1. Introduction
The application of gas-solid fluidized beds in industry is highly valued due to their abilities of providing an excellent interaction between solid particles and the gas medium, which in turn enhances energy conversion. Gas-solid fluidized beds are widely applied and have two main fields of application: (i) chemical engineering (i.e., catalytic cracking, mixing/segregation of powders), (ii) energy conversion (i.e., steam and hot water production in boilers) [
1,
2]. The hydrodynamics of gas-solid fluidized beds are complex, primarily determined by the combined effects of solids’ behaviour and bubbles’ characteristics in terms of development, movement, and burst. The application of chemical-looping combustion (CLC) lies between these two points, i.e., it is both part of the chemical engineering application and in energy conversion and steam production.
It has been well documented that utilizing a mixture of metal oxides significantly improves the oxygen storage capacity of oxygen carrier particles in CLC systems that consist of dual interconnected circulating fluidized bed (CFB) [
3,
4]. This mixture of oxygen carrier was used as a single particles carrier containing both ingredients. However, other systems utilize a binary-mixture system (BMS) as two separate species that differ in sizes and/or densities in bubbling fluidized beds [
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16], and in a single-column fluidized bed [
17,
18,
19,
20,
21]. Utilizing a BMS that differs in size and/or density will raise a number of operational uncertainties associated with the mixing/segregation of particles and the hydrodynamics of these complex systems. Mixing and segregation of binary solids that differ in both size and density have been investigated previously, where an operating map was developed to avoid any type of segregation (i.e., local or components segregation) in CLC systems [
22,
23,
24]. The hydrodynamics of BMS in a dual interconnected CFB are yet to be investigated extensively. Specifically, the effects of parameters including particle size and density, mixture composition, total solids inventory (TSI) on the pressure profile, solids holdup and solid circulation rate have not been adequately examined [
2,
25,
26,
27,
28,
29,
30]. In particular, studies concerning the role of the riser in determining the solid circulation rate, solids holdup and the stability of these CFBs, are very limited.
In addition, to investigate the hydrodynamics of CFB, time-series analysis of pressure fluctuation (gauge or differential pressure) or other signals such as solids holdup can also be used to describe the flow regimes [
25,
26,
29,
30]. Bai et al. (1996) studied the pressure fluctuation in a single column fluidized bed to characterise different fluidization regimes. They reported the solids holdup (
ϕs) values of 0.35 <
ϕs < 0.6 for the bubbling fluidization, 0.15 <
ϕs < 0.35 for the turbulent fluidization, 0.05 <
ϕs < 0.15 for the fast fluidization and
ϕs < 0.05 for the pneumatic transport [
25]. The standard deviation (
SD) of pressure fluctuation was also used to determine the minimum fluidization velocity of the binary mixture of solids [
31]. Lue and Wu (2000) used the sum of the
SD of pressure fluctuation of each species to determine the
SD of the binary mixture by multiplying the fraction of each species in the mixture to its corresponding individual
SD under the same operating conditions [
31]. Pressure fluctuation analysis has also been used to determine the combustion region in a fluidized bed reactor by studying the
SD of the pressure fluctuation, and it was found that the combustion region was related to the regions that have high
SD values [
27]. In addition, the dominant power spectral density (
PSD) is also helpful information for investigating the hydrodynamics in a fluidized bed with the method of fast Fourier transform (FFT). The irregular behaviour over time is due to the linear summation of periodic or random fluctuations, which is assumed by spectral and statistical analyses of electrical signals [
28]. Conversely, owing to the complexity of gas–solid interactions, the hydrodynamics of the fluidized bed feature high non-linearity, anisotropy and different time scales that point to the non-stationary nature of the bed dynamics [
1,
32,
33]. Therefore, the gas–solid fluidized bed has been considered as a chaotic system [
34]. The
PSD analysis by means of FFT analysis has been generally applied to the time series of pressure signals in a fluidized bed [
35,
36].
Most of the previous hydrodynamics studies have been focused on the SSS systems [
37,
38,
39,
40,
41,
42,
43,
44] with a view to gaining insight into: (i) the bed operating regime, (ii) solid entrainment, (iii) gas leakage, (iv) particle residence time, and (v) pressure profile [
43,
45,
46]. The total solids inventory has been identified to play a significant role in defining the specific solid circulation rate whereby an increase in the inventory leads to an increase in solid circulation rate [
23,
45,
47]. Also, attempts were also made towards: (i) improving the gas–solid interaction over the total height of the fuel reactor; (ii) reducing total inventory requirements of the fuel reactor by improving the solids-gas interaction; and (iii) increasing the cross-sectional area of the fuel reactor to provide steadier global solids holdup [
43,
44,
45]. Operating the fuel reactor in the turbulent or fast fluidization regimes were also proposed as an effective means to enhancing the gas–solid contact, thus to potentially reduce the total solids inventory which becomes of great relevance for increased plant capacities [
43].
However, the riser is one of the major components of all CFB systems (such CLC) which contributes to the global solid circulation rate. The objective of this study is to investigate systematically the hydrodynamics of SSS and BMS in a cold-flow model (CFM) of CLC systems, in particular in the riser and its related component i.e., the air reactor, through a detailed pressure fluctuation analysis. The effects of the system key parameters including fluidization superficial gas velocity, particle size and density, mixture composition, total solids inventory, on the pressure fluctuation characteristics are investigated systematically based on the PSD and SD analyses.
2. Materials and Methods
Experiments were performed on a 10-kW
th CFM–CLC system located at the Priority Research Centre of Frontier Energy Technologies and Utilisation at the University of Newcastle, Australia (
Figure 1), with system total solids inventory capacity is between 1 to 3 kg. This is a lab scale 10 kW
th CFM–CLC unit which was developed and designed based on Glickman scaling laws [
48,
49,
50]. The total solids inventory was varied between 1 to 2 kg, which represents the specific inventory norms of 100–200 kg/MW
th [
39,
51]. Particles physical properties, operating conditions, and operating dimensionless numbers are listed in
Table 1 and
Table 2, respectively. The apparatuses’ component geometry in terms of air and fuel reactors, risers and loop seals, and the applied experimental procedure, methodology and analysis in this work were described in greater detail in previous related studies (i.e., [
23,
52,
53,
54]).
Glass bead (GB) particles with an apparent density of 2462 kg/m
3 was used in the SSS experiments (
Table 2); this is because it is the dominant species in the BMS experiments. GB particles Reynolds number in the air reactor (
Redp(A.R.)) is between 1.75–3.8 and in the riser (
Redp(riser)) is between 7–13.6. For the BMS, particles of different size GB and polyethylene (PE) with apparent densities of
ρPE = 939 kg/m
3 as shown in
Table 3 were used for the hydrodynamics studies (PE with
Redp(A.R.) = 4–6.42 and
Redp(riser) = 16.5–25.7). Different sizes of glass beads (GB) and polyethylene (PE) particles were prepared and used in the experiments, ranging from 98 to 328 μm, at three compositions of polyethylene particles, i.e., 5 wt.%, 10 wt.%, and 20 wt.%. A small amount of Larostat antistatic agent was added into the mixture to prevent electrostatic effects during fluidization. In the cold flow model, N
2 was used in the air reactor and air in the fuel reactor. The reason for using N
2 in the air reactor is to generate a stable fluidization of particles. This is because of the limited capacity of the available compressed air system (in term of supplied pressure), especially for the air reactor. Alternatively, N
2 was used that has very similar density and viscosity to those of air (i.e.,
ρ(air)/
ρ(N2) = 1.04 and μ
(air)/μ
(N2) = 1.04) to supply the required pressure [
48]. At room temperature in cold-flow model studies, it is believed that the error induced only by this subtle difference in density ratio would be negligible. The density ratios between the utilized particles and fluid (
ρp/
ρf) are 2113 and 806 for GB and PE, respectively.
CFM–CLC are usually made out of Perspex material, therefore, there are two important reasons of not using active metal oxide in CFM–CLC units: (i) abrasion effect if optical measurement methods were utilised, and (ii) the expensive use of active metal oxides in the CFM–CLC experiments. In the hydrodynamic studies of fluidized bed (specially in CLC systems) dimensionless analysis is conducted, which is essential for selecting special particles, these particles should relatively represent the same physical features in terms of their sizes and densities as those of metal oxide particles used in an actual CLC process [
43,
52]. This dimensionless analysis considers the density and the viscosity of the fluid at the required temperature. The selected particles were calculated based on the Archimedes number (
Ar) expressed as:
where
dp is the particle size,
g is the acceleration due to gravity,
μ is the gas viscosity,
ρf and
ρp are the fluid (i.e., gas) and particle densities, respectively. By assuming that the dimensionless
Ar number for the CFM at 25 °C is equal to the
Ar number at the actual hot CLC process (i.e., 500–1000 °C):
where
Ar(25 °C) was determined using
μ and
ρf for air at 25 °C,
ρp and
dp of GB and PE, and
Ar(hot) is expressed as,
By substituting Equation (3) into Equation (2),
The diameter of the metal oxide in the actual CLC process can be obtained by knowing the metal oxide density, which is well defined in the literature, and by knowing the density and the viscosity of the fluid at the required temperature. The diameter of the metal oxide in the actual CLC process, which represents the particles used in the CFM–CLC system in terms of density and size, can be found using the following expression:
Therefore, based on this analysis, 98–138 μm (GB) at ambient operating conditions (i.e., 25 °C) relatively represents particles such as CuO (
ρp = 6315 kg/m
3, 117–313 μm), Fe
2O
3 (
ρp = 5242 kg/m
3, 125–330 μm), NiO (
ρp = 6670 kg/m
3, 115–307 μm) and SiO
2 (
ρp = 2684 kg/m
3, 156–417 μm) at 500–1000 °C using air in the air reactor (
Ar = 78–217) and CH
4 in the fuel reactor (
Ar = 48–133). Similarly, 234–275 μm (PE) at ambient operating conditions (i.e., 25 °C) relatively represents particles such as CuO (
ρp = 6315 kg/m
3, 203–457 μm), Fe
2O
3 (
ρp = 5242 kg/m
3, 216–487 μm), NiO (
ρp = 6670 kg/m
3, 200–449 μm) and SiO
2 (
ρp = 2684 kg/m
3, 270–608 μm) at 500–1000 °C using air in the air reactor (
Ar = 403–1111) and CH
4 in the fuel reactor (
Ar = 247–679). Under constant
Ar number, to have hydrodynamic similarity, the Froude number ratio (i.e., between the hot and cold model) can be close to 1 [
43,
48,
49,
50]. The velocity of the air reactor in the hot model should be between 0.9–1.8 times that of the CFM. The applicability of the result obtained from this lab-scale CFM–CLC using the particles in
Table 1 is validated and might only be applied to this or similar units. The scaling from 10 kW
th CFM–CLC (i.e., lab scale) to 200 kW
th CFM–CLC (i.e., demonstration pilot plant) using similar particles and approach was also validated in previous work and showed hydrodynamic similarity [
48].
At fixed air reactor fluidization velocity (
ug(A.R.)), each case was conducted at least 3 times with five fuel reactor fluidization velocities (
ug(F.R.)) in the range of 0.0294–0.1470 m/s. After conducting the experiments and steady-state operating conditions being reached (
Figure A1), the pressure readings were taken using 20 pressure ports allocated on the apparatus (i.e., Honeywell micro switch sensoring and control, 142PC05D, 164PC01D37 and 142PC01D, Morris Plains, New Jersey, USA) that recorded pressure at different points in the system, and in the same time the solid circulation rate was measured. The solid circulation rate (i.e., g/min) was obtained using the direct measurement method [
53] during a steady state operation by stopping the aeration of loop seal 1 and measuring the time required to reach a particle bed height accumulation of 1–2 cm (repeated for 3 times). The solids holdup was obtained using pressure drop measurement (i.e.,
ϕs = Δ
p/[
ρpgΔ
h]) from the pressure ports installed along the CFM–CLC system (the numbers in
Figure 1 are the location of each pressure port). Two types of pressure data throughout the CFM–CLC system were recorded simultaneously at steady-state operation conditions (i) an averaged pressure value from 100 readings of each pressure port (i.e., 10 readings per 1 s for 10 s interval), and (ii) transient pressure data recorded for each port with a rate of 500 readings per second for a period of 10–60 s (i.e., 5000–30,000 readings). The pressure at each point in the system was compared with the pressure values of the zero reference point measurements that were taken at the beginning of each experiment. Thus, the absolute pressure values of the system could be obtained. The confidence intervals of the error associated with the pressure measurements were less than ±0.03 kPa at 95%, where also the uncertainty analysis in our study showed that the associated error of the solid circulation rate measurements was rather reasonable, falling within 5%–15% [
52,
53].
To understand the effect of using BMS in the CFM–CLC, a systematic investigation was conducted using firstly SSS, and later another species was added into the system for BMS analysis. For both systems, the effect of changing the superficial gas velocity i.e., in the air reactor (
ug(A.R.)) and riser (
ug(riser)), total solids inventory, particle size, along with the effect of adding another species at constant fuel reactor
ug(F.R.) on riser pressure fluctuation were analysed qualitatively and quantitatively. In the CFM–CLC, the air reactor has larger diameter (i.e.,
din(A.R.) = 80 mm and
h = 300 mm) and operates under turbulent to fast fluidization regimes (to increase the particle residence time in the reactor), while the riser has smaller diameter (i.e.,
din(riser) = 40 mm and
h = 1150 mm) and operates under pneumatic conveying regime to give the particles that departed the air reactor the additional momentum to circulate the system, as shown in
Figure 1. For the first part of the experimental investigation, the pressure fluctuation was taken for 60 s (i.e., for SSS). However, to save experimental time and since there is no change in the global solid circulation in the system at steady state, it was decided to be taken at the same sampling rate, however, only for 10–15 s. It is worth noting that the transient pressure fluctuation figures are averaged for every 60 readings to avoid a thick line that cannot be read and compared, all pressure fluctuation figures are produced for 11–12 s. The averaged pressure, solids holdup, and solid circulation rate profiles of the examined conditions along with their detailed discussions can be found in our earlier studies [
23,
52].
Pressure Fluctuation Analysis
The complex Fourier method was used in the analysis of the pressure signals monitored in the air reactor and the riser. For the complex Fourier method, the first step is to find the complex Fourier expansion amplitudes for given temporal-varying pressure:
in which
t denotes to the time of the monitored variable,
ωi is the angular frequency (
ωi =
i × (2π/
P)),
P is the monitored period (
P = 11.0 s),
i is an integer and
X and
Y are the amplitude at the angular frequency
ωi.
The nature of the Fourier analysis as shown in Equation (6) is to fit all monitored data using an analytic function expressed as the right hand side (RHS) of Equation (6). There are 2N + 1 unknowns (
X0,
X1,
X2…
XN;
Y1,
Y2, …
YN) which are solved by the Fourier series analysis. During 11 s period, 5500 monitored points match well with the prediction from the Fourier coefficients (
X0,
X1,
X2…
XN;
Y1,
Y2, …
YN) as shown by
Figure 2.
The signal-sampling rate for the pressure fluctuation was 500 s
−1 over a time period of 10–60 s at the steady-state operation. The pressure fluctuation
SD was calculated by:
where
N is the number of the sampling points; Δ
Pi is the pressure drop for each sampling point and Δ
PAverage is the average pressure drop over the entire sampling points. Some of the
SD data are found to be slightly higher or lower due to the artefact of the local solids holdup just before or after the area of interest (i.e., point 3–4 in the air reactor and point 6–9 in the riser). Thus, for some points the average of three
SD values was taken (e.g.,
SD for pressure at point 6,
SD for pressure at point 9 and
SD for pressure drop point 6–9).