Online continuous measurement of the size

Flow Measurement and Instrumentation 40 (2014) 163–168

Contents lists available at ScienceDirect

Flow Measurement and Instrumentation journal homepage: www.elsevier.com/locate/flowmeasinst

Online continuous measurement of the size distribution of pneumatically conveyed particles by acoustic emission methods Yonghui Hu a, Xiangchen Qian a, Xiaobin Huang a, Lingjun Gao b, Yong Yan b,a,n a b

School of Control and Computer Engineering, North China Electric Power University, Beijing 102206, PR China School of Engineering and Digital Arts, University of Kent, Canterbury, Kent CT2 7NT, UK

art ic l e i nf o

a b s t r a c t

Available online 23 July 2014

The particle size distribution of pulverized fuel in pneumatic conveying pipelines is an important physical characteristic closely related to mill energy economy, fuel flow property, combustion efficiency and pollutant emissions. In order to determine the size distribution of pneumatically conveyed fuel particles on an online continuous basis, an instrumentation system based on acoustic emission (AE) method is developed. This method extracts information about particle size distribution from the impulsive AE signals generated by the impacts of particles with a metallic waveguide introduced into the particle flow. Analytical modeling of the particle impact is performed in order to establish the relationship between the particle size and the peak AE voltage. With the particle velocity obtained from an electrostatic velocimetry system and the pulse magnitude determined using a peak detection algorithm, the particle size is computed directly using the developed model. Experimental results obtained with glass beads on a laboratory-scale particle flow test rig demonstrate that the system is capable of discriminating particles of different sizes from the AE signals. The system has several appealing features such as online continuous measurement, high sensitivity, simple sensor structure and low cost, which make it well suited for industrial applications. & 2014 Elsevier Ltd. All rights reserved.

Keywords: Particle flow On-line particle sizing Acoustic emission Hertz theory of contact

1. Introduction The pneumatic conveying of pulverized fuel (coal or biomass) through enclosed pipelines is an important technique widely used in the power generation, steel and cement industries. One of the key parameters in such gas–solid two-phase flow systems is the size distribution of pulverized coal or biomass particles. Many aspects of the combustion process, such as fuel flow property, pipe erosion, flame stability, combustion efficiency and pollutant emissions, are heavily affected by the size of fuel particles [1–4]. Online continuous measurement of particle size distribution could help realize optimized mill operation settings, balanced coal flow to burners and improved combustion performance. The strong demand for online sizing of fuel particles in pneumatic conveying pipelines has let, over the years, to the development of a variety of techniques and instruments. Most of the proposed particle sizing techniques are based on well known physical principles, such as laser diffraction [5], digital imaging [6], ultrasound attenuation [7], mechanical vibration [8] and electrostatic sensing [9]. The particle size distribution can either be

n Corresponding author at: School of Engineering and Digital Arts, University of Kent, Canterbury, Kent CT2 7NT, UK. Tel.: þ44 1227823015; fax: þ 44 1227456084. E-mail address: [email protected] (Y. Yan).

http://dx.doi.org/10.1016/j.flowmeasinst.2014.07.002 0955-5986/& 2014 Elsevier Ltd. All rights reserved.

constructed by measurement of individual particles and counts of particles of similar size, or be extracted from a combined signal for all particles. Although the feasibility of these techniques has been verified in laboratory conditions, their deployments in the field remain problematic due to the hostile environmental conditions, high implementation and maintenance costs, or requirements on sensor accuracy, reliability and durability. For instance, digital imaging is deemed a promising technique for on-line continuous measurement of particle size, but the contamination of optical access windows due to fine dust deposition even coupled with air purging mechanism makes the technique unsuitable for long-term, routine operation in an industrial environment. In power plants, a common technique for the measurement of fuel fineness has long been isokinetic sampling and sieving method. Because the method is both cumbersome and costly, inspections of fuel fineness are conducted only periodically or when anomalous operating conditions are detected. The long time delay between particle sampling and the results analysis also precludes the possibility of real-time feedback control and optimization. Therefore, it is desirable to develop a new practical technique that is accurate, reliable, simple and cost-effective for online particle sizing. This paper is concerned with online particle sizing using acoustic emission (AE) method. AE refers to the generation of transient elastic stress waves due to the rapid release of energy from localized sources within a material. There have been some

164

Y. Hu et al. / Flow Measurement and Instrumentation 40 (2014) 163–168

Fig. 1. Signal shaping chain.

preliminary studies of the AE method for particle size analysis. The seminal work conducted by Buttle et al. [10] demonstrated that the peak amplitude and rise time of the AE signal arising from particle impact upon an aluminum target plate could be used to recover particle size information under certain conditions. Due to the complexity of the instruments used, elaborate theoretical analysis and precisely controlled laboratory conditions are required and their technique is thus impractical for industrial applications. Bastari et al. [11] proposed a system identification technique to establish the relations between the acquired AE signals and the size distribution of impacting coal powder. A multi-step procedure, consisting of wavelet packet decomposition, multivariate data analysis and neural network mapping, is employed for particle sizing under limited operating conditions. Uher and Benes [12] validated experimentally the applicability of Hertz theory of contact in the measurement of particle size distribution in solid particle flow. The agreement between the theoretical and experimental data in spectral characteristics suggests possibility of distinguishing particle mixtures by means of spectral analysis of the AE signal. Boschetto and Quadrini [13] used AE signals to recognize different size distributions of metallic and inorganic powder. A simple threshold-based time-counting technique is used to derive approximate particle sizing results. It can be concluded from the review of all existing literature that the AE-based particle sizing technique is still at an early stage of development. There exist many scientific and technical issues to be examined, such as AE generation and propagation mechanisms in particle flow, optimal probing design, quantitative characterization of AE signals and efficient algorithms for online particle sizing. Moreover, a cost-effective instrument operating on this principle in a real industrial environment remains to be developed. This paper presents for the first time the design, implementation and experimental assessment of an AE-based online particle sizing instrument. The developed instrument allows the solid particles to strike a metallic waveguide protruding into the particle flow for generation of impulsive AE signals. The impact signals are then detected by a piezoelectric AE sensor and analyzed in an online continuous manner to infer the particle size during each impact event. The effectiveness of the prototype instrument is experimentally validated under laboratory conditions. The instrument enjoys several advantages such as online continuous measurement, high sensitivity, simple sensor structure and low cost and is thus well suited for industrial applications.

2. Methodology 2.1. Theoretical model The impacts of solid particles upon a metallic plate generate transient elastic stress waves that propagate away from the impact points. The stress wave signatures are closely related to the impulsive forces that the particles impose on the plate. One of the key parameters affecting the impact force is the size of the impinging particle. The minute surface displacements caused by stress waves can be detected by a piezoelectric transducer, located some distance away from the AE source. Through physical

modeling and digital signal processing, it is possible to quantitatively characterize the particle size distribution from individual or successive impact events. The AE signal from the transducer is dependent not only upon the particle and impact dynamics but also upon the physics of wave propagation and the instrument response to surface vibrations. As illustrated in Fig. 1, the original motion at the source is shaped by a chain of distinct processes. Assuming that the wave propagation medium and the instrument can be modeled as linear, time invariant systems, the AE signal can be expressed as [10] VðtÞ ¼ SðtÞn GðtÞn RðtÞ

ð1Þ

where V(t) denotes the measured AE voltage signal and S(t), G(t) and R(t) are the acoustic source, wave propagation and instrument response functions, respectively. The symbol n represents convolution. If the wave propagation and the instrument response functions are known, their effects can be decoupled from the measured AE signal through de-convolution. Information about the particle size is then extracted from the derived source function. The above technique, known as quantitative AE, can achieve quantitative characterization of the particle size. However, determination of the wave propagation and instrument response functions require precise modeling of the physical process and accurate calibration of the data acquisition system, both of which are challenging tasks. It is also doubtful that such a technique could be utilized in practical situations, where the particles are irregular in shape, the signals are contaminated by noise and the pulses are overlapped due to simultaneous impacts of multiple particles. For these reasons, an inferential particle sizing approach based on the peak voltages of individual impact events is employed in the present study. The source function (Fig. 1) representing the time history of the impact force can be determined by Hertz theory of contact [14]. The formulation of the theory assumes that the impact is normal and elastic, the particle is spherical and the plate is perfectly flat. The impulsive force that the particle imparts to the plate can be approximated by [15] ( SðtÞ ¼

f max ð sin ðπ t=t c ÞÞ3=2

for 0 ot o t c

0

otherwise

ð2Þ

where tc ¼ 4.53(4ρ1π(δ1 þ δ2)/3)2/5r1v0 1/5 is the time the particle 2/5 2 6/ spends in contact with the plate, and fmax ¼1.917ρ3/5 r1v0 1 (δ1 þ δ2) 5 is the peak compression force. In above equations, δi ¼ (1 μ2i )/(πEi), E and μ are Young's modulus and Poisson's ratio, respectively, and subscripts 1 and 2 refer to the materials of the particle and the plate, respectively. ρ1, r1 and v0 represent the mass density, mass equivalent radius and approach velocity of the particle, respectively. Given particle velocity and properties of both materials, the particle size can be theoretically derived from the source function. With the assumption that the wave propagation medium and the instrument are linear, time invariant systems, the maximum voltage of an AE pulse corresponds to the maximum compression and is proportional to the peak compression force. Therefore, the relationship between the maximum voltage of an AE pulse, the


165

Fig. 2. Schematic illustration of particle size measurement through AE detection.

particle velocity and the particle size can be expressed as [16] V max ¼ Kr 1 2 v0 6=5

ð3Þ

where Vmax is the maximum voltage of the AE pulse and K is a proportionality constant. Calibration of K can be performed using isometric particles of known size distributions. Once the particle velocity and the constant are determined, the particle size can be directly computed from the peak pulse voltage using Eq. (3).

2.2. Instrument design The measurement of particles in pneumatic conveying pipelines can be implemented by detecting the AE signals generated by impacts of particles either with a natural bend in the pipeline wall or with an artificial obstacle introduced into the flow. In order to isolate structural noises from the pipeline and obtain a reasonable impact frequency, an intruded waveguide is employed as the target for particle impacts. After propagating through the waveguide, the AE signals are detected by an AE sensor attached to the outer end of the waveguide. Fig. 2 shows the sensing arrangement for on-line particle size measurement. A prototype instrument for online particle sizing was designed and constructed. 3-D drawings and installation of the prototype sensing head are shown in Fig. 3. The main body of the sensing head is a 100 mm bore spool piece, which can be connected to the pipeline through the flanges at both ends. The waveguide rod, made of durable metal–ceramic composites, penetrates into the particle flow through the wall of the spool piece. The middle section of the waveguide is cylindrical for sealing with rubber bushings. The inside section is semi-cylindrical with the flat surface facing the flow, which enables normal particle impact. The outside section is also semi-cylindrical but with opposite direction of the flat surface, onto which the AE sensor is attached. The diameter of the waveguide is 10 mm and the penetration distance is 50 mm, so that the area for particle impact is 500 mm2. The instrument permits adjustment of the penetration distance in order to change the number of particle impacts per unit time. The acoustic emission signal due to particle impact covers a broad frequency range, from the audible to the ultrasonic frequencies [17]. In order to avoid the influence of audible environmental noise and to reduce the computational overhead due to high speed sampling, a low-frequency ultrasonic piezoelectric AE sensor (type SR40M, SOUNDWEL) is used. The sensor operates in the bandwidth of 15–70 kHz, with 40 kHz as the resonance frequency. Moreover, three circular electrostatic electrodes positioned upstream of the waveguide (Fig. 3) are used to measure the velocity of particles [18,19].

Fig. 3. Sensing head. (a) 3-D drawings. (b) Installation.

The signals from the sensing head are fed to a signal processing system that performs signal conditioning, data acquisition and computation. The AE signal is amplified with a voltage gain of 40 dB and then filtered through a band-pass filter with a frequency range of 20– 100 kHz. The current signals from the electrostatic sensors are converted into voltage signals and then amplified and filtered. An analog-to-digital converter digitizes the four signals simultaneously at a sampling rate of 1.25 MHz for each channel. The particle sizing algorithm is implemented on a high-performance floating point digital signal processor running at 300 MHz. The particle sizing

166


Fig. 4. Block diagram of the signal processing system.

Fig. 6. Experimental set-up.

Fig. 5. A typical AE signal during two impact events.

results are transmitted to a host computer through a RS232 interface for display and storage. Fig. 4 shows the block diagram of the signal processing system.

2.3. Particle sizing algorithm

Table 1 Physical properties of the test particles and experimental conditions. Item

Value

Waveguide diameter Young's modulus of the waveguide Poisson's ratio of the waveguide Particle diameter

10 mm

Particle density

Fig. 5 shows a typical AE signal due to two successive impact events. The test particles are glass beads with an approximate diameter of 180 μm. The impact velocity is about 8.0 m/s. As illustrated in Fig. 5, the impulsive signal arising from a particle impact event is followed by a sequence of damped oscillations. In order to achieve particle sizing, the primary peak during each impact event should be detected. Here we propose a search algorithm for peak detection. The procedure of the peak detection algorithm is as follows: Step 1. Detection of all possible peak candidates: all the local maxima between two consecutive local minima are regarded as peak candidates. Step 2. Localization of the primary peak candidates: for three consecutive peak candidates, the one located in the middle and greater than the other two is regarded as the primary peak candidate. Step 3. Removal of false peak candidates by thresholding: all the primary peak candidates smaller than a certain threshold are deemed as pseudo peaks due to noise and are thus removed. 3. Results and discussion 3.1. Experimental set-up Experiments with the prototype particle sizing system were carried out on a 100 mm bore particle flow test rig. Fig. 6 shows the layout of the test rig. The sensing head was installed in a vertical section of the polymethyl methacrylate pipeline.

Young's modulus of the particle Poisson's ratio of the particle Particle velocity Mass flow rate of particles

550 GPa 0.22 20–250 μm 2230 kg/ m3 6.4 GPa 0.20 6–12 m/s 2.0 g/s

An industrial suction system was connected to the pipeline to generate a steady air flow. Glass beads ranging from 20 μm to 250 μm in diameter were fed into the rig via a feeding hopper. The particle velocity was varied by adjusting the power of the suction system, while the impact frequency was tuned by regulating the particle discharge rate of the feeding hopper. Table 1 summarizes the physical properties of the test particles and the experimental conditions. 3.2. Experimental results The first experiment was to investigate AE signals generated by particles of different sizes. For this purpose, the glass beads were sieved into three size ranges using a mechanical sieve shaker. The particle velocity before collision was around 8.0 m/s and the mass flow rate was 2.0 g/s. Fig. 7 shows the AE signals generated by particles below 90 μm, between 90 μm and 180 μm and above 180 μm, respectively. It is apparent that, under the above conditions, signals from individual impacts could be resolved in time and larger particles generated higher signal peaks. Therefore, the theory based on which the particle sizing algorithm was derived could be partially validated. The burst AE signals generated by small particles were so weak that they were almost immersed in


167

Table 2 Measured mass of particles in different size ranges.

6 m/s 9 m/s 12 m/s

o 90 μm

90–180 μm

4180 μm

Total mass

2.18 g 6.95 g 11.46 g

2.98 g 5.40 g 8.11 g

5.62 g 5.76 g 5.68 g

10.78 g 18.11 g 25.24 g

Fig. 8. Measured mass ratio of particles in different size ranges.

the measured total mass and the mass ratio for particles in different size ranges, respectively. The prescribed mass ratio for particles below 90 μm, between 90 μm and 180 μm and above 180 μm was 50%:30%:20%. The results suggest that higher particle velocity leads to increased accuracy of the measurements. At all velocities, the mass for particles above 180 μm remained almost fixed because all the impacts were detected. At low velocity, the pulsed signals generated by small particles were comparable to the noise and can hardly be detected, which led to large errors in the measured mass ratio. As the velocity increased, more smaller particles were detected and the results became more accurate.

Conclusions

Fig. 7. AE signals generated by particles in different size ranges. (a) Particles below 90 μm. (b) Particles between 90 μm and 180 μm. (c) Particles above 180 μm.

the background noise, which was mainly caused by the strong airborne sound from the suction system. To determine a reference particle size distribution of the offthe-shelf glass beads is not straightforward. For ease of operation, glass beads in different size ranges were mixed with known mass ratio. The measured particle size was used to compute the mass of the particle, and the total mass of particles within a size range was obtained through simple summation. The measured mass ratio was compared with the prescribed mass ratio, which could indicate the effectiveness of the system. Table 2 and Fig. 8 show

This paper presented a prototype AE-based instrumentation system for the online measurement of particle size distribution in gas–solid two-phase flows. The relationship between the particle size and the peak AE voltage was theoretically established. A particle sizing algorithm was developed using peak detection techniques. Experimental results obtained with poly-dispersed glass beads have verified the possibility of online particle size measurement from burst AE signals generated by individual impact events. Significant further work is required to advance the technique. Firstly, the dynamics of particle collision and wave propagation should be investigated through numerical simulation. A deeper understanding of the physical process will shed light on the optimized design of the sensing head. Secondly, the theoretic model based on which the particle sizing algorithm is developed should be refined for a wide range of flow conditions. The accuracy of the results is expected to be improved by making more realistic assumptions. Thirdly, effective denoising techniques and precision signal conditioning electronics will be employed in order to extend the lower limit of the measurable particle size. Finally, the performance of the instrument under a wider range of conditions, such as variations in test material, particle concentration and installation location on the pipeline, should be investigated.

168


Acknowledgment The authors wish to acknowledge the National Natural Science Foundation of China (No. 51375163), the Ministry of Science and Technology of the PR China (No. 2012CB215203) and the Chinese Ministry of Education (No. B12034). This work is also supported by the Fundamental Research Funds for the Central Universities (Nos. 13QN16 and 13QN17). References [1] Holtmeyer ML, Kumfer BM, Axelbaum RL. Effects of biomass particle size during cofiring under air-fired and oxyfuel conditions. Appl Energy 2012;93: 606–13. [2] Ninomiya Y, Zhang L, Sato A, Dong Z. Influence of coal particle size on particulate matter emission and its chemical species produced during coal combustion. Fuel Process Technol 2004;85(8–10):1065–88. [3] Yu D, Xu M, Sui J, Liu X, Yu Y, Cao Q. Effect of coal particle size on the proximate composition and combustion properties. Thermochim Acta 2005; 439(1–2):103–9. [4] Shibaoka M. Carbon content of fly ash and size distribution of unburnt char particles in fly ash. Fuel 1986;65(3):449–50. [5] Black DL, McQuay MQ, Bonin MP. Laser-based techniques for particle-size measurement: a review of sizing methods and their industrial applications. Prog Energy Combust Sci 1996;22(3):267–306. [6] Gao L, Yan Y, Lu G, Carter RM. On-line measurement of particle size and shape distributions of pneumatically conveyed particles through multi-wavelength based digital imaging. Flow Meas Instrum 2012;27(10):20–8.

[7] Su M, Xue M, Cai X, Shang Z, Xu F. Particle size characterization by ultrasonic attenuation spectra. Particuology 2008;6(4):276–81. [8] Coghill PJ. Particle size of pneumatically conveyed powders measured using impact duration. Part Part Syst Charact 2007;24(6):464–9. [9] Tajdari T, Rahmat MF, Wahab NA. New technique to measure particle size using electrostatic sensor. J Electrostat 2014;72:120–8. [10] Buttle DJ, Martin SR, Scruby CB. Particle sizing by quantitative acoustic emission. Res Nondestr Eval 1991;3(1):1–26. [11] Bastari A, Cristalli C, Morlacchi R, Pomponi E. Acoustic emissions for particle sizing of powders through signal processing techniques. Mech Syst Signal Process 2011;25(3):901–16. [12] Uher M, Benes P. Measurement of particle size distribution by the use of acoustic emission method. In: Proceedings of 2012 IEEE international instrumentation and measurement technology conference, Graz; 2012. p. 1194–8. [13] Boschetto A, Quadrini F. Powder size measurement by acoustic emission. Measurement 2011;44(1):290–7. [14] Stronge WJ. Impact mechanics. Cambridge, UK: Cambridge University Press; 2000. [15] Pecorari C. Characterizing particle flow by acoustic emission. J Nondestr Eval 2013;32(1):104–11. [16] Gao L, Yan Y, Carter RM, Sun D, Lee P, Xu C. On-line particle sizing of pneumatically conveyed biomass particles using piezoelectric sensors. Fuel 2013;113(11):810–6. [17] McLaskey GC, Glaser SD. Hertzian impact: experimental study of the force pulse and resulting stress waves. J Acoust Soc Am 2010;128(3):1087–96. [18] Qian X, Yan Y. Measurement of biomass and blended biomass fuels in pneumatic conveying pipelines using electrostatic sensor-arrays. IEEE Trans Instrum Meas 2012;61(5):1343–52. [19] Qian X, Yan Y, Shao J, Wang L, Zhou H, Wang C. Quantitative characterization of pulverized coal and biomass-coal blends in pneumatic conveying pipelines using electrostatic sensor arrays and data fusion techniques. Meas Sci Technol 2012;23(8):085307.

Online continuous measurement of the size

Online continuous measurement of the size

Suggest Documents