Physical mechanisms involved in slug transport and pipe blockage during horizontal pneumatic conveying

Powder Technology 262 (2014) 82–95

Contents lists available at ScienceDirect

Powder Technology journal homepage: www.elsevier.com/locate/powtec

Physical mechanisms involved in slug transport and pipe blockage during horizontal pneumatic conveying I. Lecreps a,⁎, O. Orozovic b, T. Erden c, M.G. Jones b, K. Sommer c a b c

TUNRA Bulk Solids Handling Research Associates, NIER, The University of Newcastle, Callaghan 2308, NSW, Australia Centre for Bulk Solids and Particulate Technologies, NIER, The University of Newcastle, Callaghan 2308, NSW, Australia Lehrstuhl für Verfahrenstechnik disperser Systeme, Technische Universität München, 85350 Freising, Germany

a r t i c l e

i n f o

Article history: Received 19 February 2014 Received in revised form 11 April 2014 Accepted 17 April 2014 Available online 26 April 2014 Keywords: Slug Horizontal pneumatic conveying Stresses Porosity Conveying mechanisms Pipe blockage

a b s t r a c t Moving slugs of plastic pellets were investigated in-situ during low velocity pneumatic conveying in horizontal pipelines. Slug characteristics including the profile of pressure, pressure gradient, particle velocity, porosity, radial and wall shear stresses, aspect and behaviour were combined to obtain a complete picture of moving slugs. The objective was to gain unique knowledge on the physical mechanisms involved in slug formation, transport, and decay and the occurrence of pipe blockage. Slugs in both stable and unstable states were analysed. A strong correlation between particle velocity and wall stresses was found, which suggests that the stresses responsible for the high pressure loss characterising slug flow may result mostly from the transfer of particle impulses to the pipe wall. Most slugs were found to be denser at the rear where particle velocity was the highest, thus leading to slug shortening over time. This phenomenon was successfully modelled using both Newton's 2nd law and the ideal gas law and prediction of particle velocity showed good agreement with experimental values. In contrast, other slugs were found to extend due to the particles at the front moving faster than the particles at the rear. Pipe blockage was found to result from insufficient permeation of the slug by the conveying gas, indicating that sufficient material permeability is a condition for slug flow to occur. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Despite increased interest in dense phase pneumatic conveying since the seventies and the development of special dense phase conveying systems, the real establishment of such systems for industrial applications has been somewhat slow. The resistance typically comes from practitioners for whom the random performances of low velocity pneumatic conveying systems result in too higher risk. In fact, even if operating a pneumatic conveying system constitutes a relatively easy task, the design of such a system is often problematic. Since the end of World War II, research teams in both industry and academia have been working actively to assist designers of pneumatic conveying systems by developing design guidelines for the selection of system parameters like pressure, mass flow rate and velocity of both the gas and solid phases. The goal is to furnish equipment manufacturers design tables, diagrams and equations to aid in the design of new conveying systems. While this has been satisfactorily achieved in the field of high velocity pneumatic conveying by integrating friction factors as in the transport of gas alone, the design of low velocity pneumatic conveying systems and particularly slug flow conveying systems still remains a problem. This is because the complex physical mechanisms involved ⁎ Corresponding author. Tel.: +61 2 40339030. E-mail address: [email protected] (I. Lecreps).

http://dx.doi.org/10.1016/j.powtec.2014.04.058 0032-5910/© 2014 Elsevier B.V. All rights reserved.

in the transport of high particle concentrations in a gas phase have still not been fully understood. In fact, in dense phase, the flow phenomena occurring in the pipeline are influenced by not only the gas velocity, solid properties, pipeline direction and configuration and solid feeding devices, but also the particle–particle and particle–wall interactions that are of great importance and should be taken into account. In addition, the transport of solids by a gas stream can cause some unique phenomena that often are not encountered in gas–liquid flows or single-phase flows, like the production of electrostatic charges, which again increases the complexity of the flow phenomena and their description. A frequent approach to describe and systematise slug flow pneumatic conveying consists of extrapolating the physical parameters characterising the conveying process, such as average gas velocity, slug velocity and pressure loss to the behaviour of individual slugs. However, since information such as number and length of slugs is usually unavailable and each individual slug is in a particular state of formation, stability or decay, generally little information can be gained. Therefore, the converse approach in which measurements performed on individual slugs are extrapolated to the entire pipeline has also been applied. This approach presents great advantages in that if sufficient information is available, the actual physical mechanisms leading to slug flow and the resulting pressure loss can be identified and put into equations, which in turn can be used to predict overall slug flow design parameters. Some workers including Ramachandran [1], Konrad

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

[2], Mi [3], Krull [4], Mason [5], and Daoud [6] focused on observing slug flow and investigating experimentally some of the slug characteristics, usually the slug velocity and pressure gradient. While some general knowledge on slug flow could be gained, because usually only one or two characteristics were investigated, the mechanics of slug formation, stability and decay could not be identified. One of the issues is that because slug flow is a dynamic process, significant parameters such as slug porosity and internal stress states are particularly difficult to determine experimentally, in particular because the measurements must be nonintrusive to avoid flow disturbance. As a result, complex measuring devices must be specifically designed or adapted to this application. In addition, the variations of porosity and stresses along a slug or between slugs are generally low and their recording necessitates highly accurate measuring devices, which were not easily available until the last decade. The experimental studies of Niederreiter [7], Pahk [8], Lecreps [9] and Nied [10] in this area are particularly relevant. Technological advances also gave rise to the development of numerical analysis, which are increasingly applied and permit slug flow analysis without the requirement of pneumatic conveying trials. Tsuji [11], Kuang [12], Wensrich [13] and Levy [14] are among the authors who took up the challenge of simulating slug flow pneumatic conveying using diverse approaches such as Discrete Element Methods, Computational Fluid Dynamics or a combination of the two. Numerical simulations have the advantage of identifying some precise physical phenomena that physical measuring devices would only pick up with difficulty. In addition, they can deliver a complete picture of a moving slug with data on local stresses, porosity, velocity, pressure and also shape. However, computer simulations rely on many assumptions and require pre-calibration of the models, which usually necessitate experimental data. In addition, most simulation works consider steady-state slug flow and a short length of pipe. Thus, those results are often more qualitative than quantitative. Only in recent years, simulation results for unstable slug flow became available [15]. This paper deals with detailed in-situ investigations of moving slugs during pneumatic conveying in horizontal pipelines. It addresses both experimental and theoretical investigations performed with the aim of identifying the main physical mechanisms playing a role in the formation, transport and decay of slugs and the occurrence of pipe blockage. In particular, focus has been on the mechanisms driving the flow instabilities and leading to pipe blockage through establishment of relationships between profiles of pressure, porosity, particle velocity and wall stresses along moving slugs. By combining all those characteristics, a unique insight into the physical mechanisms involved in the transport of slugs in horizontal pipes was obtained. Slugs in both steady and unsteady states as well as occurrence of pipe blockage were analysed. 2. Thirty five years of research to understand slug flow 2.1. Flow observation Ramachandran was possibly the first to study the flow of solid–gas mixtures using long transparent pipes to enable flow observation in large pipe diameters [1]. He noticed that the ease of movement is better in the case of coarser particles due to lower compaction of the mass. He also noticed that the material follows different modes of flow along the pipeline and proposed that the increase of material velocity down the pipe may be due to the expansion of the air from higher to lower pressures, which leads to the increase of the size of the interstices between particles, i.e. decrease of the compaction degree which, in turn, leads to the increase of the particle velocity. However, no measurement was performed to verify this premise. When Konrad [2] proposed that the material is conveyed only in the slugs and in the regions just in front of and behind them with the material being picked up from the stationary layer by the moving slug, transported along the pipe and then dropped off the back of the slug to form a stationary layer of the same thickness, he actually suggested that slug flow is no steady-state

83

transport. Also numerical results obtained by Levy [14] indicated that slugs are continuously created and destructed. Further, Kuang traced numerically the process of particle exchange between the settled layer and horizontal moving slugs [12]. He found that the particles in the centre of a settled layer move into the upper part of a slug while the particles initially located in the lower part of a settled layer move into the lower area of the slug. Nevertheless, many workers including Konrad [16] assumed that plugs are like moving packed beds with all particles within each plug fixed relative to each other and moving with the same velocity, even though the transport occurs in a wave motion. 2.2. Particle, slug and gas velocity Tomita investigated slug flow pneumatic conveying in a horizontal pipeline numerically and found that the gas velocity increases preceding the slug arrival. He mentioned that this would explain the jump of particles in front of the slug that has frequently been observed. He also found that the slug velocity is not always constant but changes sinuously [17]. Mi [18] and Krull [4] also measured the slug velocity and both established a linear correlation between slug and air velocity. Klinzing suggested as a rough estimate of slug velocity that it achieves about 70% of the air velocity in horizontal pipes [19]. Kuang [12] reported from numerical analysis that the gas mainly flows in the empty part of the pipe over a settled layer before encountering a slug, then redistributes itself to cover the entire cross-sectional area at the rear of a slug and finally becomes a partial flow again after passing through the slug. When the gas flow rate was low and the particles in a settled layer were stationary, he sometimes observed a backflow of gas inside the settled layer before and after a slug. Recently, Kuang also reported that an increase of the friction coefficient leads to the decrease of particle velocity but increase of solid concentration and pressure drop [15]. 2.3. Slug length Daoud noticed that the plug length decreases with the gas flow rate while for a given mass flow rate, the plug length increases along the pipe. He explained the changes of plug length and velocity along the pipe by establishing relationships based on mass balances at the front face and rear of the plug. In particular, he explained the increase in plug length along the pipe by the velocity difference between the front and rear of the plug [6]. It is however questionable whether those experimental results are indicative of a coming pipe blockage. Hitt also found that the waves increase in length along the pipeline [20]. Mason identified the same phenomenon and observed further over the pipeline short waves close together with a long gap before another series of waves [5]. This implied that the waves increase in length and then break up. Krull suggested that factors such as the pressure gradient and the slug velocity largely influence the length of a slug [4]. The numerical simulations carried out by Levy [14] also revealed that both the shape and the length of the slugs change along the pipe. Recent simulation results by Kuang [15] indicate that the slug length increases and the slug velocity decreases when the friction coefficient is increased. All those results indicate that horizontal slug flow pneumatic conveying is far from being a steady-state process. 2.4. Porosity distribution/permeation through a slug Aziz investigated the pressure loss variation across a plug according to the possibility for the gas to permeate through the plug [22]. He concluded that the pressure drop across the plug varies either linearly if permeation of the transport gas into the plug is allowed or in an exponential fashion if the plug is consolidated at its back and its initial dense state solid packing is maintained. He also concluded that the transport of material is made easier if a certain amount of permeation is possible. Kuang investigated the porosity distribution by means of computer simulation and found that the solid concentration is denser

84

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

in the centre of a slug than close to the wall [12,21]. In addition, he observed a region with high solid concentration at the bottom of a settled layer in front of a slug and suggested that it is caused by the compression of the slug. He also observed that although the distribution of solid concentration in the pipe is not uniform, the average solid concentration of a slug across the pipe cross-section fluctuates around a constant that is lower than the bulk density. Lecreps [23] also noticed that slugs have density lower than the bulk density when investigating the porosity of horizontal moving slugs experimentally. Using both an indirect method based on the Ergun equation and a direct method involving a slug-catcher, she showed that all slugs are slightly fluidised with a porosity that is independent of the supply air velocity. In addition, she showed that slugs are homogeneously fluidised over the pipe height. While applying ECT (Electrical Capacitance Tomography) to analyse horizontally moving slugs, Nied [10] also stated that slugs are slightly fluidised entities with highest particle concentration at the pipe wall and a porosity that increases towards the centre of the pipe. An increasing air supply velocity seemed to enhance this effect. Those recent findings contradict Kuang's numerical results [12]. 2.5. Stress states All models developed to predict the pressure loss during slug flow pneumatic conveying involve the determination of the stresses induced by moving slugs. Models based on bulk solid mechanics also involve the so-called stress transmission coefficient, which defines the fraction of axial stress transmitted into radial direction, as per Janssen's theory [24]. The determination of the stresses and stress transmission coefficient generally occurred theoretically or semi-empirically, more rarely fully experimentally. Among others, Yi developed an apparatus for direct measurement of both the acting axial stress and resulting radial stress in which a bulk solid sample retained between two porous plates, thus providing a range of compression states, was pulled upward along the test pipe [25]. The measurements were carried out on a very short, non-aerated slug so that the results are unlikely to reflect the actual mechanisms occurring in pneumatic conveying. Mi determined values of wall pressure by subtracting the static air pressure from the total air pressure that he obtained experimentally. He reported that the wall pressure is higher at the bottom of the pipe than at the top due to the weight of the slug [26]. He also calculated semi-empirical values for the stress transmission coefficient and suggested that the interparticle stresses in a horizontal moving slug are in the active state. However, the method used was later proved to be false [25]. Vasquez investigated the wall friction during slug flow pneumatic conveying by means of pressure transducers for both the total pressure and air pressure and used a bending type transducer to measure the total radial stress [27]. He found values for the stress transmission coefficient lower or higher than 1 according to the material investigated, indicating that either passive or active stress case takes place. Krull [4] criticised both the measurement technique and results obtained by Vasquez and developed a test rig described as a large-scale static pressure transducer that permits direct measurement of the radial stress within a slug of granular material without obstructing the flow. Since this test chamber provided an averaged radial stress around the circumference of the slug, it was particularly suitable to measure the radial stress by vertical pneumatic conveying where the stress can be assumed to be homogenously distributed around the pipe circumference. Krull reported that a linear relationship exists between the pressure gradient through a slug and the radial wall pressure measured. He established a linear decreasing correlation between the stress transmission coefficient and the ratio between slug velocity and air mass flow rate. Niederreiter [7] and later Lecreps [28] measured both the radial stress and wall shear stress induced by moving slugs at the pipe wall by using a small measuring plate simulating a piece of the pipeline wall which was connected to two miniature force sensors. Their results

are fundamentally innovative in that they concluded that transfer of axial stress into radial direction does not take place during slug flow pneumatic conveying and the high values of stresses detected at the pipe wall have a different origin. Instead of using traditional bulk solid mechanics, they proposed to apply a modified form of the kinetic theory to quantify the impulse transfer of particles to the pipe wall. Lecreps stated that for all supply air velocities, the stresses are greater at the side of the pipe, followed by the bottom and finally the pipe top. The stresses were found to be about constant in the area of stable slug flow and then increased when slug flow transitioned to unstable flow.

3. Experimental testing 3.1. Test material and test rig 3.1.1. Test material Conveying experiments were carried out with polypropylene (PP) granules. The granules are white and have a regular shape and a form between lentils and spheres. Table 1 lists the main physical properties of the material tested. The test material is non-cohesive and has a bulk porosity of 0.38. To improve the optical tracing of the slugs, black particles were added to the white polypropylene pellets in a ratio of 1/100 to play the role of tracers without changing the characteristics of the product transported.

3.1.2. Pneumatic conveying rig Conveying trials were performed in an i.d. 80 mm industrial scale pilot plant of 35.3 m length from product feeding to the last bend after which the material conveyed enters the last vertical section downward to the storage silo (Fig. 1). The product transport took place in-batch using a dual silos system. The supply of the air mass flow rate was controlled using 4 de Laval nozzles, which ensured a constant air mass flow rate independent of the back pressure changes. Stable slug flow was achieved with supply air velocities between 6.8 and 7.5 m/s. Higher conveying velocities led to occasional self-released pipeline blockages. At the end of the first horizontal section, different measurement devices were integrated in the pipeline to investigate individual slugs without disturbing the conveying process. These elements are identified by 2 to 5 in Fig. 1. Element 5 is a slug-catcher that was part of a separate study focusing on slug porosity published in [23]. The other elements relevant to this study are described in Section 3.2.

3.2. Investigations on individual slugs 3.2.1. Particle and slug velocity, and slug aspect The particle velocity was determined by following the horizontal motion of black tracer particles recorded by a CCD camera (Element 4 in Fig. 1) with a frequency of 30 Hz during passage in transparent plastic pipes (Element 2). The particle velocity was measured at the pipe midheight. The same CCD camera provided information on slug aspect, including shape of individual slugs, in particular front and rear and height of the stationary layer between two slugs.

Table 1 Physical characteristics of the test material. Test material

Polypropylene

Aerodynamic equivalent diameter (AED) Equivalent spherical diameter (ESD) Particle density Bulk density Wall friction angle on stainless steel

3.0 mm 4.3 mm 889 kg/m3 553 kg/m3 9.7° ± 0.7°

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

85

13.7 m

P6

6.5 m

P5

P4

P3 5

3

2.00 m

4

P2

5.08 m

P1

1

2

Air

15.0 m Fig. 1. Industrial-scale pilot plant in stainless steel for conveying trials.

3.2.2. Pressure, wall shear stress and normal stress within a slug Niederreiter [7] developed and constructed a measurement device (Element 3 in Fig. 1) that allows the in-situ and simultaneous detection of pressure within a slug and stresses that a slug induces at the pipe wall. The measurement probe has the particularity to allow both the radial and wall shear stresses to be measured simultaneously and locally around the pipeline circumference without disturbing the conveying process. For this purpose, a short pipeline section has been instrumented with both pressure and force sensors (Fig. 2). Two piezoelectric force sensors (type 208C01, PCB Piezotronics Inc., USA) are mechanically arranged perpendicular to each other and connected to a measurement plate of 970 mm2 surface area, which simulates a piece of pipeline wall and transfers the stresses to the force sensors. In addition, the probe includes several miniature pressure sensors. One of the sensors is located in the measuring chamber and aims to control that pressure equilibrium between the two sides of the measuring plate (conveying section and measurement chamber) is well provided by the small holes located on each side of the plate. Further miniature pressure sensors, which measure the pressure inside a slug over a distance of 35 mm each, are used to determine the slug porosity as described in Section 3.2.3. The design of the probe and the calibration method are detailed in [28].

3.2.3. Slug porosity The slug porosity was determined indirectly from pressure loss and particle velocity measurements by applying the principles of gas flow through porous columns of bulk material. The Ergun equation, which correlates pressure loss, porosity and relative velocity between gas and particles, was found to be suitable to determine the profile of porosity along moving slugs [23]. Hence, by applying Eq. (1), the porosity within a slug can be determined by means of pressure loss measurements over a given distance where the relative velocity between air and particles is known. The determination procedure is summarised in Fig. 3. Note that a detailed description of the method and a discussion of its applicability can be found in [23,28]. ð1−ε Þ  η f i ð1−ε Þ  ρ f i 2 ΔP i ¼ 150   vslipi þ 1:75   vslipi ΔL ε  dp ε  dp 2

3

2

3

ð1Þ

where ΔP is the pressure loss, ΔL is the bed length, ε is the bed porosity, ηf is the fluid dynamic viscosity, dp is the particle diameter, vslip is the relative velocity between fluid and particles, and ρf is the fluid density. The pressure loss was measured using high accuracy miniature piezoresistive pressure sensors, model XTM-190 from the firm Kulite,

Pressure sensor Cap Piezoelectric force sensors

Connecting strap Pipe element Seal

Conveying pipe Pressure sensors Fig. 2. Image and exploded assembly of the probe for simultaneous detection of wall shear stress, normal stress and pressure inside a slug [7].

86

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

Pressure loss

P

P2

P1

vf (T0, P1)

vf (T0, Patm)

Ergun Equation = f(vslip, P)

vp

Flow direction vslip (T0, P1) Fig. 3. Schematic procedure for porosity determination based on the semi-empirical equation of Ergun [23].

positioned only 37.5 mm apart so that the air expansion resulting from the pressure decrease could be neglected. The pressure was sampled with a frequency of 10,000 Hz. The procedure to determine the particle/slug velocity was presented in Section 3.2.1. The air velocity and resulting slip velocity were calculated for each increment i by applying the ideal gas law under assumption of isotherm conditions.

4.2. Profile of particle velocity

Fig. 4 shows that both the radial and wall shear stresses induced by a slug at the pipe side (a and b) and bottom (c and d) follow the exact trend of the particle velocity. In addition, the amplitude of the stresses and particle velocity are correlated, i.e. low stress values are obtained when the particle velocity is low and high stress values are linked to

As illustrated in Fig. 4, in most cases, the particle velocity was found to be greater towards the back of the slug than at the front (additional results can be found in [23,28]). This implies that the particles located at the back tend to move closer to the particles in front of them. This phenomenon is caused by the existence of a decreasing pressure

3500 3000

Radial Stress [Pa] Wall Shear Stress [Pa] Particle Velocity [m/s]

a)

3.5 3.0

2000

2.0

1500

1.5

1000

1.0

500

0.5

0 0.0

0.0 0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

3500 3000

Radial Stress [Pa] Wall Shear Stress [Pa] Particle Velocity [m/s]

3.0

2.5

2000

2.0

1500

1.5

1000

1.0

500

0.5

0 0.0

0.0 0.2

0.4

0.6

2.5

2000

2.0

1500

1.5

1000

1.0

500

0.5

0.0 0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2

Time [s]

0.8

1.0

1.2

1.4

1.6

1.8

3500 3000

Radial Stress [Pa] Wall Shear Stress [Pa] Particle Velocity [m/s]

d)

3.5 3.0

2500

2.5

2000

2.0

1500

1.5

1000

1.0

500

0.5

0 0.0

0.0 0.1

0.2

0.3

0.4

0.5

0.6

Time [s]

Fig. 4. Correlation between particle velocity and radial and wall shear stresses along individual slugs.

0.7

0.8

0.9

1.0

Particle Velocity [m/s]

2500

Radial and Wall Shear Stress [Pa]

c) Particle Velocity [m/s]

Radial and Wall Shear Stress [Pa]

3000

3.0

Time [s] 3.5

Radial Stress [Pa] Wall Shear Stress [Pa] Particle Velocity [m/s]

3.5

2500

Time [s] 3500

b)

Particle Velocity [m/s]

2.5

Particle Velocity [m/s]

2500

Radial and Wall Shear Stress [Pa]

4.1. Correlation between particle velocity, radial and wall shear stresses

Radial and Wall Shear Stress [Pa]

4. Results and discussion

particles which move quicker. This behaviour was observed on each slug without exception. Those results indicate that a strong functional relationship exists between particle velocity and radial and wall shear stresses. This, in turn, suggests that the stochastic agitation of the particles, which transfer their momentum to the wall during impact, may be at the origin of the largest fraction of the wall stresses induced by moving slugs and the resulting significant pressure gradient characterising slug flow pneumatic conveying.

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

Eq. (5) describes the pressure loss as a function of particle mass and velocity. On the other hand, the pressure loss can also be described by applying the ideal gas law. In that case, the pressure P in the slice is a function of the interstitial gas velocity and is given by:

gradient from the front towards the rear. To show the relationship between pressure gradient and particle velocity, a model was built that involves both Newtown's 2nd law and the ideal gas law. Using this model, the profile of particle velocity was then analysed considering either a given time and spatial increments x or a given spatial location and time increments t. In other words, the particle velocity was analysed either along a slug at a given instant or at a given position along the pipe over time. Applying the developed model, the particle velocity was predicted and compared with experimental values.

P ðxÞ ¼

dvs ¼ ½P ðxÞ−P ðx þ ΔxÞ  A dx



 ˙ RT vðx þ ΔxÞ  vðxÞ 2  m 2 :  vs ¼ vðx þ ΔxÞ−vðxÞ ρp  A  ε  ð1−ε Þ

vs

4.2.2. Profile of particle velocity along a slug Let us consider two consecutive slices as shown in Fig. 6 with pressure and interstitial velocity defined  at the extremity of each slice. The  xþΔxÞvðxÞ can then be defined combined interstitial velocity term vvððxþΔx Þ−vðxÞ for each slice. The right hand side of Eq. (8) is comprised of air, particle and slice properties which reasonably all can be assumed to be constant values. Therefore, the left hand side of Eq. (8) also equals a constant and: 

ð3Þ

0

Z vs  dvs ¼

P1 P0

−

ΔP  dx Δx

v1  v0 v1 −v0



   2  vs ¼ Cte:

ð9Þ

Based on Eq. (9), for a given time, if the combined velocity term  increases along the slug from the rear towards the front, then

v1 v0 v1 −v0

the velocity of the slice vs, i.e. the particle velocity is greater at the rear than at the front. Thus, an analysis of how the combined velocity term   v1 v0 v1 −v0 is varying along the slug will give information on the velocity

ð4Þ

of the particles at various locations along the slug. As the absolute pressure decreases along the slug, from Eq. (6) it can be seen that the gas velocity must increase to maintain the mass flow, hence v2 N v1 N v0. Therefore the numerator of the combined interstitial

which results in: 1 2  ρ  ð1−ε Þ  vs ¼ P 0 −P 1 ¼ ΔP 1 : 2 p

ð5Þ

Core

x

Back

P(x)

ð8Þ

Eq. (8) can be used to investigate the profile of slice velocity, i.e. particle velocity for instance at a given time and various locations of x along a slug or over a time t at a given location.

where ρp is the particle density. Substituting Eq. (3) into Eq. (2), simplifying, rearranging and setting integral bounds give: Z

ð7Þ

Setting ΔP1 = ΔP2 and hence Eq. (5) equals Eq. (7) and rearranging the variable velocity terms to one side result in:

where ms is the mass of a slice, vs is the velocity of the slice, P(x) is the pressure at a given location x, ΔP is the pressure loss over that slice and A is the pipe cross-sectional area. Note that the velocity of all particles in a given slice is considered as a constant. Assuming a constant porosity over the slice of length Δx, the mass of the slice can be given by:

ρp  ð1−ε Þ 

ð6Þ

  ˙ RT 1 1 m − : ΔP 2 ðxÞ ¼ A  ε vðxÞ vðx þ ΔxÞ

ð2Þ

dms ¼ ρp  ð1−ε Þ  A  dx

˙ RT m vðxÞ  A  ε

where m˙ is the gas mass flow rate, R is the gas constant in J/mol·K, T is the temperature in K, and v is the superficial gas velocity. And hence the pressure drop over a slice is:

4.2.1. Particle velocity model The fluctuations of particle velocity within a slug were described using both Newtown's 2nd law to describe the pressure loss as a function of particle mass and velocity and the ideal gas law as a second expression for the pressure loss. Combining those two equations allows definition of a relationship between existing pressure gradient at a given time or location and particle velocity. The model development is as follows. We consider the core of the slug where the entire cross-section of the pipe is filled with material. In particular, we consider one slug slice of length Δx (Fig. 5). By assuming that the net force is a measured pressure drop over a slice then by Newton's 2nd law: dms  vs 

87

dms

P(x+ x)

x x+ x P0 P1

Flow direction Fig. 5. Slice of a horizontal slug for force balance.

Front

88

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

when stable slug flow occurs. This is due to fluctuations in the total mass of material and total slug length in the pipeline over time.

Flow Direction Fig. 6. Consecutive slices at a fixed time with defined pressure and interstitial velocity and combined interstitial velocity term.

velocity term is higher towards the front, i.e. v1 ⋅ v2 N v1 ⋅ v0. To see how the denominator is behaving, let us consider one big slice that starts at Position 0 and finishes at Position 2. The combined interstitial velocity v0 . We know for the second slice that the combined for this slice is vv22−v 0 v1 v1 v0 . As v2 N v1 N v0, it can be seen that vv22−v N vv22−v . velocity term is vv22−v 1 1 0 Hence, it can also be concluded that: v2  v1 v1  v0 N : v2 −v1 v1 −v0

ð10Þ

Combining Eqs. (9) and (10) results in Eq. (11), which indicates that at a given time, the particles of a given slice i are faster than the particles of the slice i + 1 in front of them. This is true provided that the average porosity values in two adjacent slices are equal. vsi Nvsiþ1

ð11Þ

Therefore, in the case of a slug where the average cross-sectional porosity of a given slice is equal to the porosity of the adjacent slice, the particles located at the slug rear will tend to move closer to the particles located in front of them as a result of the pressure which decreases along the pipe. 4.2.3. Profile of particle velocity over time Besides the profile of particle velocity along a slug, it is also of interest to understand the particle velocity behaviour over the time, i.e. at a given location but different instants t. Let us consider a general slice xþΔxÞvðxÞ with a combined interstitial velocity term given by vvððxþΔx Þ−vðxÞ which now varies over time. Applying Eq. (6), the following can be shown: vðx þ ΔxÞ  vðxÞ ¼

  ˙ RT 2 1 m  P ðxÞ  P ðx þ ΔxÞ Aε

˙ RT 1 1 m − : vðx þ ΔxÞ−vðxÞ ¼  A  ε P ðx þ ΔxÞ P ðxÞ

ð12Þ

ð13Þ

Dividing Eq. (12) by Eq. (13) and simplifying, the general form of the combined interstitial velocity term is obtained:   ˙ RT vðx þ ΔxÞ  vðxÞ m 1 ¼ :  vðx þ ΔxÞ−vðxÞ P ðxÞ−P ðx þ ΔxÞ Aε

ð14Þ

Eq. (14) shows that the combined interstitial velocity term is inversely proportional to the pressure drop P(x) − P(x + Δx) over the slice. When these effects are considered with respect to Eq. (8) and multiple short time increments, Eq. (14) indicates that an increasing time varying pressure drop will increase the slice velocity, and vice versa. This knowledge is of particular relevance as the pressure at a given point along a pneumatic conveying pipeline fluctuates with time, even

4.2.4. Comparison of predicted and experimental particle velocities Fig. 7 illustrates the particle velocity along a slug calculated from local pressure and pressure loss measurements by applying Eq. (9) and a comparison with the particle velocity measured by means of a camera and tracer particles (see Section 3.2.1). Note that the step-like pattern observed on the experimental velocity curve results from the fact that the velocity was sampled at a rate of 30 Hz (camera limitation) while all other data was sampled at 10,000 Hz. Eq. (9) was found to not only describe the profile of particle velocity along a slug with reasonable accuracy but also the value of the particle velocity in each slice could be satisfactorily predicted. It was found that stable slugs have a nearly linear profile of particle velocity with particles at the rear having higher velocity than the particles at the front. Note that a slug was considered as stable when both picture and signal analyses showed constant length and velocity of the slug when passing through the measuring section. In some cases, the particle velocity was found to be fluctuating along the slug length and/or the particles were found to move quicker at the front than at the rear. Both the prediction model and the measurements were able to detect those fluctuations. Examples are given in Fig. 7a and f. Those slugs are believed to be in an unstable state, as will be discussed in Section 4.3.2. Those results show the applicability of Eq. (9) to predict the particle velocity along a slug based on pressure loss measurements. Experimental values of pressure loss comprise the effects of numerous physical characteristics including interactions between particles and wall, particle–particle interactions, porosity, and fluid and solid characteristics. This can be considered as either a strength or weakness of the model. A strength because all characteristics are included into one single parameter, which is the pressure loss, or weakness because the effects of each individual characteristic on the resulting particle velocity cannot be identified. 4.3. Physical insight into moving slugs Individual slugs were investigated with respect to internal stress state, porosity, pressure loss, particle velocity and physical aspect. All those characteristics were combined to gain unique knowledge on the physical mechanisms involved in slug transport in horizontal pipelines. The analysis of slugs conveyed with supply air velocities from 6.8 to 8.5 m/s permitted identification of similarities in the transport behaviour and physical characteristics, which were independent of the gas velocity. Based on a systematic analysis of all characteristics, the slugs could be classified into two categories, namely stable slugs and unstable slugs. Slugs in unstable state were found to be either in a process of shortening or extension. This section presents the physical characteristics of those three classes by means of examples. 4.3.1. Properties of moving slugs in stable state Fig. 8 shows the pressure, stresses, pressure loss and particle velocity measured on a slug conveyed with vf = 6.8 m/s and the porosity calculated by applying the Ergun equation. The stresses were measured at the pipe side. When passing through the measurement section, this slug was found to be representative of typical stable slugs, which are characterised by the following. First, the comparison between pictures, pressure and stress signals indicated that the slug length remained exactly the same over the 0.5 m separating the camera and the stress measurement plate. In fact, the slug needed 1.01 s to pass both in front of the camera and on the stress measurement plate. Provided that the slug velocity was constant over 0.5 m, this indicates constant slug length over the same distance. Second, both the radial and wall shear stresses tend to increase over the slug length. On this particular slug, the stresses showed local minima, particularly at t = 0.64 s when the pressure loss also showed a significant and rapid decrease. At this

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

Fig. 7. Comparison of experimental and predicted particle velocities.

89

90

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

40 30

2000 20 1000 10

Front face Pressure [kPa]

Radial and Wall Shear Stress [Pa]

a)

Radial stress Wall shear stress Pressure 1 Pressure 2

3000

0.2

0.4

0.6

0.8

1.0

1.2

t = 0.200 s t = 0.233 s

0 1.6

1.4

FRONT t = 0.266 s

Time [s]

1000 800

2000 600 400

1000

200 0

0

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Pressure Loss [Pa/35 mm]

1200

b)

Radial stress Wall shear stress Pressure loss

3000

Radial and Wall Shear Stress [Pa]

t = 0.166 s

0 0

1.6

1.2 1.0

3.0 2.5

0.8

2.0 0.6 1.5 0.4

1.0

0.2

Porosity Particle velocity

0.0 0

0.2

0.4

0.6

0.8

1.0

1.2

Time [s]

1.4

0.5 0.0 1.6

t = 0.333 s

Rear t = 1.200 s t = 1.233 s

REAR t = 1.300 s

3.5

Particle Velocity [m/s]

c)

t = 0.300 s

t = 1.266 s

Time [s]

Porosity [-]

t = 0.133 s

t = 1.333 s t = 1.366 s t = 1.400 s t = 1.433 s t = 1.466 s

Fig. 8. Physical characteristics of a stable slug (vf = 6.8 m/s) — Stresses measured at the pipe side.

position over the slug length, the particles moved with the same velocity as the particles directly in front of and behind them, as shown in Fig. 8c. However, the slug locally displayed higher porosity. Many air cavities were present at the top of the slug. Fig. 9 shows the air cavity which generated the local porosity increase at t = 0.64 s. Fig. 8a indicates that the presence of air cavities leads to local decrease of both stresses. Third, the particle velocity showed a decreasing trend from the slug rear towards the front, i.e. the particles located in a given slug slice moved faster than the particle in front of them, as is expected along a stable slug as a result of the decreasing pressure gradient (see Section 4.2). Note that other authors noticed the lower particle velocity at the front face and explained it with the fact that particles are picked up from the stationary layer by the front face and have to be accelerated to the slug velocity. Finally, while the particle velocity gradient showed a negative trend in the direction of the flow, the porosity showed the converse trend, i.e. the rear of the slug was denser than the front.

t = 0.64 s

Fig. 9. Air cavity at the top of the slug illustrated in Fig. 8.

The combination of decreasing stresses, decreasing particle velocity and increasing porosity towards the front of a slug was found to be representative of most conveying slugs. In other words, the stresses induced by slugs are generally greater at the rear where particles move faster and the slug is denser. 4.3.2. Properties of moving slugs in unstable state 4.3.2.1. Insight into a shortening slug. Fig. 10 illustrates the behaviour of another slug conveyed with vf = 6.8 m/s. In this case, the stresses were measured at the top of the pipeline. The curves illustrating the pressure loss and the stresses follow the exact same trend and are typical of a slug in stable state with a linear decrease of both stresses and pressure gradient from the slug rear towards the front. Fig. 10b shows that as soon as the pipe cross-section is entirely filled with particles, the pressure loss increases significantly and significant stresses are detected at the top of the pipeline. Similarly, a sharp decrease of the signal is also detected as soon as the slug rear leaves the measurement area, i.e. the pipe is no longer filled up to the top. As usually observed when only a small fraction of the pipe is filled with stationary particles before a slug arrives, a large amount of suspended particles flying with high velocity over the stationary layer preceded the slug front itself (photos marked in red in Fig. 10). Those suspended particles have been picked up by the slug front where particularly turbulent flow conditions prevailed and were transported faster

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

a)

Radial stress Wall shear stress Pressure 1 Pressure 2

1500

35 30

1000 25 500

20

0

Pressure [kPa]

Radial and Wall Shear Stress [Pa]

Front face

40

2000

t = 0.066 s

t = 0.100 s

t = 0.133 s

15 t = 0.166 s

10 0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Time [s]

t = 0.200 s

1000

b)

Radial stress Wall shear stress Pressure loss

1500

800

1000

600

500

400

0

200 0 0

0.2

0.4

0.6

0.8

1.0

1.2

Pressure Loss [Pa/35 mm]

2000

Radial and Wall Shear Stress [Pa]

91

t = 0.233 s

t = 0.266 s FRONT t = 0.300 s

t = 0.333 s

t = 0.366 s

1.4

Time [s] t = 0.400 s

1.2

Porosity [-]

2.5

0.8

2.0

0.6

1.5

0.4

1.0

0.2 0.0

Porosity Particle velocity

0

0.2

0.4

0.6

0.8

1.0

Time [s]

1.2

1.4

0.5

Particle Velocity [m/s]

c)

1.0

3.0

Rear REAR t = 1.066 s

t = 1.100 s

t = 1.133 s

0.0 t = 1.166 s

Fig. 10. Physical characteristics of a shortening slug (vf = 6.8 m/s) — Stresses measured at the pipe side. (For interpretation of the references to colour in this figure, the reader is referred to the web version of this article.)

than the rest of the slug in the form of isolated particles less prone to friction forces. The turbulent flow conditions prevailing at the slug front face facilitated the lifting of particles from the stationary layer. In fact, Tomita [17] noticed an increase in gas velocity preceding the slug arrival and suggested that it would explain the jump of particles frequently observed in front of a slug. In addition, the shear force that flying particles generated at the surface of the stationary layer could have initiated the particle lifting. This shear force results from the momentum exchange of particles hitting the surface of the stationary layer. Finally, the backflow of gas that Kuang [12] sometimes observed inside the settled layer before and after a slug conveyed with very low velocity may play a significant role in the initiation of the particle lifting. Despite their high velocity, i.e. energy, those flying particles did not generate any stress that could be detected at the top of the pipe where the measurement plate was positioned. In addition, those particles did not induce any significant pressure loss, which suggests that the suspension is very light with particles moving in a stream with high velocity. Note that the oscillations of the wall shear stress signal in front of the slug are caused by the measurement plate, which oscillates with its natural frequency. As observed on most slugs, the particles located in a given slug slice moved faster than the particles in front of them. In other words, the closer the particles are to the rear of the slug, the faster they move. This has a consequence that during transport, the particles located at

the rear tend to move closer to the particles in front of them, which, over time, results in a tendency for the slug porosity to decrease and for the slug length to reduce. This phenomenon could be well observed by combining information gained from the recorded pictures and stress signals. In fact, the core of the slug needed 0.76 s to pass in front of the camera but only 0.52 s to pass over the stress measurement plate located 50 cm downstream. This suggests that over the small distance between the camera and the measurement plate, the velocity of the slug increased and/or the slug became shorter by losing material or compacting itself. Further analysis including the gas pockets, height of particle layer between slugs and properties of the previous and succeeding slugs revealed that the mean velocity of the slug remained constant but its length changed due to compression. The results are shown diagrammatically in Fig. 11. This finding emphasizes the unstable character of slug flow. Due to the existing pressure gradient along a slug, the particles located at the rear of a slug will tend to move faster than the particles in front of them. As a result, over time, they will move closer to those particles, the slug porosity will tend to decrease and the slug will become shorter. This state is continuously changing and hence is somewhat unstable. Changes in porosity will affect the gas permeability, gas velocity, pressure gradient and resulting conveying behaviour, which makes prediction of slug flow parameters a difficult challenge. Previous results

92

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

0

2

4

6

8

t = 0.76 s

t = 4.33 s

t = 4.56 s

10

Time [s]

t = 3.16 s

t = 3.16 s

t = 0.52 s

Fig. 11. Scheme illustrating the compression of the slug (presented in Fig. 10) occurring over time from the rear towards the front end.

4.3.2.2. Insight into an extending slug. Fig. 12 shows the characteristics of a slug conveyed with vf = 8.5 m/s. The stresses were measured at the pipe side. Although the transport still took place in the slug flow

Front face 2500 2000

a)

Radial stress Wall shear stress Pressure 1 Pressure 2

60 t = 0.100 s

55

1500 50 1000 45 500

Pressure [kPa]

Radial and Wall ShearStress [Pa]

showed that slugs are slightly fluidised entities and this fluidisation state is a requirement for conveying. This sets one of the major limitations of slug flow pneumatic conveying.

0

t = 0.200 s t = 0.233 s

35 0

0.2

0.4

0.6

0.8

1.0

1.2

1.4 t = 0.266 s

i

2500 2000

b) 600

1500 1000

400

500 200 0 0

0

0.2

0.4

0.6

t = 0.300 s

800

Radial stress Wall shear stress Pressure loss

0.8

1.0

1.2

Pressure Loss [Pa/35mm]

Radial and Wall ShearStress [Pa]

t = 0.166 s

40

Time [s]

1st FRONT t = 0.366 s t = 0.400 s t = 0.433 s

2nd FRONT t = 0.500 s

1.2

3.5 3.0 2.5

0.8

2.0 0.6 1.5 0.4

1.0

0.2

Porosity Particle velocity

0.5

Particle Velocity [m/s]

c)

1.0

0.0 0

t = 0.333 s

t = 0.466 s

1.4

Time [s]

Porosity [-]

t = 0.133 s

t = 0.533 s

Rear t = 0.800 s t = 0.833 s REAR t = 0.866 s

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

t = 0.900 s

Time [s] t = 0.933 s t = 0.966 s t = 1.000 s Fig. 12. Physical characteristics of an extending slug (vf = 8.5 m/s) — Stresses measured at the pipe side.

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

mode, the first flow instabilities characterising the process suggested that the operations were performed close to the lower boundary of the so-called transition zone. The signals presented in Fig. 12 correspond to the passage of two slugs, which however can be considered as one and the same slug since no stationary layer separated both parts. The first part of the slug filled the entire cross-section over a very short distance only. By entering the pipe cross-section where the stress measurement plate was located, the slug's 1st front generated an increase of the stresses. This front was made of particles moving faster than the particles composing the second front. The 2nd front can be recognised through the rapid increase of both the stresses and pressure loss at t = 0.50 s. Although the images show that the particles of the preslug filled the pipe cross-section completely, the elevated Ergun porosity of about 0.75 indicates high degree of fluidisation with particles flowing through the cross-section with high velocity. Contrary to the two slugs presented above, the core of this slug displayed stress values that remained globally on the same level over the entire slug length. However, the stress signals showed significant oscillations, especially the radial stress, which oscillated around a value of 1500 Pa (Fig. 12b). Moreover, in contrast to most slugs where the particles located at the rear moved faster than the particles in front of them, this slug displayed a positive particle velocity gradient from the rear towards the front. The particles located at the front end moved over twice as fast as the particles at the back with a velocity over 3 m/s. Furthermore, instead of the usual decrease in density from the rear towards the front, this slug displayed porosity that remained constant over the length of the main slug and increased slightly at the slug rear. Further analysis revealed that the structure of the slug changed over the short distance separating the camera from the stress measurement plate. In fact, the slug needed 0.50 s to pass in front of the camera but 0.77 s to pass the measurement plate. Since the time separating the entering of the two front faces in the camera objective and later onto the measurement plate remained unchanged, it can be concluded that the slug velocity remained constant. This implies that the slug increased in length, either by extending itself or by picking up more material than it deposited. Fig. 13 shows diagrammatically the process of extension observed on the slug illustrated in Fig. 12. The theory of extending length is supported by the positive velocity gradient from the rear towards the front and the presence of a pre-slug. In fact, the closer the particles are to the front face, the faster they move. As a result, the slug tends to extend from the rear towards the front. While the front of the main slug displayed Ergun porosity values around 0.6, which is a value that usually encountered at the front of the slugs, the rear displayed a higher porosity. A stable moving slug is characterised by a higher density at the rear than at the front, with values at the rear usually close to the bulk density. In this case, the slug is no longer in a stable state. Over time, each particle tends to gain some space on the particles following it. The particles located at the rear tend to catch up with particles moving in front of them so that the usual increasing porosity gradient from rear to front slowly disappears and the slug extends from the rear towards the front. If particles at the front face move significantly faster, those particles may part from the rest of the slug, as observed in Fig. 12 and form a short slug on their own. This slug then may grow further by picking up more material than it deposits.

0

1

2 t = 2.76 s

t = 2.49 s

93

4.4. Physical mechanisms involved in pipe blockage and release The transport of polypropylene pellets in the form of slugs occurs problem-free as long as operations take place far enough from the lower and upper boundaries of slug flow. In close proximity to the lower boundary, the flow steadily slows down until slugs remain motionless. In close proximity to the upper boundary of slug flow, the flow undergoes instabilities caused by the high pressure fluctuations due to the alternation between slug flow and strand flow. Those instabilities easily lead to pipeline blockage. However, during the experimental investigations, the occurring pipe blockages cleared themselves without further intervention. It should be pointed out that the occasional blockage of the pipeline occurred solely when operations were performed close to the upper boundary of slug flow (vf = 8.5 m/s). Fig. 14 illustrates the results obtained during the investigation of a slug that remained temporary blocked in the pressure and stress measurement probe. The stress measurement plate was located at the pipe side. The arrival of the front face was accompanied by the usual increase of pressure, pressure loss and stresses. Without showing any changes in its velocity, the slug front face picked up the particles of the stationary layer. The front part of the slug denoted by A displays a similar behaviour as stable slugs: The stresses and the particle velocity increase over the slug length whereas the porosity slowly decreases. From t = 1.17 s, the slug no longer moved constantly forward but displayed a pulsative movement. The blockage was due to the compressive state of the slug, which no longer permitted air percolation. Until air percolation through the slug improved, the slug transport was stopped and the upstream pressure increased. Once the pressure was high enough to overcome the friction forces at the pipe wall, the entire slug moved forward by a small distance. During this movement, particles rearranged themselves but the rapid decrease of pressure behind the slug led to a new blockage. The sudden motion of the slug was accompanied each time by a significant decrease of the stresses at the pipeline wall. The stresses increased again as soon as a new blockage occurred. This process repeated itself until the particle rearrangement was optimal, i.e. the porosity reached a critical value to allow the transport to continue. The oscillations on the porosity signal in area B are similar to the oscillations on the stress and pressure curves. This is indicative of the particles' rearrangement. The slug rear was locally compacted, as indicated on the porosity curve. This local compressive state led to the release of particles in the direction opposite to the flow (marked on the pictures by red circles). The particle release allowed better percolation of the air through the slug. Afterwards, the slug, which was almost motionless, moved abruptly forward. Those results support the hypothesis that a certain degree of permeation and thus particle fluidization is mandatory for stable slug flow pneumatic conveying to occur. 5. Conclusions The results presented are part of a project aiming to investigate insitu moving slugs in order to identify, measure and comprehend the mechanisms involved in both the formation, stability and decay of slugs and pipe blockage. The understanding and modelling of the physical phenomena at slug scale will lead to a better understanding of the conveying process over the entire pipeline. Recent improvement in

3 t = 0.50 s

t = 0.77 s

4

5

6

t = 2.19 s

t = 2.19 s

Fig. 13. Scheme illustrating the extension of the slug (presented in Fig. 12) from the rear towards the front.

Time [s]

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

10000

a)

Radial stress Wall shear stress Pressure 1 Pressure 2

8000

60

6000 40

4000 2000

20

A

0

Front face

80

Pressure [kPa]

Radial and Wall ShearStress [Pa]

94

B

Pressure Loss [Pa/35 mm]

t = 0.233 s t = 0.266 s t = 0.300 s

0 0

0.2

0.4 0.6 0.8 1.0 1.2 1.4

1.6

1.8 2.0 2.2 2.4

t = 0.333 s

Time [s] 1200

t = 0.366 s

b)

1000

FRONT t = 0.400 s

800 600

t = 0.433 s

400

Rear

200 0

t = 2.066 s

0

0.2

0.4

0.6 0.8

1.0 1.2 1.4 1.6

1.8 2.0 2.2 2.4 t = 2.100 s

Time [s] 1.2

2.5

c)

t = 2.133 s

2.0

0.8 1.5 0.6 1.0

0.4 0.2

0.5

0.0

0.0 0

0.2

0.4 0.6 0.8 1.0 1.2 1.4

1.6

Particle Velocity [m/s]

Porosity Particle velocity

1.0

Porosity [-]

t = 0.200 s

t =2.166 s REAR t = 2.200 s t = 2.233 s t = 2.266 s

1.8 2.0 2.2 2.4

Time [s]

t = 2.300 s

Fig. 14. Analysis of a slug that remained temporarily blocked in the conveying pipeline. (For interpretation of the references to colour in this figure, the reader is referred to the web version of this article.)

instrumentation and computational ability now offers the possibility to gain an insight into slug behaviour and properties that neither numerical simulation nor simple theoretical approaches can provide. Individual slugs were investigated with respect to particle velocity, slug porosity, radial and wall shear stresses, pressure and pressure gradient, slug aspect and flow behaviour. All characteristics were combined to obtain a unique picture of moving slugs and permit identification of relationships between parameters and systematisation of slug behaviour. A strong relationship was found between the profile of particle velocity along a slug and the radial and wall shear stresses that a slug induces at the pipe wall. The higher the particle velocity, the higher the stresses. This observation strongly supports the approach developed by Niederreiter [7] and Lecreps [9,28] who proposed to apply the kinetic theory to describe the stresses induced by a moving slug at the pipe wall and the resulting high pressure drop characterising slug flow pneumatic conveying. This is in contrast to the bulk solid mechanics approach that has been widely used until it was shown that slugs are slightly fluidised structures and compact structures cannot be conveyed pneumatically [9,23,28]. In most cases, slugs were found to display a negative particle velocity gradient from the rear towards the front, i.e. the closer the particles to the slug front, the slower they move. By applying both Newton's 2nd law and the ideal gas law, it was shown that the particles located in a

slug slice will move quicker than the particles of the slice in front of them as a result of the existing decreasing pressure gradient in the flow direction. This implies that over time, slugs will tend to shorten by compression of the bulk in the flow direction. In fact, while most slugs were found to be denser at the rear with a density slowly decreasing towards the front, some slugs were found to have constant density over their core. This may be a result of the compression process, which has reached its limitation, i.e. the minimum porosity required for gas permeation and slug transport. Indeed, in this investigation, it was found that the reason for the observed pipe blockage was the compressive state of the slug and permeability was a condition for slug flow to occur. However, other slugs did not follow this pattern and were found to undergo an extension process due to the particles at the front moving much faster than the particles at the rear. Eventually, this phenomenon may lead to separation of the slug front from the rest and subsequent formation of two independent slugs. The results obtained indicated that slugs are far from being steady structures and much more than just waves moving with an average velocity, picking up material at the front and depositing a similar amount at the back. Slugs are continuously evolving structures presenting a gradient of particle velocity, stresses, pressure and porosity over their length, which in turn determine their movement. While for this reason,

I. Lecreps et al. / Powder Technology 262 (2014) 82–95

modelling of slug flow pneumatic conveying on a slug by slug basis would be too challenging, the knowledge gained offers unique opportunity to develop an innovative pressure loss prediction model. The physical relationship between particle velocity and pressure gradient can be used in a predictive model since in general, pressure loss and slug velocity are the two unknowns to be determined. The identified correlation between particle velocity and stresses at the wall has led to the development of a predictive model based on stochastic agitation of particles, gas conservation law and slug permeability recently published in [29]. Further development of this model will be part of a following publication. Further experimental work is currently being undertaken to investigate slugs at two different locations over a pipeline. In particular, focus is on determining whether slug characteristics such as length, porosity, stresses, the profile of particle velocity and general aspect remain similar further down the pipe, provided that conditions for stable slug flow are still fulfilled. If this is the case, these characteristics are likely to be properties of the bulk material itself and slug characteristics such as porosity and stresses may be obtained from bench-scale tests.

References [1] P.S. Ramachandran, D.G. Burkhard, B.E. Lauer, Studies of dense phase horizontal flow of solids–gas mixtures, Indian J. Technol. 8 (1970) 199–204. [2] K. Konrad, D. Harrison, J.F. Davidson, R.M. Nedderman, Prediction of the pressure drop for horizontal dense phase pneumatic conveying of particles, Pneumotransport 5, Proceedings of International Conference on Pneumatic Transport of Solids in PipesBHRA Fluid Engineering, U.K., 1980, pp. 225–244. [3] B. Mi, Low-Velocity Pneumatic Transportation of Bulk Solids, PhD thesis University of Wollongong, Australia, 1994. [4] T. Krull, Slug Flow Pneumatic Conveying: Stress Field Analysis and Pressure Drop Prediction, PhD thesis The University of Newcastle, Australia, 2005. [5] D.J. Mason, A Study of the Modes of Gas–Solids Flow in Pipelines, PhD thesis Thames Polytechnic, London, UK, 1991. [6] K. Daoud, A. Ould Dris, P. Guigon, J.F. Large, Experimental study of horizontal plug flow of cohesionless bulk solids, Kona 11, 1993, pp. 119–124. [7] G. Niederreiter, Untersuchung zur Pfropfenentstehung und Pfropfenstabilität bei der pneumatischen Dichtstromförderung, PhD thesis University of Munich, Germany, 2005. [8] J.B. Pahk, G.E. Klinzing, Voidage measurement for a moving plug in dense phase pneumatic conveying using two different methods, Part. Sci. Technol. 28 (2010) 511–519. [9] I. Lecreps, K. Wolz, K. Sommer, Stress states and porosity within horizontal slug by dense-phase pneumatic conveying, Part. Sci. Technol. 27 (4) (2009) 297–313. [10] C. Nied, E. Frank, H. Dauth, K. Sommer, Experimental analysis of dynamic porosity changes during plug flow using electrical capacitance tomography, Proceedings of 11th International Conference on Bulk Material Storage, Handling and Transportation, Newcastle, Australia, 2013. [11] Y. Tsuji, T. Tanaka, T. Ishida, Lagrangian numerical simulation of plug flow of cohesionless particles in a horizontal pipe, Powder Technol. 71 (1992) 239–250. [12] S.B. Kuang, K.W. Chu, A.B. Yu, Z.S. Zou, Y.Q. Feng, Computational investigation of horizontal slug flow in pneumatic conveying, Ind. Eng. Chem. Res. 47 (2008) 470–480. [13] R.E. Stratton, C.M. Wensrich, Horizontal slug flow pneumatic conveying: numerical simulation and analysis of a thin slice approximation, Powder Technol. 214 (2011) 477–490.

95

[14] A. Levy, Two-fluid approach for plug flow simulations in horizontal pneumatic conveying, Powder Technol. 112 (2000) 263–272. [15] K. Li, S.B. Kuang, R.H. Pan, A.B. Yu, Numerical study of horizontal pneumatic conveying: effect of material properties, Powder Technol. 251 (2014) 15–24. [16] K. Konrad, Boundary element prediction of the free surface shape between two particle plugs in a horizontal pneumatic transport pipeline, Can. J. Chem. Eng. 66 (1988). [17] Y. Tomita, K. Tateishi, Pneumatic slug conveying in a horizontal pipeline, Powder Technol. 94 (1997) 229–233. [18] B. Mi, P.W. Wypych, Pressure drop prediction in low-velocity pneumatic conveying, Powder Technol. 81 (1994) 125–137. [19] Z.B. Aziz, G.E. Klinzing, Plug flow transport of cohesive coal: horizontal and inclined flows, Powder Technol. 55 (1988) 97–105. [20] R.J. Hitt, An Investigation into the Low Velocity Pneumatic Conveying of Bulk Solids, PhD Thesis Thames Polytechnic, London, UK, 1985. [21] S.B. Kuang, A.B. Yu, Micromechanic modelling and analysis of the flow regimes in horizontal pneumatic conveying, AICHE J. 57 (10) (2011) 2708–2725. [22] Z.B. Aziz, G.E. Klinzing, Optimizing the performance of a plug flow system, Powder Technol. 62 (1990) 77–84. [23] I. Lecreps, O. Orozovic, M. Eisenmenger, M.G. Jones, K. Sommer, Methods for in-situ porosity determination of moving porous columns and application to horizontal slug flow pneumatic conveying, Powder Technol. 253 (2014) 710–721. [24] H.A. Janssen, Versuche ueber Getreidedruck in Silozellen, Zeitschrift des VDI 29 (1895) 1045–1049. [25] J. Yi, Transport Boundaries for Pneumatic Conveying, PhD thesis University of Wollongong, Australia, 2001. [26] B. Mi, P.W. Wypych, Investigations into wall pressure during slug-flow pneumatic conveying, Powder Technol. 84 (1995) 91–98. [27] N. Vasquez, L. Sanchez, G.E. Klinzing, Dense-phase plug flow analysis: experimental findings, Proceedings of the Conference World Congress of Particle Technology WCPT4, Sydney, Australia, 2002. [28] I. Lecreps, Physical Mechanisms Involved in the Transport of Slugs During Horizontal Pneumatic Conveying, PhD thesis University of Munich, Germany, 2011. [29] I. Lecreps, O. Orozovic, M.G. Jones, K. Sommer, Application of the principles of gas permeability and stochastic particle agitation to predict the pressure loss in slug flow pneumatic conveying systems, Powder Technol. 254 (2014) 508–516. Dr. Isabelle Lecreps has been a consulting engineer and research fellow at TUNRA Bulk Solids, Newcastle, Australia, since 2010. She graduated with a master's degree in Food Technology at Polytech Lille and a master's degree in Process Engineering at the University of Compiegne, France, in 2003. She was awarded a doctorate degree from the Technical University of Munich for her work on the physical mechanisms involved in slug transport in 2011. Ognjen Orozovic is a Ph.D. student at the Centre for Bulk Solids and Particulate Technologies at the University of Newcastle, Australia. He graduated with a bachelor's degree in Mechanical Engineering at the University of Newcastle, Australia, in 2012. His current research project is in the area of slug flow pneumatic conveying, focusing on the formation, stability and decay mechanisms of slugs. Prof. Mark Jones is Head of the School of Engineering and Director of TUNRA Bulk Solids and has held the Chair in Bulk Solids Handling at the University of Newcastle, Australia, since 1999. He is currently Honorary Secretary of the Australian Society for Bulk Solids Handling and Vice-President of the International Federation of Measurement and Control of Granular Materials and has been appointed Guest Professor of Central South University, Changsha, China and Visiting Professor at Teesside University in the UK. Prof. Karl Sommer has led the Institute for Process Engineering of Disperse Particles at the Technical University of Munich since 1982. His expertise focuses on bulk solid mechanics with particular interest for sampling, mixing, agglomeration, adhesion and pneumatic conveying.