203

An analytical approach to predicting particle deposit by fouling in the axial compressor of the industrial gas turbine
T W Song1, J L Sohn1 , T S Kim2, J H Kim3, and S T Ro1 1 School of Mechanical and Aerospace Engineering, Seoul National University, Seoul, Korea 2 Department of Mechanical Engineering, Inha University, Inchon, Korea 3 Turbomachinery Department, Korea Aerospace Research Institute, Daejon, Korea The manuscript was received on 13 May 2004 and was accepted after revision for publication on 30 September 2004. DOI: 10.1243/095765005X7547

Abstract: The gas turbine performance deteriorates with increased operating hours. Fouling in the axial compressor is an important factor for the performance degradation of gas turbines. Airborne particles entering the compressor with the air adhere to the blade surface and result in the change of the blade shape, which directly inuences the compressor performance. It is difcult to exactly understand the mechanism of compressor fouling because of its slow growth and different length scales of compressor blades. In this study, an analytical method to predict the particle motion in the axial compressor and the characteristics of particle deposition onto blade is proposed as an approach to investigating physical phenomena of fouling in the axial compressor of industrial gas turbines. Calculated results using the proposed method and comparison with measured data demonstrate the feasibility of the model. It was also found that design parameters of the axial compressor such as chord length, solidity, and number of stages are closely related to the fouling phenomena. Likewise, the particle size and patterns of particle distributions are also important factors related to fouling phenomena in the axial compressor. Keywords: fouling, gas turbine, axial compressor, collection efciency

INTRODUCTION

Performance degradation of the gas turbine is directly related to the change of the blade proles of compressor and turbine due to their deterioration caused by fouling, erosion, corrosion, foreign object damage (FOD), etc. Among these degradationrelated factors, compressor fouling, dened as the deposition process of airborne particles on compressor blades, is known to be the source of about 70 85 per cent of performance degradation of industrial gas turbines [1]. The atmosphere includes various types of airborne particles such as dirt, dust, salt, etc. that are sources of compressor fouling. Although the ltration system removes most of these particles,

Corresponding author: School of Mechanical and Aerospace

Engineering, Seoul National University, Seoul, Korea.

unltered particles enter into the compressor and adhere to the blade surface as a mixture of moisture and lubricants. Severe fouling problems can be resolved by installing highly efcient lters at the compressor inlet, but their use is restricted due to heavy cost and large pressure drop. In the axial compressor of the industrial gas turbine, fouling results in the change of the shape of leading edges of the blades and their surface roughness and, as a result, the airow in the compressor cascade becomes distorted. This alters the compressor characteristics. For example, as a result of compressor fouling, after 100 h of continuous operation, a low-pressure compressor experienced about 3 4 per cent drop in pressure ratio with 2 4 per cent drop in efciency and 10 percent drop in pressure ratio with 6 7 percent drop in efciency for a high-pressure compressor [2]. The turbine output power decreases with the compressor
Proc. IMechE Vol. 219 Part A: J. Power and Energy

A06304 # IMechE 2005

204

T W Song, J L Sohn, T S Kim, J H Kim, and S T Ro

performance degradation. This also results in the reduction in surge margin and dramatically unstable operation of the whole gas turbine. There are many previous studies for the prediction of the overall performance degradation of the gas turbine due to fouling phenomena. Saravanamuttoo and Lakshminarasimha [3] studied the effect of fouling on the gas turbine performance by assuming only one stage (rst or last stage) of the axial compressor affected by fouling. Aker and Saravanamuttoo [4] insisted that fouling progresses linearly along compressor stages and, as a result, the degradation of compressor stage decreases linearly in the downstream direction. Seddigh and Saravanamuttoo [5] quantied the effect of compressor fouling to the gas turbine performance degradation. Massardo [6] investigated the effect of fouling on pressure rise and efciency of the axial compressor. Detailed characteristics of particle depositions can be investigated using advanced computational tools such as computational uid dynamics [7]. Slater et al. [7] showed particle deposition characteristics on turbine blades in turbulent ow environment in a gas turbine using a computational uid dynamics technique based on a fully Eulerian two-uid approach. Their results provide detailed inter-related mechanisms among particle deposition, geometric characteristics of the blade and particle inertia, etc. Tarabrin et al. [8] invented an analytical model of the fouling mechanism considering the motion of foulant particles by simplication of a compressor blade as a cylinder. Although this simplication is good enough, real physical phenomena such as ow conditions around a blade cannot be reected. In the present study, a newly modied analytical model based on the cylinder model of Tarabrin et al. [8] is suggested. With this model, the parametric studies of a blade prole and the ow conditions are performed and the effects of particle size and particle distribution are also investigated. 2 CASCADE COLLECTION EFFICIENCY

the general mechanisms of the deposition of solid particles on a cylinder in the ow eld. A particle along a streamline encounters diffusion by Brownian motion, or is intercepted by the cylinder surface, or collides due to its inertia. When a particle oats in a slow ow eld, diffusion is the dominant physical phenomenon and, if the particle size is not less than that of a cylinder, interception is dominant [9]. Impact occurs in a fast ow eld due to inertia if the cylinder diameter is very large compared with the particle size. In this study, it is assumed that fouling occurs only when particles hit on the cylinder [8]. In order to examine the fouling effect due to impact, the collection efciency, E, is dened as the ratio of the number of particle impacts on the cylinder to that of all particles entering perpendicularly to the cylinder, as shown in Fig. 1. the number of particle impacts H on a cylinder E the number of particles entering L perpendicularly to the cylinder

(1)

Here, impacts to the cylinder means that, when particles enter perpendicularly to a cylinder, some particles are separated from the ow eld around cylinder by their inertia and collide with the cylinder. Equation (1) assumes that the entering particles are uniformly distributed and all collided particles do not bounce off but adhere. The collection efciency dened in the ow eld around a cylinder can be modied for the simplied cascade ow, as shown in Fig. 2, around a single axial compressor blade as follows H H L c sin (bb b1 )

(2)

Fouling is dened as the deposition process of airborne particles on the solid surface. Figure 1 shows

Fig. 1

Characteristics of particle motion around a cylinder

Fig. 2

Flow characteristics between two blades in the axial compressor [8]

Proc. IMechE Vol. 219 Part A: J. Power and Energy

A06304 # IMechE 2005

Predicting particle deposit by fouling

205

If considering a blade row instead of a single blade, the cascade collection efciency based on the above expression can be dened as Ec E c sin (bb b1 ) s sin b1 (3)

Cc in the above equation is the slip correction factor, which corrects the change of a small particles slip in the ow eld [11]. The slip correction factor must be taken into account when micro scale particles are dealt with and can be expressed as &  ' 0:55 Cc 1 Kn 2:514 0:8 exp Kn (8)

The cascade collection efciency means that ratio of the number of particle impacts on a blade to that of particles entering between two blades. From the above two equations, it is clear that the cascade collection efciency can be determined by blade characteristics and the width of ow eld containing particles possible to collide with a blade is H. Tarabrin et al. [8] assumed a blade as a cylinder and H was obtained using two-dimensional potential ow around a cylinder and derived the cascade collection efciency as follows   1 0:77 1 c sin (b b b1 ) Ec St s sin b1 where St denotes the Stokes number dened as St
2 tw {rp dp =(18m)}w Lc 2c sin (bb b1 )

(4)

Since particles move in the ow eld, the governing equation for the particle motion [equation (6)] must be solved simultaneously with governing equations of the ow eld, which requires CFD. Because of the physically complex ow eld in the compressor blade row, any advanced CFD tool cannot exactly capture all important physical characteristics of the ow eld. To derive a relatively simple analytic model, it is assumed, in this study, that there exist innite number of blades in an axial compressor row, and the blade shape can be represented by a plate, as shown in Fig. 3. The analysis can be categorized by two cases according to the ow characteristics between blades, as described below.

(5) 3.1 Model 1: constant velocity model

The Stokes number is the ratio of a particles stopping distance to a characteristic length. Zero Stokes number means that all the particles move along with the ow eld and there are no particle impacts on the cylinder (or a blade). When the Stokes number increases, the possibility may increase that a particle falls out by impact on the cylinder [10]. As shown by equation (4), if the ow velocity at the blade inlet or its angle of attack to the blade increases, the cascade collection efciency also increases.

Assuming that the velocity between blades is not variable, the right hand side of equation (6) becomes constant. With the x-axis set to be parallel with the blades and the y-axis perpendicular to the x-axis as indicated in Fig. 3, scalar components of equation (6) can be expressed as follows dvx vx w dt dvy vy 0 t dt

(9) (10)

A NEW ANALYTICAL MODEL TO PREDICT CASCADE COLLECTION EFFICIENCY

The expression of the cascade collection efciency, equation (4), derived by Tarabrin et al. [8], has a limitation when applied to the axial compressor due to the assumption of the ow over a cylinder. In this study, a new model for the cascade collection efciency is proposed by modelling a blade as a plate instead of a cylinder. The governing equation of a particle motion can be described as follows

dv vu dt
2 1 d p rp C c 18 m

(6)

where t is the relaxation time [9] dened as

(7)

Fig. 3

Simplied cascade model assuming at-plates as compressor blades

A06304 # IMechE 2005

Proc. IMechE Vol. 219 Part A: J. Power and Energy

206

T W Song, J L Sohn, T S Kim, J H Kim, and S T Ro

where vx dx , dt vy dy dt (11)

The solutions of the above equation can be categorized into three cases: Case I D c 1:12tw0 . 0 x 0:28tvyi f1 exp ( t=t)g c tan bb 0:28 tan bb r& c c l2 exp (l1 t) l1 exp (l2 t) t D 0:28   vyi vxi exp (l1 t) exp (l2 t)g tan bb

Analytic solutions of the above equations can be expressed as x wt t(vxi w)(1 et=t ) y vyi t(1 e
t=t

(12) (13) where

(19)

Here, vxi and vyi are x and y components of the particle velocity at the blade row inlet, respectively. Combining the above two equations, an expression for the particle position between the blades is derived as   y y x w t ln 1 (vxi w) vyi t vyi (14)

r 1 1 D l1 2t 2t c r 1 1 D l2 2t 2t c Case II D c 1:12tw0 0 x

(20) (21)

When a particle reaches the exit of a blade row, its position can be expressed as xc y tan bb (15)

0:28tvyi f1 exp ( t=t)g c tan bb 0:28 tan bb  &   ' vyi t c c vxi exp t 2t 0:28 tan bb 0:56t (22)

From the above two equations, the y value of a particle at the exit of a blade row can be calculated. With this value, the collection efciency, E, and the cascade collection efciency, Ec, are expressed as E H y/ sin bb L c sin (bb b1 )/ sin b1 c sin (bb b1 ) y s sin b1 s sin bb (16) (17)

Case III D c 1:12tw0 , 0 r !  ( t c 1 D cos t x exp 2t 0:28 2t c p )   sin 1=2t D=ct vyi c p vxi tan bb 0:56t 1=2t D=c 0:28tvyi f1 exp ( t=t)g c tan bb 0:28 tan bb (23) With these solutions, the collection efciency E and the cascade collection efciency Ec can be computed by the same expressions as equations (16) and (17). Detailed derivations to obtain equations (19) (23) are described in the Appendix 2.

Ec E

3.2

Model 2: variable velocity model

The main function of the compressor is to raise pressure by decreasing the ow velocity passing through blade rows. To account for the decrease in relative ow velocity inside the blade rows, this model assumes that the inlet and outlet ow velocities are satised with de Hallers criterion w2 =w1 . 0:72 and that the ow velocity along a blade decreases linearly. With these assumptions, the x-component particle motion can be expressed as below

RESULTS AND DISCUSSION

x vyi t(1 et=t )/ tan bb dvx vx w0 1 0:28 dt c (18)

&

'

where w0 is the relative ow velocity at the blade row inlet.

Proc. IMechE Vol. 219 Part A: J. Power and Energy

The feasibility study of the proposed analytical model is conducted with the 12 stage axial compressor proposed by Iwamoto et al. [12], and results are compared with those of Tarabrin et al. [8, 13]. Effects of compressor design parameters such as chord length and solidity on the fouling mechanism
A06304 # IMechE 2005

Predicting particle deposit by fouling

207

Table 1 Specication of blade and velocities of the compressor of Iwamoto et al. [12]
Chord, c Mean radius, rm Solidity, c/s Flow coefcient at design point, fd Degree of reaction Relative outlet angle from previous blade, a Stagger angle, bb Axial velocity Blade tip speed (rst stage) 7 cm 18.7 cm 1.5 0.547 0.5 158 40.68 150 m s21 350 m s21

are examined. The role of particle sizes and their distributions on the fouling process are also investigated. The geometric characteristics and design parameters of the compressor blade considered in the present study are shown in Table 1.

Fig. 5

Distribution of the cascade collection efciency over chord length and particle size (s 1.5)

4.1

Particle sizes and particle sizes when the solidity is xed. Here, the reference chord length (Cref) is selected as the chord length of the compressor of Iwamoto et al. [12]. It is shown that the cascade collection efciencies decrease with the increase in chord length. This may be caused by the decrease in the probability of particle impacts on the blade due to the increase in time during which particles move in the ow eld between the blades. Figure 6 shows that the cascade collection efciency increases linearly with the increase in solidity when the chord length is xed. This is not caused by the increase in particle impacts on the blades but is due to the decrease in the number of inlet particles by closer spacing of the blades when the solidity is low. These results correspond to those of Tarabrin et al. [8]. Those two gures also show that the particle size is an important parameter in this study as predicted by Fig. 4.

Figure 4 shows distributions of the cascade collection efciency with various particle sizes. As expected, the cascade collection efciency increases with growing of particle size. The cascade collection efciencies predicted by the cylinder-based model of Tarabrin et al. [8] are larger than those by the platebased model. This is caused by the increase in particle impacts on the cylindrical geometry, which has larger curvature than at-plates. The reason why the collection efciencies predicted by model 2 are larger than model 1 is due to the effect of increase in inertia force of the particles caused by the decrease in ow velocity between the two plates.

4.2

Blade chord and solidity

Figure 5 represents the distributions of the cascade collection efciencies with various chord lengths

Fig. 4

Distributions of cascade collection efciencies with various particle sizes

Fig. 6

Distribution of the cascade collection efciency over solidity and particle size (c 7 cm)

A06304 # IMechE 2005

Proc. IMechE Vol. 219 Part A: J. Power and Energy

208

T W Song, J L Sohn, T S Kim, J H Kim, and S T Ro

4.3

Flow rates

As expressed in equation (4), the cascade collection efciency is increased with the decrease in the inlet ow angle (b1) or the increase in the relative ow velocity (w) at blade row inlet. These two parameters are dependent on the magnitude of the ow rate. Figure 7 shows that the cascade collection efciencies decrease with the increase in the ow coefcient. Here, the ow coefcient is normalized by the ow coefcient at the design point of the compressor of Iwamoto et al. [12]. The decrease in the ow coefcient means a reduced ow rate of the compressor at constant rotational speed. The increase in cascade collection efciency with a reduced ow coefcient is dominated by an increase in particle sizes. From these results, it can be concluded that compressor fouling will be severe when the inlet air contains relatively large fouling particles entering the compressor at low velocity.

Fig. 8

Inuence of the inlet ow angles on the cascade collection efciency with various particle sizes

4.5

Compressor stages

4.4

Inlet ow angles

The multi-stage axial ow compressor in heavy-duty gas turbines is generally equipped with variable inlet guide vanes (VIGV) or variable stator vanes (VSV) for a stable start-up and the enhancement of exhaust heat recovery [14]. The closure of the guide vanes changes the direction of incoming ow and brings the decrease of the inlet ow angle (b1). Figure 8 shows the inuence of the inlet ow angle on the cascade collection efciency with a condition of xed ow rate. The decrease of inlet ow angle enhances the possible interaction of incoming particles on the blade surface as shown in Fig. 2 and, therefore, increases the cascade collection efciency. This result can also be clearly understood from equation (4).

Assuming that particles entering a cascade are uniformly distributed and specications of all stages of the axial compressor are the same, when particles enter into a cascade in proportion to s sin b1 , some particles deposit on the blade surface in proportion to Ec sin (bb b1 ). Then, the number of particles carried away through a stage is proportional to s sin b1 Ec sin (bb b1 ). As described in Figs 48, the amount of deposited particles on the blade surface increases when the particle size is large. This result shows that most of the large particles are deposited on the front stages and, as a consequence, the number of particles entering the downstream stages is decreased. Figure 9 shows that the number of particles passing through the stages is decreased in the downstream stages. The decrease in the number of particles in the downstream stages is due to the deposition of large particles on front stages. This result is that the fouling is dominant especially in the front stages of the axial compressor.

Fig. 7

Distribution of the cascade collection efciency over the ow coefcient for various particle sizes

Fig. 9

Distribution of the number of particles captured by consecutive compressor stages for the variation of the particle size

Proc. IMechE Vol. 219 Part A: J. Power and Energy

A06304 # IMechE 2005

Predicting particle deposit by fouling

209

4.6

Characteristics of particle distributions

In the above discussions, it was assumed that the ow entering the axial compressor contains the same size of particles for each operating condition. However, in a real situation, there exist various sizes of particles in the atmosphere and, therefore, it is also necessary to consider characteristic distributions of particle sizes contained in the inlet air. Since the distribution of particle sizes may change according to the location and time, effects of arbitrary particle distributions on the overall fouling mechanism are considered in this study. In most industrial gas turbines, a ltration system must be installed in front of the inlet duct to protect the blades from fouling and erosion by airborne particles in the incoming air. The type of the lter determines the maximum particle diameter. Filters can be categorized such as inertial lters, self-cleaning pulse lters, high efciency lters etc. Among these lters, the inertial lter is the simplest and cheapest one, which can remove particles with diameters over 20 mm [1]. In the present study, the maximum allowable particle diameter is set as 10 or 20 mm depending on the lter characteristics. For the simplicity of the analysis, it is assumed that the stability time of the deposition layer, dened as the time elapsed from the beginning of particle deposition on the blade to the instant when the deposition layer does not grow any more, is the same for each stage of the axial compressor. And, it is also assumed that the rate of different particle sizes in each stage is the same, as shown in Fig. 9. This tendency is preserved regardless of the distribution of particle sizes of the rst stage. The distributions of particle sizes of the rst stage used in this study are shown in Fig. 10. Here, P(dp) means the mass (mg) of a particle with its diameter of dp (mm) per unit volume (m3) of air. Two kinds of maximum particle diameters such as 10 and 20 mm are chosen as described above. The non-dimensional collection mass of particles in each stage is dened as the collection of particle masses from the smallest size (0.1 mm) up to largest one (for example 20 mm) in the stage normalized by the particle mass in the rst stage. The nondimensional collection mass of particles in the jth stage, aj, can be expressed as follows P i (Ni,j Ni,j1 ) P(dp,i ) aj P i (Ni,1 Ni,2 ) P(dp,i ) (24)

Fig. 10

Distribution patterns of particles adopted in the present study. (a) Cases A C; (b) cases DF; (c) cases A, D, H, and G

where subscript i denotes ith particle size in each stage. Figure 11 shows non-dimensional collection masses of particles fallen out in each compressor stage for three different patterns of particle distributions with their maximum particle size of
A06304 # IMechE 2005

20 mm. The predictions are compared with measured data of the experimental compressor [13]. Differences between measured and predicted data especially in downstream stages are due to the reduction in the adhesion force of particles to the blades by the evaporation of moisture contained in the air in the high temperature and pressure environment. In case of a dominant amount of
Proc. IMechE Vol. 219 Part A: J. Power and Energy

210

T W Song, J L Sohn, T S Kim, J H Kim, and S T Ro

Fig. 11

Effect of patterns of particle distributions on the non-dimensional collection mass of particle with maximum particle diameter of 20 mm

Fig. 13

Effect of patterns of particle distributions on the non-dimensional collection mass of particles with different maximum particle sizes

large particles (case A), most of the large particles adhere to the front stages and, as a result, nondimensional collection masses of particles in downstream stages are relatively small. On the other hand, when the mass of small particles is dominant (case C), non-dimensional collection masses of particles are increased in downstream stages because of the small portion of mass of large particles. Figure 12 represents similar results as Fig. 11 with the maximum size of particles of 10 mm. Figure 13 shows that the slope of the particle size distribution does not have a strong inuence on the pattern of particle adhesion on the blades. Instead, comparing with case D, it can be concluded that the maximum size of particles, which is governed by the inlet lter, strongly inuences the characteristics of particle adhesion to each stage of the axial compressors.

CONCLUSIONS

An analytic method has been developed for the analysis of fouling phenomena in the axial compressor of the industrial gas turbine based on a at-platecascade model considering variable velocity between blades. Effects of geometric parameters and ow characteristics of the axial compressor on fouling mechanism are investigated as results of parametric studies. The impact of the characteristics of fouling particles, such as their size and distribution, on the fouling phenomena are also studied. Some important conclusions disclosed by the present study are as follows: (1) Fouling is closely related to the characteristics of geometry and ow of the compressor stage. Adhesion of particles to blades, dened as the cascade collection efciency, is increased with the decrease in chord length and the increase in solidity. Also, fouling is expedited with the decrease in the ow rate and the inlet ow angle. (2) A large particle size increases the cascade collection coefcient. Deposition of large particles in front stages makes fouling in front stages dominant. Small particles pass through front stages and inuence downstream stages. (3) Distribution of particle sizes is an important parameter to determine the fouling level. Also, the maximum size of particles, which is governed by the inlet lter characteristics, plays an important role to the fouling characteristics. ACKNOWLEDGEMENT

Fig. 12

Effect of patterns of particle size distributions on the non-dimensional collection mass of particles with a maximum particle diameter of 10 mm

This work was supported by the Electric Engineering and Science Research Institute of the Ministry of Commerce, Industry and Energy in Korea.
A06304 # IMechE 2005

Proc. IMechE Vol. 219 Part A: J. Power and Energy

Predicting particle deposit by fouling

211

REFERENCES
1 Diakunchak, I. S. Performance deterioration in industrial gas turbines. ASME J. Engng Gas Turbines Power, 1992, 114, 161 168. 2 Mezheritsky, A. D. and Sudarev, A. V. The mechanism of fouling and the cleaning technique in application to ow parts of the power generation plant compressors. ASME Paper 90-GT-103, 1998. 3 Saravanamuttoo, H. I. H. and Lakshminarasimha, A. N. A preliminary assessment of compressor fouling. ASME Paper 85-GT-153, 1985. 4 Aker, G. F. and Saravanamuttoo, H. I. H. Predicting gas turbine performance degradation due to compressor fouling using computer simulation techniques. ASME J. Engng Gas Turbines Power, 1989, 111, 343 350. 5 Seddigh, F. and Saravanamuttoo, H. I. H. A proposed method for assessing the susceptibility of axial compressors to fouling. ASME J. Engng Gas Turbines Power, 1991, 113, 595 601. 6 Massardo, A. F. Simulation of fouled axial multistage compressors. Proceedings of IMechE Conference on Turbomachinery, 1991, Paper C423/048, 243 252. 7 Slater, S. A., Leeming, A. D., and Young, J. B. Particle deposition from two-dimensional turbulent gas ows. Int. J. Multiphase Flow, 2003, 29, 721 750. 8 Tarabrin, A. P., Schurovsky, V. A., Bodrov, A. I., and Stalder, J. P. An analysis of axial compressor fouling and a blade cleaning method. ASME J. Turbomachin., 1998, 120, 256 261. 9 Heinsohn, R. J. and Kabel, R. L. Sources and Control of Air Pollution. 1999, Prentice Hall, Engelwood Cliffs, NJ. 10 Hinds, W. C. Aerosol Technology: Properties, Behavior, and Measurement of Airborne Particles, 1982 (John Wiley & Sons, New York). 11 Jenning, S. G. The mean free path. J. Aerosol Sci., 1988, 19, 159 166. 12 Iwamoto, T., Ikesawa, K., Kanayama, T., Nagai, K., Yukinari, A., and Nakagawa, T. Development of a high-pressure ratio axial ow compressor. Proceedings of the 1991 Yokohama International Gas Turbine Congress, 1991, II, pp. 79 86. 13 Tarabrin, A. P., Schurovsky, V. A., Bodrov, A. I., and Stalder, J. P. Inuence of axial compressor fouling on gas turbine unit performance based on different schemes and with different initial parameters. ASME Paper 98-GT-416, 1998. 14 Kim, J. H., Song, T. W., Kim, T. S., and Ro, S. T., Model development and simulation of transient behavior of heavy duty gas turbines. ASME J. Engng Gas Turbines Power, 2001, 123, pp. 589 594.

Cx dp E Ec H Kn L Lc N P(dp) s St t u U v vxi vyi w

axial velocity of uid, m s21 particle diameter, m collection efciency cascade collection efciency width of particle groups impacting on the cylinder, m Knudsen number radius of cylinder, m characteristic length, m number of entering particles particle distribution function, mg m23 pitch, m Stokes number time, s velocity vector of uid, m s21 tangential speed, m s21 velocity vector of particle, m s21 x-component of particle velocity at blade inlet, m s21 y-component of particle velocity at blade inlet, m s21 relative ow velocity in blade passage, m s21 relative blade outlet angle of previous stage, deg ow or particle angle, deg stagger angle, deg ow coefcient (Cx/U ) dynamic viscosity, kg m21 s21 particle density, kg m23 solidity (c/s) relaxation time, equation (7)

a b1 bb w m rp s t

Subscripts b p 1 2 blade particle in out

APPENDIX 2 Detailed procedure for the solutions of the variable velocity model Equation (18) can be arranged in terms of xcomponent as follows

APPENDIX 1 Notation aj c Cc dimensionless deposited mass, equation (24) chord, m slip correction coefcient, equation (8)

d2 x 1 dx 0:28w0 w0 0:28w0 vyi x (1 et=t ) dt 2 t dt ct t c tan bb (A1) The procedure for solving linear second-order ordinary differential equation [equation (A1)] has three steps as follows.
Proc. IMechE Vol. 219 Part A: J. Power and Energy

A06304 # IMechE 2005

212

T W Song, J L Sohn, T S Kim, J H Kim, and S T Ro

Step 1 Find the general solution, xh of the homogenous part of equation (A1) d2 x 1 dx 0:28w0 x0 dt 2 t dt ct (A2)

By inserting equation (A6) into equation (A1), c1(t) and c2(t) can be obtained as et(l1 1=t) fet=t w0 (1 l1 t) 0:28w0 vyi t= c1 (t) c tan bb et=t (1 l1 t) l1 tg l1 (l1 l2 )t(1 l1 t) (A7a) et(l2 1=t) fet=t w0 (1 l2 t) 0:28w0 vyi t= c2 (t) c tan bb et=t (1 l2 t) l2 tg l2 (l2 l1 )t(1 l2 t) (A7b)

The general solution of the homogenous problem is classied into three according to the sign of the determinant coefcient of the characteristic equation. The characteristic equation and the determinant of equation (A2) are obtained as 1 0:28w0 l2 l 0 t ct D c 1:12tw0 (A3) (A4)

Then, the particular solution becomes xp (t) c1 (t)el1 t c2 (t)el2 t 0:28tvyi f1 exp ( t=t)g c tan bb 0:28 tan bb (A8)

According the sign of D, the general solution of equation (A2) can be solved as follows (1) D . 0 (l l1 , l2 [ R) (2) D 0 (l1 l2 [ R) (2) D , 0 (l a + ib) xh (t) c1 el1 t c2 el2 t (A5a) xh (t) c1 el1 t c2 tel1 t (A5b) xh (t) e (c1 cos bt c2 sin bt)
at

Therefore, the general solution of the full nonhomogenous problem is x(t) xh (t) xp (t) c1 x1 (t) c2 x2 (t) 0:28tvyi {1 exp ( t=t)} c tan bb 0:28 tan bb

(A9)

(A5c)

where c1 and c2 are constants, which can be determined by the initial conditions. Step 2 Find the particular solution of the non-homogenous part of equation (A1). The particular solution is obtained by using the method of variation parameters. Let the particular solution be xp (t) c1 (t)x1 (t) c2 (t)x2 (t) (A6)

Here, x1(t) and x2(t) can be determined with equation (A5) based upon the properties of the determinant of the characteristic equation.

Step 3 Find coefcients c1 and c2 of equation (A9) with initial conditions; x0 (0) vxi and x(0) 0. According to the determinant of the characteristic equation, equation (A9) becomes equation (19), (22), or (23).

Proc. IMechE Vol. 219 Part A: J. Power and Energy

A06304 # IMechE 2005