Minimum gas speed in heat exchangers to avoid particulate fouling

3946 M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955Fig. 2. Particle size distributions (PSD, left) and scanning electron microscope (SEM, right) photos of glass (a,b), bronze (c,d) andcopper particles (e,f), respectively.The cooling airflow rate is also controlled through acritical nozzle and a pressure regulator.Fouling of the heat exchanger tube is monitoredusing a digital camera through a glass window installedin the heat exchanger section as shown in Fig. 1b. Thedigital camera is the Finepix S602 zoom, with a resolutionof 2832 pixels · 2128 pixels, i.e. 6.03 million pixels.The digital camera has a 6· optical zoom which comparedto small image format has a zoom area of 35–210mm. The fouling layer thickness is measured optically.Images are taken of the heat exchanger tubes before andduring fouling. The clean heat exchanger tube radius istaken as a reference. The tube radii in the camera imagesand hence the fouling layer thicknesses are determinedwith an accuracy of ±1 pixel which corresponds to ±0.04mm. The temperatures of the hot air and the cooling airbefore and after the heat exchanger are measured usingK type thermocouples with an accuracy of ±0.4 °C. Theair speed at the exit of the heat exchanger section ismeasured by a hot wire anemometer with an accuracy of±0.015 m/s. Air exhaust from the setup is filteredthrough a bag filter before it goes into the atmosphere.Each fouling experiment is carried out in the followingprocedure: Hot and cold airflow rates are adjustedto the desired flow rates at the beginning of theexperiment. The hot air temperature is adjusted throughthe air heater and the experiment is kept running till asteady state is reached. After a steady state has beenreached, particles are injected at the desired rate and thefouling process is monitored.3. Results and discussion3.1. Influence of flow speed on growth of fouling layerThe following experiments are performed to see theinfluence of flow speed on the growth of the fouling layerand to determine the flow speed at which fouling ceases.Two types of particles are used in the experiments, thenon-metallic glass particles and the metallic bronze andcopper particles defined in the previous section.In the first experiments, glass particles have beeninjected in the flow at different gas flow speeds and thefouling behavior is monitored. The average airflowspeed between the tubes has been changed from 2.7 to3.8 to 5 m/s and in each case the operating time was 9 h.The air temperature is 200 °C and the particles

M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955 3947Fig. 3. Fouling layer at the end of 9 h of operation for different flow speeds. The foulant is spherical glass particles of diameter 21 ± 16lm. (a) V ¼ 2:7 m/s; (b) V ¼ 3:8 m/s; (c) V ¼ 5 m/s.Table 1Comparison between the fouling behaviors for glass particles under different operating conditionsAverage airflow speed between the HE V ¼ 2:7 m/s V ¼ 3:8 m/s V ¼ 5 m/stubesFinal thickness (mm) 1 0.75 0.3Circumferential and radial growth Continues Starts very fast and thendecreases abruptlyStarts very fast and thendecreases abruptlyconcentration is 0.15 g/N m 3 of flowing air. Fig. 3 showsthe fouling layer at the end of operation for the abovementionedcases. Table 1 shows a comparison betweenthe three different cases. It is shown that as the speed ofthe flow increases, the fouling layer surface area andthickness are reduced. Increasing the average air speedbetween the heat exchanger (HE) tubes from 2.7 to 5 m/sreduces the final fouling layer thickness from 1 to 0.3mm after 9 h of operation. When the average flow speedbetween the HE tubes is 2.7 m/s, it is found that thecircumferential and radial growth of the fouling layercontinued from the beginning till the end of the experiment.Increasing the average flow speed between the HEtubes to 3.8 and 5 m/s showed a different foulingbehavior. The fouling rate was very high at the beginning,especially for the first 3 h. Then the fouling ratestarted to decrease abruptly for the next 6 h of operation,till hardly any radial or circumferential growth canbe seen.Figs. 4 and 5 show the growth of the fouling layer forglass particles over the heat exchanger tube as a functionof time. The average air speed between the tubes is 2.7m/s for Fig. 4 and 3.8 m/s for Fig. 5. Fig. 4a shows thatat the beginning of the fouling process, some particlesstart to deposit at intermittent and distant positions.These points of deposition start to grow radially, Fig.4b–e, till the whole surface of the heat exchanger tube iscovered completely with particles, Fig. 4f. The surficialgrowth of the fouling layer is emphasized by the encircledareas in Figs. 4a–f. The surficial growth of thefouling layer can be explained by the fact that fineparticles are most likely to stick first to the HE tubesbecause of their higher sticking velocity compared tocoarse particles [7]. Then larger particles, which depositon the heat exchanger tube and stand alone in theshearing flow, start to roll over the heat exchanger tubeif the shearing flow speed is above a certain limit. Theparticle rolling motion is stopped when it is blocked by astanding heap of particles, which leads to the surficialgrowth of the fouled areas till they cover the unfouledsurfaces. The flow speed required to roll a particleresting on a flat surface, and, the sticking velocity of aparticle, will respectively be discussed in detail in thefollowing subsequent sections.Fig. 5 shows that at the beginning of the foulingexperiment, Fig. 5b, the glass particles in the flow startedto deposit on the rear end of the HE tube and continuedto the circumference, Fig. 5c–e. The rear end of the heatexchanger, Fig. 5f, is subjected to the lowest shearingforces due to separation of flow from the tube surface,Achenbach [9]. The flow field for the examined HEshown in Fig. 1 is solved using the finite volume methodapplied by the commercial CFD-package CFX version4.2. The low Reynolds number k–e model is used to solvethe flow field around the HE tubes. For all but theadvection terms, use is made of a second-order centraldifferencing scheme. In the equations for the velocitycomponents, use is made of the third order QUICKinterpolation scheme for the advection terms. For the kand e equations the first order HYDRID-scheme is appliedwhich is more stable than the QUICK scheme andensures the k and e values to remain positive in theconverged solution. In the computations use is made ofthe SIMPLEC pressure-correction scheme in which thepressure field is indirectly specified via the continuityequation. For the u and the v velocity use is made of theStone’s method, for the turbulence quantities a linerelaxationmethod is used and an algebraic multi-gridmethod is applied for the pressure equation. The flowfield is solved for a turbulent flow, at Re ¼ 2 10 5 , which

3948 M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955Fig. 4. Progress of the fouling layer over the HE tube within 9 h of operation. The foulant is spherical glass particles of diameter21 ± 16 lm. The average air speed between the HE tubes is 2.7 m/s: (a) after 1 h of operation, (b) after 1.5 h of operation, (c) after 2 h ofoperation, (d) after 2.5 h of operation, (e) after 3 h of operation, (f) after 9 h of operation.Fig. 5. Growth of the fouling layer over the HE tube within 6 h of operation. The foulant is spherical glass particles of diameter 21 ± 16lm. The average air speed between the tubes is 3.8 m/s: (a) at the beginning, (b) after 0.5 h of operation, (c) after 1 h of operation, (d)after 1.5 h of operation, (e) after 2 h of operation, (f) after 6 h of operation.corresponds to a superficial flow velocity of 10 m/s. Fig. 7shows the velocity contours in the y and x directions forthe HE shown in Fig. 1. Although the CFD calculationsshow that the flow around the middle centered tube is notperiodic, this centered tube can be regarded as a representativefor a tube in a tube bundle. It is shown that for

M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955 3949Fig. 6. A summary for the fouling layer circumferential growth shown in Fig. 5.the middle centered tube the flow keeps attached to thetube surface and then separates at an angle of 120° fromthe stagnation point. From the numerical results it couldalso be concluded that there is more air circulation at thefront than in the rear end of the middle centered HEtube, Fig. 7c and d. Therefore particles that stick to therear end of the HE tube are more likely to remain inposition than particles that stick to the sides of the HEtube. A particle that sticks to the side of the HE tube rollsover the HE tube if the airflow speed is above a certainlimit, and stops if it is blocked by the resting particles atthe rear end of the heat exchanger tube. Therefore foulingstarts at areas that are subjected to minimum shearingforces and then spreads due to particle transport tothe already deposited particles. Fig. 6 summarizes thecircumferential growth of the fouling layer shown in Fig.5.3.2. Limiting fouling speedThe previous set of experiments has shown that asthe flow speed in the HE increases, the thickness and thesurface area of the fouling layer deposited over the HEtube are reduced. The following experiments are done todetermine the flow speed at which fouling of the heatexchanger is prevented. The results of the experimentwill be compared to the results achieved with an analyticalmodel. The analytical model is only valid forspherical non-charged particles. The analytical model isbased on the minimum shear velocity required to roll aparticle on a flat surface. It can be applied for a cylindricaltube if the particle diameter is much smaller thanthe cylinder diameter, as is the case here. The shear flowspeed required to roll a particle deposited on a flatsurface is called the critical flow velocity and it is afunction of the particle material and size and the surfacematerial. The critical flow velocity can be calculatedwhen assuming that the hydrodynamic rolling momentdue to the drag force, F d , lift force, F L , and buoyancyforce, F b , is greater than the resting adhesion momentdue to the surface adhesion force, F a , and the force ofgravity, F g . The ratio between the hydrodynamic rollingmoment and the adhesion resting moment is defined byZhang et al. [10] as RM and equal toRM ¼¼Hydrodynamic rolling momentAdhesion resting momentF d ð1:399R aÞðF a þ F g F b F L Þd=2 ; ð1Þwhere R, d and a are the particle radius, contact diameterand the particle radius of deformation, respectively.If the RM value is greater than 1, rolling will takeplace. Abd-Elhady et al. [5] have calculated the criticalflow velocity for spherical copper particles of diametersranging from 5 to 50 lm, the results are shown in Fig. 8.The main flow stream velocity, i.e. the critical flowvelocity, required to roll a copper particle of 10 lm is10.5 m/s while for a 50 lm copper particle this velocity is4.5 m/s.Due to the comparison with the analytical modelpresented above, in the experiments only sphericalmetallic particles are used. The HE is operated at differentair flowing speeds and different particle sizes todetermine the operating speed at which fouling is preventedas function of the particles size in the flow. Theflow speed at which fouling is prevented, is investigatedfor copper particles of average diameter 10 lm with astandard deviation of 5 lm suspended in the airflow andalso for bronze particles of average diameter 55 lm witha standard deviation of 6 lm. The air temperature is 20°C and the particle concentration is 2 g/m 3 of airflow.The heat exchanger is kept running for 8 h in eachexperiment. Experimental results of the fouling experimentare plotted in Fig. 8. Fig. 8 shows that when air iscontaminated with the copper particles of averagediameter 10 lm the HE tubes fouled when the averageflow speed between the tubes was lower than 9.5 m/s, i.e.

3950 M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955Fig. 7. Velocity contours in the y (left) and x (right) directions for the HE section (a,b) shown in Fig. 1b and for the middle centeredtube (c,d), respectively. Reynolds number is 2 · 10 5 .7.5 and 2.7 m/s. It is also shown that when the air iscontaminated with the bronze particles of averagediameter 55 lm the HE tubes fouled when the averageflow speed between the tubes was lower than 5.5 m/s, i.e.4.5 and 2 m/s. Fig. 9 shows images for the HE tube at theend of the fouling experiments in case the airflow iscontaminated with bronze particles of average diameter55 lm. Fig. 9 shows that when the flow speed betweenthe tubes is equal to 2 m/s the fouling layer thicknessbecomes 0.3 mm and diminishes when the flow speedbecomes 5.5 m/s. The minimum flow speed at whichfouling in a heat exchanger is prevented will be called thelimiting fouling speed.The copper particles used in the experiments have awide particle size distribution ranging from 1 to 20 lm,as shown in Fig. 2e. Surprisingly, no fouling was found

M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955 3951Air flow velocity (m/s) .1614121086420Copper particles0 10 20 30 40 50 60Particle's diameter (µm).of the copper particles smaller than 10 lm at a flowvelocity of 9.5 m/s. This could be due to the smallamount of the fine particles present and due to the removalof them by larger hitting particles. The removal offine particles by the large depositing particles leads to alower limiting fouling speed than the limiting foulingspeed, which corresponds to the smallest particle in thegas flow. The effect of the particle size distribution onthe fouling mechanism still has to be investigated.The critical flow velocity shown in Fig. 8 is calculatedfor a metallic non-charged spherical copper particleresting on a good conductor surface. In case of anelectrically charged particle resting on a bad conductorsurface like glass an electrostatic adhesion force isdeveloped which leads to a higher critical flow velocitythan in case of a non-charged particle. The critical flowvelocity then should be calculated based on the nature ofthe particles in the flow and the flow conditions.3.3. Sticking of particlesCritical flow velocity (Theoretical).No Fouling Condition (Experimental).Fouling Condition (Experimental).Bronze particlesFig. 8. The critical flow velocity and limiting fouling speedversus particle diameter.In practical applications of heat exchangers especiallyin waste incinerators, the flue gases contain a wide rangeof particle sizes. To prevent fouling of heat exchangersbased on the critical flow velocity, the critical flowvelocity should be selected based on the particle sizemost likely to stick first to the heat exchanger tubes. Thefollowing experiments have been done to determinewhich particle size deposits first from a batch of particlesflowing over a heat exchanger tube. A mixture of equalmasses of bronze and copper particles is injected into theairflow. The bronze and copper particles used are definedin the experimental setup section. The particlesconcentration is 2 g/m 3 of airflow. Hot air at 200 °C andflow rate of 13E)3 Nm 3 /s is passed to the heat exchangersection. This results in a Reynolds number of700 based on the HE tube diameter and the superficialflow velocity. An airflow rate of 1.5E)3 Nm 3 /s and atemperature of 20 °C was used for cooling the heat exchangertube internally. The heat exchanger has run for9 h. At the end of the fouling experiment samples aretaken from the fouling layer deposited on the heat exchangertube and these are examined by SEM. Twosamples of particles are taken from the fouling layerdeposited on the middle centered tube of the heat exchangersection: The first sample of particles was scratchedfrom the top of the fouling layer. The secondsample of particles was scratched from the first foulinglayer of particles, which had deposited on the tube of theheat exchanger. This sample was taken by blowing airthrough the HE at a very high speed such that most ofthe fouling layer was blown off and only a very thin layerof particles remained on the tube of the heat exchanger.Fig. 10 shows SEM pictures for the particles at thebottom (near the heat exchanger tube) and at the top ofthe fouling layer respectively. First the fine copper particleshave accumulated over the heat exchanger tubeand then afterwards the larger bronze particles. The topof the fouling layer consisted mainly of the large bronzeparticles surrounded by the fine copper particles.The same kind of experiment above is repeated usingspherical glass particles instead of using a mixture ofFig. 9. Photos for the examined heat exchanger tube at the end of operation under different airflow velocities: (a) V ¼ 2 m/s, a bronzelayer of thickness 0.3 mm is formed on the HE steel tube; (b) V ¼ 4:5 m/s, a very fine layer of bronze particles is formed; (c) V ¼ 5:5 m/s, no fouling.

3952 M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955Fig. 10. Scanning electron microscope photos for the particles at the bottom (left) and top (right) of the fouling layer. The fouling layeris composed of a mixture of copper and bronze particles.Fig. 11. Scanning electron microscope photos for the particles at the bottom (left) and top (right) of the fouling layer. The particles aremade of glass.bronze and copper particles. Fig. 11 shows SEM photosfor the particles at the bottom (near the heat exchangertube) and top of the fouling layer. At the top of thefouling layer large particles have accumulated with smallparticles in between. In the fouling layer near the tube ofthe heat exchanger, the finest particles in the flow haveagglomerated together to form islands of fine particlesconnected to bigger particles. Fig. 12 and Table 2 show acomparison between the particles size distribution of theinjected glass particles in the airflow and the glass particlesin the fouling layer near the HE tube. It is shownthat the average diameter of the glass particles has reducedfrom 20.4 lm in the airflow to 14.7 lm inthefouling layer near the HE tube. 90% of the injected glassparticles in the airflow have a diameter less than 43.1 lmwhile 90% of the glass particles in the fouling layer nearthe HE tube have a diameter less than 25.2 lm. Thereduction in particle size in the fouling layer near the HEtube from the particle size in the airflow is due to thehigher sticking velocity of small particles compared tolarge particles as defined by Rogers and Reed [7].Rogers and Reed [7] developed a model that describesthe adhesion of a particle to a surface following anelastic–plastic impact. The model is based upon considerationof the energy balance during a normal impactbetween a sphere and a massive plane. The model describesthe adhesion of a particle arising from only elasticdeformations occurring during the approach of the particleto the surface. When the particle velocity is reducedto zero during the approach phase, part of the particleinitial kinetic energy, Q k , is converted into stored elasticenergy, Q E , while the remainder, Q L , is dissipated. This isthen followed by the recovery of the stored elastic energy,which is converted into kinetic energy of the reboundingparticle. For a particle of mass m impacting a surfacenormally with a velocity V im we have1mV 22 im þ Q A ¼ Q E þ Q L ;ð2Þwhere Q A is the energy due to the attractive forces betweenthe incoming particle and the surface. If the storedelastic energy, Q E , is greater than the adhesive energyrequired to separate the particle from the surface, Q 0 A ,then the particle will rebound otherwise it sticks to thesurface. The rebound speed, V r , is calculated from1mV 22 r¼ Q E Q 0 A : ð3ÞThe mentioned energy terms are described in detail byRogers and Reed and in van Beek [11]. The maximum

M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955 3953Fig. 12. A comparison between the particle size distribution of the injected glass particles in the airflow and the glass particles in thefirst fouling layers deposited on the HE tube.Table 2A comparison between the particle size distribution of the injected glass particles in the airflow and the glass particles in the first foulinglayers deposited on the HE tubea b cMean diameter (d M ) Standard deviation (SD) D 10 D 50 D 90Injected glass particlesin airflowGlass particle in thefouling layer nearthe HE tube20.4 16.2 3.6 16.2 43.114.7 8.4 5.9 13.3 25.2All dimensions are in micrometers.a The particle diameter at which 10% of the particles are smaller than it.b The particle diameter at which 50% of the particles are smaller than it.c The particle diameter at which 90% of the particles are smaller than it.impact speed at which an incident particle will stick tothe impacted surface, i.e. V r ¼ 0, is defined as the criticalsticking velocity. To show the variation of the stickingvelocity with the incident particle diameter the Rogersand Reed model is solved to calculate the stickingvelocity for copper particles of diameters 10 and 4 lmhitting a solid steel surface. Fig. 13 shows the variationof the coefficient of restitution, e, with the impact speed,V im , for copper particles of diameters 10 and 4 lm. Thecoefficient of restitution is defined as the ratio betweenthe normal rebound speed, V r , and the normal impactspeed, V im , for a particle hitting a solid surface. Fig. 13shows that if a copper particle of diameter 10 lm hits asteel surface at a speed lower than 0.025 m/s it will stick,while a copper particle of diameter 4 lm sticks if it hitsat a velocity lower than 0.12 m/s. Therefore, if we have amixture of particles of diameters 10 and 4 lm in air,which is flowing with an average speed of 1 m/s over aHE tube, there will be a larger number of particles thathit the HE tube with a speed lower than 0.12 m/s thanthe number of particles that hits with a speed lower than0.025 m/s. This can explain why small particles are likelyto dominate at the beginning of the fouling process.3.4. Sintering of fouling layersOnce fouling has started and particles accumulate tobuild up the fouling layer, sintering takes place. Sinteringchanges the fouling layer structure from a weakpowdery layer to a solid stable structure strongly attachedto the heat exchanger tubes [12]. Sintering is afunction of time and of the hot gas temperature flowingover the fouling layer [13]. Fig. 14 shows a cross sectionalview for a fouling layer taken from the previousexperiments mentioned above. The particles in the layerare bronze and were subjected to hot air at 200 °C for

3954 M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955Coefficient of restitution, e = V r / V im.10.90.80.70.60.50.40.30.20.10Sticking limit =0.025 m/sSticking limit = 0.12 m/sdp=10 micrometersdp=4 micrometers0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5Impact speed V im (m/s).Fig. 13. Coefficient of restitution versus impact speed for copper particles hitting a solid steel surface.Fig. 14. A cross sectional image through the fouling layer, thearrows show the position of the scratches.9 h during the fouling experiments. Fig. 2d shows a SEMphoto for the bronze particles before usage in the foulingexperiments. If we compare Figs. 2d and 14 we find thatmost of the bronze particles in the fouling layer havescratches on the surface compared to the initial particles.These scratches are probably the areas of contact whichhave been made up due to sintering and were brokenduring splitting of the fouling layer for examinationunder the SEM. Sintering makes removal of the foulinglayer by soot blowing afterwards difficult.An experimental setup has been built and developedto study the mechanism of fouling over heat exchangerstubes as a function of flow speed and contaminatingparticles. Fouling starts at areas that are subjected tominimum shearing forces, the rear end of the tubes andthen spreads to the circumference due to particle transportto the already deposited particles. Fine particlesstick first to the heat exchanger tubes because of theirhigher sticking velocity than coarse particles. Then thelarge particles deposit and the fouling layer starts tobuild up. As the flow speed in the heat exchanger increases,the thickness and the surface area of the foulinglayer deposited over the heat exchanger tube are reduced.The flow speed at which fouling ceases is calledthe limiting fouling speed. The limiting fouling speed iscompared to the critical flow velocity required to roll aspherical particle resting on a flat surface and subjectedto a shear flow. Depending on the particle size distributionin the gas flow, fine particles deposited on the HEtube may be removed by the large depositing particles.The removal of fine particles by the large depositingparticles leads to a lower limiting fouling speed than thelimiting fouling speed, which corresponds to the smallestparticle in the gas flow. The influence of the particle sizedistribution on the selection of the limiting fouling speedstill has to be investigated in more detail. To preventfouling, the gas speed of a heat exchanger should belarger than the critical flow velocity that corresponds tothe particle size most likely to stick first and remain onthe HE tube. As the fouling layer starts to build upsintering takes place. Sintering gradually changes thelayer from a loose powdery structure to a more solidstructure strongly adhered to the heat exchanger tube.Sintering makes it difficult afterwards to clean the tubesby soot blowing.4. Concluding discussionAcknowledgementsWe would like to acknowledge the staff members ofthe laboratory of Thermal Engineering at the Universityof Twente for their assistance in preparing and performingthe experiments.

M.S. Abd-Elhady et al. / International Journal of Heat and Mass Transfer 47 (2004) 3943–3955 3955References[1] J. Taborek, T. Aoki, R.B. Ritter, J.W. Palen, J.G.Knudsen, Fouling: the major unresolved problem inheat transfer, Chem. Eng. Process 68 (2) (1972) 59–67.[2] H. M€uller-Steinhagen, F. Reif, N. Epstein, P. Watkinson,Influence of operating conditions on particulate fouling,Can. J. Chem. Eng. 66 (1988) 42–50.[3] J.M. Grillot, J. Icart, Fouling of a cylindrical probeand a finned tube bundle in a diesel exhaust environment,Exp. Thermal Fluid Sci. 14 (1997) 442–454.[4] F.J. Cabrejos, G.F. Klinzing, Incipient motion of solidparticles in horizontal pneumatic conveying, PowderTechnol. 72 (1992) 51–61.[5] M.S. Abd-Elhady, C.C.M. Rindt, J.G. Wijers, A.A. vanSteenhoven, Removal of particles from powdery fouledsurfaces, in: J. Taine (Ed.), Proceedings of the TwelfthInternational Heat Transfer Conference, Grenoble,France, vol. 2, 2002, pp. 687–692.[6] M.C. van Beek, C.C.M. Rindt, J.G. Wijers, A.A. vanSteenhoven, Analysis of fouling in refuse waste incinerators,Heat Transfer Eng. 22 (2001) 22–31.[7] L.N. Rogers, J. Reed, The adhesion of particles undergoingan elastic–plastic impact with a surface, J. Phys. D: Appl.Phys. 17 (1984) 677–689.[8] M. Fichman, D. Pneuli, Sufficient conditions for smallparticles to hold together because of adhesion forces,ASME J. Appl. Mech. 52 (1985) 105–108.[9] E. Achenbach, Distribution of local pressure and skinfriction around a circular cylinder in cross-flow up toRe ¼ 5 10 6 , J. Fluid Mech. 34 (4) (1968) 625–639.[10] F. Zhang, A.A. Busnaina, M.A. Fury, S.Q. Wang, Theremoval of deformed submicron particles from siliconwafers by spin rinse and megasonics, J. Electron. Mater. 29(2000) 199–204.[11] M.C. van Beek, Gas-side fouling in heat-recovery boilers,Ph.D. Thesis, Eindhoven University of Technology, Eindhoven,The Netherlands, 2001.[12] A.L. Robinson, S.G. Buckley, N. Yang, L.L. Baxter,Experimental measurements of the thermal conductivity ofash deposits: Part 2. Effects of sintering and depositmicrostructure, Energy Fuels 15 (2001) 15–84.[13] H.R. Rezaei, R.P. Gupta, G.W. Bryant, J.T. Hart, G.S.Liu, C.W. Bailey, T.F. Wall, S. Miyamae, K. Makino, Y.Endo, Thermal conductivity of coal ash and slags andmodels used, Fuel 79 (2000) 1697–1710.