Ultrasonics 38 (2000) 813–816 www.elsevier.nl/locate/ultras
Ultrasound process tomography system for hydrocyclones H. In˜aki Schlaberg *, Frank J.W. Podd, Brian S. Hoyle University of Leeds, School of Electronic and Electrical Engineering, Leeds LS2 9JT, UK
Abstract The implementation of a laboratory-based ultrasound tomography system to an industrial process application is not straightforward. In the present work, a tomography system with 16 transducers has been applied to an industrial 50 mm hydrocyclone to visualize its air-core size and position. Hydrocyclones are used to separate fine particles from a slurry. The efficiency of the separation process depends on the size of the air core within the cyclone. If the core is too large due to spigot wear, there will be a detrimental effect on the slurry throughput. Conversely, if the throughput is increased to an extent where the air core becomes unstable or disappears, the particle separation will no longer take place, and the processed batches may become contaminated. Ultrasound tomography presents a very good tool with which to visualize the size, position and movement of the air core and monitor its behaviour under varying input parameters. Ultimately, it could be used within this application both to control the input flow rate depending on the air core size and to detect spigot wear. This paper describes the development of an ultrasonic tomography system applied to an instrumented hydrocyclone. Time-of-flight data are captured by a dedicated acquisition system that pre-processes the information using a DSP and transfers the results to a PC via a fast serial link. The hardware of the tomography system is described, and cursory results are presented in the form of reconstructed images of the air core within the hydrocyclone. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Tomography; Ultrasonic process monitoring
1. Introduction The size of the air core in hydrocyclones is important to the efficiency of the separation process for which they are used. It is often difficult to obtain this parameter by optical means as the slurries are generally opaque. Some observations have been made by using cameras over the top of relatively large units [1]. Smaller units, however, are totally enclosed, and optical access is not possible. As the slurries are mostly water-based suspensions, ultrasound presents, in principle, a relatively simple way of determining the air core size and position. If the air core is located along the centre axis and remains stationary, a single transducer would be sufficient to determine its size. It may, however, be the case that the air core oscillates, and a larger number of transducers are necessary to establish its size and position in this case. Ultrasonic reflection tomography [2] can aid in providing a graphic representation of the size and position of the air core by taking measurements * Corresponding author. Tel.: +44-113-233-2074; fax: +44-113-233-2032. E-mail address:
[email protected] (H.I. Schlaberg)
from multiple views around the perimeter of the hydrocyclone. For this work, as many transducers as were practically feasible have been used to provide a tomographic image of a cross-section of the hydrocyclone.
2. Hydrocyclones Hydrocyclones are widely used in chemical and mineral industries to separate particles by size or density. They have a conical shape, and a liquid containing the particles is fed tangentially to its top. There are openings at the top and bottom of the cone. The slurry rotates from the top downwards, and part of it leaves the hydrocyclone via the lower opening (apex), through an attachment called a spigot. At the centre of the cone is an air core, and part of the slurry that reaches the bottom travels upwards along this core and finally leaves the hydrocyclone through the top (vortex) (see Fig. 1). Larger particles are collected in the bottom flow while the top flow carries the smaller particles. The efficiency of the separation process is linked to the size of the air core. It is thus desirable to increase
0041-624X/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S0 0 4 1 -6 2 4 X ( 9 9 ) 0 0 18 9 - 4
814
H.I. Schlaberg et al. / Ultrasonics 38 (2000) 813–816
Fig. 1. Operating principle of a hydrocyclone. Fig. 2. Hydrocyclone with and without sensors.
the incoming flow rate until the air core has a very small size. In this case, more slurry and less air go through it, thus making the process more effective. If, however, the flow rate is increased to an extent that the air core disappears, the separation process stops, and a flow with mixed particles will leave through both outlets. It is possible for the spigot at the bottom to wear over time, and this has a detrimental effect on the efficiency as it increases the size of the air core.
3. Equipment and experiment The purpose of this work is to visualize the air core of the hydrocyclone so that the process can be driven at its most efficient regime and also to detect spigot wear. A 50 mm upper diameter hydrocyclone was instrumented with 16 ultrasonic transducers (see Fig. 2) and a close-by set of pulser/preamplifier circuits ( Fig. 3). The number of transducers used was mainly determined by the physical size of the transducers (3 mm wide) and the available space around the top of the hydrocyclone. This set-up was attached to a box of electronics with time gain controlled amplifiers and a DSP that sends the converted information to a host computer. The transducers are numbered consecutively (1–16) around the ring starting at an arbitrary position. The hardware system sends information in the form of emitter, receiver and time-of-flight ( TOF ) data packets to the host, where an image of the air core is reconstructed (see Fig. 4). That is, if an echo was detected, the number of the pulse originating transducer and the number of the detecting sensor, together with the time it took for the pulse to
Fig. 3. Circuit ring around hydrocyclone top.
get from one to the other, are transmitted. The hardware system is described in more detail in Refs. [3,4]. The operation principle is as follows. The DSP receives a command from the host computer over the fast (20 Mbit/s) serial link to acquire one frame of data. Subsequently, it triggers the first transducer and starts the time-gain-compensation ( TGC ) circuit. Signals from all the transducers (including the emitter) are then TGCamplified, and if they exceed a pre-set threshold, a digital output is asserted. The DSP samples these outputs at regular intervals and stores the results in memory. When sufficient time has elapsed for the furthest possible echo to arrive, the
H.I. Schlaberg et al. / Ultrasonics 38 (2000) 813–816
815
Fig. 4. Block diagram of the hardware.
sampling process stops. The system is paused for a short period to allow any echoes to die down. The next transducer is then triggered, and the process is repeated until all the sensors have been activated. Between sampling periods the DSP has enough time to initiate a data conversion process that translates the stored values of the 16 (one per sensor) digital outputs from the transducers into data packets that have the format of: emitter, receiver, time-of-flight ( TOF ). This operation reduces the amount of information transmitted to the host and also decreases the processing necessary to reconstruct an image. The image is reconstructed by drawing arcs along a path where an object might be located, with the given data of the emitter, receiver, TOF and the fan angle of the transducer (see Fig. 5). Superimposing the arcs of all the data items generates an image where the boundary of the object is highlighted by the intersection of these arcs. It is possible to accelerate the image reconstruction by using multiple processors in parallel and dividing the image into sections for each processor. Alternatively,
Fig. 5. TOF projection paths of possible object locations using a single transducer.
every nth data set could be allocated to a different processor, and the partial images would be combined to form the final image with the complete data set. Using this approach, it is possible to obtain images in real-time, and reconstruction speeds of around 30 frames per second have been shown in Ref. [5].
4. Results Fig. 6 shows an image obtained with only a partial subset of the available transducers while the hydrocyclone was operating. At the captured frame rate of about six frames per second (on a 266 MHz PC ), the air core appeared stationary and in a steady state. The image has a resolution of 50×50 pixels, and each pixel corresponds to a physical dimension of about 1 mm. This is determined by the sampling frequency of the acquisition system, which checks whether an echo is present at intervals of about 1.3 ms. This proves that ultrasonic tomography can be used
Fig. 6. Captured image with a partially operating sensor.
816
H.I. Schlaberg et al. / Ultrasonics 38 (2000) 813–816
to visualize the air core inside a hydrocyclone when other methods such as optical inspection are not possible. To determine the size of the air core, the radial distance of the most highlighted pixels can be measured. Alternatively, methods using the Hough transform, as described in Ref. [6 ], can be employed. Here, an algorithm processes the pixel field with parameters of radius and circle centre coordinates and seeks a high correlation between the parameters and the image. This is a fairly robust method as it can find circles that are only partially complete due to noise or sensor failure.
reconstruction algorithm leading to faster capture and visualization would make it possible to find out if the air core was oscillating at frequencies higher than those that can be detected with the present set-up.
5. Conclusions
References
Ultrasound reflection tomography offers a opportunity to measure the size and position of the air core of a hydrocyclone when optical methods are not viable. It has been shown that an air core can be visualized using a relatively small number of transducers. The measurements can be improved by using higher sampling rates and more transducers and by using the transmission information available from the sensors. The hardware is capable of producing data at around 200 frames per second, but the data transmission, image reconstruction and display make it only possible to observe images at around six frames per second. Improvements in the
[1] O. Castro, F. Concha, J. Montero, J. Miranda, J. Castro, D. Urizar, Proc. Hydrocyclones-96, Cambridge, UK, April (1996) 229. [2] H.I. Schlaberg, M. Yang, B.S. Hoyle, Electron. Lett. 32 (17) (1996) 1571. [3] H.I. Schlaberg, M. Yang, B.S. Hoyle, M. Beck, C. Lenn, Ultrasonics 35 (1997) 213. [4] M. Yang, H.I. Schlaberg, B.S. Hoyle, M. Beck, C. Lenn, IEEE Trans. Ultrasonics Ferroelect. Freq. Contr. 46 (3) (1999) 492. [5] M. Yang, H.I. Schlaberg, B.S. Hoyle, M. Beck, C. Lenn, J. RealTime Imag. 3 (4) (1997) 295. [6 ] M. Yang, H.I. Schlaberg, B.S. Hoyle, M. Beck, C. Lenn, Frontiers in Industrial Process Tomography Conf., Delft, The Netherlands, April (1997) 349.
Acknowledgements The authors gratefully acknowledge the support of English China Clays and the collaboration of the PACE group at the Camborne School of Mines, University of Exeter, UK.