1.1 Literature Review

One of the most challenging problems in gas turbine engines operating in dusty environments is turbomachinery erosion1,2,3 The presence of solid particles in the working medium has an adverse effect on the performance and lifetime of turbomachinery As a result of the erosion caused by the solid particles not only aerodynamic characteristics of the compressor and turbine blade changes but can also lead to structural failure The mechanisms associated with particle ingestion are (a) the vortex from engine inlet-to-ground during high power setting with the aircraft standing idle or moving on the runway; (b) storms transporting sand to several thousand feet altitude; (c) thrust reverser afflux at low airplane speed blowing sand, ice and other particles into the engine inlets Combustion process may also produce erosive particles, from the burning of different types of heavy oils or synthetic fuels In particular, helicopter engines are highly susceptible to large amounts of dust and sand ingestion during hover, takeoff and landing

Without taxing the performance of gas turbine engines, it is extremely difficult to remove all solid particles from the gas stream4,5 Even small particles of one to thirty

burning turbines6 Surface erosion by particle impacts increases tip clearances and blade surface roughness and changes the blade shape especially near the leading and trailing edges7 In particular, TBC erosion is extremely damaging to the thermal protection of gas turbines, and has been identified as a life limiting factor for gas turbines8,9,10 With the increasing use of thermal barrier technology to protect highly loaded and rotating turbine components11 further studies are required to enhance thermal protection for life extension

Multiple process combinations such as cutting, fatigue, brittle fracture, and melting mechanisms have been identified to govern material removal by erosion Experimental studies of blade alloys and coating materials have concluded that their erosion rates are dependent upon particle impact velocities and impingement angles, as well as temperature for a given particle target material combination3 These tests require special facilities that control particle-laden flow around the target to achieve the desired impact conditions over the tested targets12 Prior TBC erosion test results13,14 have demonstrated that electron beam-physical vapor deposited (EB-PVD) TBC erosion rates are an order of magnitude less than plasma sprayed (PS) TBC at 90o particle impingement angle and that both increase with particle’s impact velocity when impacted normally It is also necessary to perform experimental studies of particle surface impact

to provide particle rebound characteristics over the range of impinging conditions that are

encountered in turbomachines3,7 However, very little is available in the open literature for TBC

Erosion of blades in turbomachines is mainly affected by rotational speed and flow conditions, blade geometry and blade row location, blade material, and particle characteristics3,7 The gas and particles experience different degrees of turning through the blade passages under two-phase flow conditions Deviation from the gas flow path increases with particle inertia causing multiple impacts with the various surfaces The direction and velocity of the particles as well as their distribution through subsequent stages are altered by surface impacts Experimental studies that simulate erosive particle impact conditions in the engine environment are instrumental to the development of blade and coating materials The associated blade surfaces erosion requires knowledge of the particle three-dimensional trajectories and their impact statistics on the various engine surfaces in addition to surface material erosion behavior under the impact conditions

Trajectory simulations of the particles are based on the numerical integration of the equations of motion of the particles, through the successive turbomachinery stationary and rotating blade rows The simulations require the flow field and blade passage as input, and a model for particle restitution conditions following each surface impact The basis for particle trajectory simulations in turbomachines continues to be Eulerian-

dimensional flow field solutions of the Reynolds Averaged Navier-Stokes equations for

turbulent flow through the blade passages16 are used in turbomachinery trajectory simulations

Hamed & Tabakoff18 developed a methodology to predict turbomachinery blade surface erosion using blade surface statistical impact data computed from particle trajectory simulations, and correlations based on blade material erosion test results It was used in the prediction of blade erosion in both axial and radial compressors and turbines3and erosion of turbine blade coating developed for automotive19 and ground based17 gas turbine applications

1.2 Motivation for Research

Currently, most of the gas turbines operate at very high temperatures and hence for thermal protection the base material of the turbines is coated with special coatings called Thermal Barrier Coatings (TBC) So once these coatings get severely eroded the turbine blade has to be replaced Hence erosion rate of these TBC coatings can essentially determine the life time of the turbine blades

Therefore, a detailed experimental and computational effort is required to understand the mechanism of erosion of these coatings and the factors on which they depend This has been the main motivation behind this research In the current research erosion rate and particle rebound characteristics of alumina particles when impacted on a TBC developed by NASA GRC has been studied

Moreover, very little research has been done to study the erosion pattern and intensity in very small turbines operating at high pressures and high speed So, to understand the erosion mechanism in these kinds of turbines combined experimental and computational techniques are required

1.3 Need for Experimental Study

Experimental studies are required to characterize blade and coating material erosion and particle rebound characteristics with different impact conditions For each kind of impacting particle and target material combination the erosion rate varies with the impact conditions and is essentially different from other pairs Also, each impacting particle and target material combination has different rebound characteristics compared to other pairs Subsequently, in order to develop empirical models to predict erosion rate

and particle rebound characteristics, to determine particle trajectories a vast database of experimental data is required for the particular particle and target material pair

Therefore, in the present study experiments are conducted to obtain erosion rate and rebound characteristics data for different impact conditions, for the particular TBC coated sample and alumina particles

1.4 Need for Computational Study

Conducting experiments to determine the erosion pattern and intensity of full scale turbines will not only be very expensive and time consuming but also very few facilities are available for that kind of research Thus, numerical simulations of particle trajectories model the effects of aerodynamic forces arising due to the interaction of the particles with the three-dimensional (3D) turbine flow field and the change in the particle velocities due to their collision with the vane and blade surfaces Vane and blade surface impact patterns using trajectory simulations are obtained at the operating conditions associated with particle ingestion into the turbine Predictions of vane and blade surface erosion patterns are based on the computed particle impact statistics and the experimentally measured data

1.5 Objective of the present Research

Objective of the present research is to characterize TBC erosion rate for different impact conditions, to accurately predict particle rebound characteristics at high speeds using high speed photography and to study the effect of alumina particle size and ingestion speed on erosion pattern and intensity of turbine blades of a HP compressor drive axial turbine

1.6 Present Research

In the present research experimental studies were conducted at the high temperature erosion wind tunnel facility at University of Cincinnati to study the erosion rate of TBC coated and uncoated turbine blade material samples for different impact conditions

Particle Image Displacement Velocimetry (PIDV) technique was used to determine the rebound characteristics of alumina particles after impacting TBC coated samples These tests were conducted at the particulate flow wind tunnel facility at University of Cincinnati

Empirical models were developed for TBC erosion rate and alumina particle restitution coefficients for different impact conditions These, models were incorporated

in ANSYS CFX solver to accurately determine particle trajectories and erosion pattern on turbine blades of a compressor drive high pressure (HP) axial turbine Numerical simulations for flow field and particle trajectory through turbine cascade are performed for a scaled NASA designed compressor drive HP turbine A parametric study was carried out to understand the dependence of erosion rate of stator and rotor blades on particle injection velocity

2 Experimental Set-up

The experimental setup consists of the following equipments:

(i) Air Supply System

(ii) Particulate Flow Wind Tunnel Facility

(iii) High Speed Camera

(iv) Data Acquisition System

2.1 Air Supply System

Two different air supply systems were used for conducting the experiments One for supplying air to the particle rebound characteristics measurement wind tunnel and the other for high temperature erosion wind tunnel But, both the air supply systems are essentially the same

Air was supplied from a high pressure tank through a 4” (101.6 mm) diameter pipe For this pipe line, the maximum pressure loading was 60 psig (5.1 x 105 N/m2) The pressure level from the pressurized air supply was adjusted using a pressure regulator The maximum inlet pressure can be set to 60 psig The mass flow rate was varied using a

flow regulating valve The air flow from the 4” pipe was fed to the particulate flow wind tunnel

2.2 Particulate flow Wind Tunnel

Fig 1 shows the schematic diagram of the particulate flow wind tunnel that was used to conduct the particle rebound tests at ambient conditions Whereas, Fig 2 shows the high temperature erosion wind tunnel facility that was used to conduct experiments to measure erosion rate of TBC coated samples For both these tunnels, the primary variables like fluid velocity, particle velocity, impact angle, particle flow rate and particle size could be easily varied and controlled using representative aerodynamic environment

In addition, for the high temperature erosion wind tunnel the temperature can be varied from ambient conditions to as high 2000 oF

The test section in the particulate flow wind tunnel as shown in Fig 3a is provided with a glass window in order to facilitate measurement of particle velocities using High Speed Photography Tabakoff and Wakeman12 described the operation and calibration procedures for the particulate flow wind tunnel and the high temperature erosion wind tunnel Both the tunnels essentially operate in the same way except for that the High Temperature Erosion Wind Tunnel has a combustion chamber and a pre heater

to heat up the flow and the particles to the desired temperature The particle feeder, part

A, shown in Fig 1 & Fig 2 is designed to operate at high pressures and has a replaceable metering orifice with an opening for controlling the particle loading

The high temperature erosion wind tunnel shown in Fig 2 works as follows: abrasive particles of a given constituency and measured weight are placed into particle feeder (A) The particles are fed into a secondary air source and blown into the particle pre-heater (D), and then to the injector (E), where they mix with the primary air supply (B), which is heated by the combustor (C) The particles are then accelerated via high velocity air in a constant-area steam-cooled duct (F) and impact the specimen in the test section (G) The particulate flow is then mixed with coolant and directed to the exhaust tank

Nominally 26 μm (500 grit) alumina particles were used for conducting the erosion tests and nominally 500 μm (36 grit) particles were used for the particle rebound tests Table 1 contains the particle type and sizes that were used for conducting the experiments

2.3 Test Section

The test section (figure 3a) is designed such that the particle laden air is directed over the sample placed in the holder (figure 3b) while preserving the aerodynamics of the fluid surrounding the blade sample This section contains several interchangeable inserts such that the fluid profile can be determined using conventional instrumentation In addition, the test section in Fig 1 has a window to facilitate recording of particle trajectories using high speed photographic or laser methods

2.4 High Speed Photography

A Phantom high-speed digital camera (model V 9.1) with a Nikon lens shown in Fig 4 was used to photograph the particulate flow The camera has a maximum frame speed of 50,000 fps For the current study a frame rate of 27,000 frames per second (37

μs interval) and resolution of 256 x 256 was found to be appropriate for the light intensity obtained from a 250 mW (wavelength 454-676 nm) laser source manufactured by Spectra-Physics Laser The laser sheet was about 1.65 mm thick and around 50 mm in height in the frame of interest The exposure time could be set as low as 3 μs

The camera was connected to a computer and the frame speed and resolution were set from the GUI of Phantom (camera) software The same software was also used to actuate the camera Details about the camera and the software can be obtained from the user manual20

The camera was focused to a reference point where the laser sheet was formed before recording the images It was ensured that the camera was properly aligned

by using a laser leveler The camera recorded images for a maximum time interval of 1 s

3 Measurement Methodology

3.1 TBC Erosion Rate Measurement

The mass erosion rate is defined as the change in mass of sample material per unit mass of impacting particles

)/(mg g particles

impacting of

mass

sample of

mass in

change rate

g cm particles

impacting of

mass

sample of

volume in

change rate

temperature erosion wind tunnel shown schematically in Fig 1 and subjected to erosion

by a calibrated mass of particles The holder protected all but one target coupon surface that was exposed to particle impacts The samples were weighed before and after the erosion tests to determine the weight loss due to erosion by the impacting particles

3.2 Data Reduction from High Speed Photography

The actual flow field is 3-D but only a two dimensional study was done in this research Since the laser sheet was in the x – z plane, so by processing the data only x – z components of the particle velocity can be obtained As the thickness of the laser sheet is very thin and only the particles lying in this thin sheet are properly illuminated the y component is not very significant Hence a 2 – D analysis like this should be sufficient for the current application

The camera was directly connected to a PC with Windows OS and the video was recorded using Phantom software After, the images were recorded they were analyzed

using Phantom Cine Viewer 663 software The images were imported to Cine Viewer software in cine format Using the Cine Viewer software successive frames were

analyzed and in each frame the location of the particles was found out from their pixel

number From this, the distance a particle moved between successive frames could be found out

The x and y co-ordinates of a particle in each frame was determined using

Phantom Cine Viewer image processing software By comparing successive images, the

trajectory of the particle is calculated using Particle Image Displacement Velocimetry (PIDV) technique21, 22, 23 The distance between the different images of each particle represents the information pertinent to the velocity of the particle, the time between each exposure being known

The 1” width sample was used for the reference length to normalize the distance the particle traveled between successive frames This gave the actual the distance the particle traveled between successive frames Since, the number of frames per second is known, the time between two successive frames could be found out Once the location of the particle between two successive frames (Figures 5 & 6) and the time between these two frames is identified the velocity of the particle could be calculated The velocity was deduced using the following relationship:

t

S S

S =

δ

t

δ = time between successive frames, sec

F = number of frames per second

V = velocity of the particle, m/s

S

δ = actual distance traveled by the particle between two consecutive frames, m

R = actual length of the reference mark in the test section, m

d = distance traveled by the particle as obtained from the pixel count

r = length of the reference mark as obtained from the pixel count

The velocity restitution coefficients and directional restitution coefficients were calculated by comparing three successive frames of interest

3.3 Statistical Analysis of High Speed Photography Data

Various error sources related to high-speed photography are outlined below

• Uneven size and shape of the particles

• Low particle concentration in the interrogation window

• Reduction of usable image pairs due to significant velocity component normal to the light sheet

• Reflection of light from the model surface

• Camera not very accurately focused

• Noise of video camera

In the present study, since the sample size is less than 30 (varied from 10 to 20), t – distribution method was used for error analysis instead of standard normal distribution

The parameters involved in t-distribution method are as follows:

ii Asymptotic to the horizontal axis

iii Symmetric about the mean

iv Dependent of the degrees of freedom If the sample size is n then the number of degrees of freedom is ν = n-1

v More variable than standard normal distribution ν(t) = ν(ν-2) for n > 2

vi Approximately equals the standard normal if ν is large

1

/)

2 2

s , y can be replaced by ev and eβ

μ = population mean of the restitution ratios

n = sample sizes of restitution ratios

CIα :

n

s t

y± 1−α,n 1−

CIα is the α% confidence interval for the sample sizes

α = significance level of the confidence interval

t, value can be obtained from the table provided in Dowdy and Wearden24

μ lies within the following limit:

1 , 1 1

,

1 − − ≤ ≤ + − −

−t n y t n

−t1− ,n−1

y α lower limit of the mean value

=+t1− ,n−1

y α Upper limit of the mean value

For the present research a confidence interval of 95% was chosen

Based on the analysis using t-distribution method the percentage error varied from ~ 3%

to 20% for the whole data set Table 2 contains the results for error estimates based on distribution method

t-4 Computational Analysis

Numerical simulations were conducted to determine the 3D flow field and associated particle trajectories through a single stage axial turbine Table 3 lists the geometrical parameters of the NASA developed turbine for automotive applications25 However, the turbine operational conditions were scaled as listed in Table 4 to meet the

requirements of a modern rotorcraft engine comparable to Bell Ranger 206

4.1 Turbine Geometry and modeling

Solid model of the turbine stator and rotor blades as shown in Fig 7 was generated from the airfoil profile data at three sections, the hub, mid span and tip

SolidWorks software was used to do the solid modeling The files were saved in igs

format so that they can be imported easily in mesh generating softwares like Gridgen

4.2 Grid generation

The Gridgen V15 software was used to create the domain of interest for numerical simulations Gridgen V 15 software is a very robust grid generating software which can

be used to generate both structured and unstructured grids More details about this software can be found in the Gridgen User’s Manual

Fig 8 presents the grid used in the turbine flow field numerical simulations The

computational grid used for the three-dimensional flow solution was of the H-type in both the stator and rotor blocks The vane passage H-type grid included 140 grid points in the stream wise direction, 80 grid points blade-to-blade, and 100 grid points in the span-wise direction The rotor passage H-type grid included 120 grid points in the stream wise direction, 60 grid points blade-to-blade, and 100 grid points in the span-wise direction Highly stretched mesh spacing was employed in the regions close to the blade passage surfaces to resolve the boundary layer next to the blade wall The combined grid consists

of 1.6 million cells The first grid point is taken at a y+ value of 30

4.3 Governing Equations

The solution to the Reynolds Averaged Navier Stokes (RANS) equations for turbulent compressible flow was obtained numerically using ANSYS CFX solver26 for unsteady N-S equations in their conservation form The turbulence model selected in this

study is the k - ω based Shear-Stress-Transport (SST) model of Menter27.The SST model was chosen because of its demonstrated capability (Bardina et al28) to accurately predict

the onset and the extent of flow separation under adverse pressure gradients Using this model proper transport behavior of turbulent shear stress can be obtained by a limiter to

the formulation of the eddy-viscosity The solver uses a high resolution advection scheme

which essentially uses a second order accurate upwind scheme everywhere except near discontinuities where it is first order accurate26 However, all turbulence equations always use the first order upwind advection scheme

The N-S equataions in conservation form:

S U U p

U U

T tot

tot

S U U U

U T

h U t

∇

∇ +

2 (

) (

4.5 Particle Trajectory Simulations

The particles were introduced after the flow solution was converged to about 8 orders of magnitude Since the particle loadings encountered in gas turbine applications are sufficiently low, the solid particle dynamics were simulated using one-way gas-

particle interaction models, which do not take into consideration the effects of dispersed

particles’ momentum exchange with the gas flow field

The particle trajectories were determined by stepwise integration of their governing equations of motion

R P

P P

D P

F

g U U F dt

U

d

+

−+

The terms on the right hand side represents aerodynamic, gravitational and forces due to rotating reference frame Inter-particle collisions and particle rotational drag are neglected The drag force is determined using the Schiller-Naumann (SN) drag model incorporated in Ansys CFX This drag model is particularly efficient in predicting drag for sparsely distributed spherical particles