For high-temperature applications, buffer waveguides only a foot or so in length (~300 mm) convectively cool down enough at their cold end so that the ultrasonic flow sensor is not troubled by the temperature of a hot (300°C) fluid or hot pipe at the other end. The secret is to isolate thermally and economically, without corrupting the signal.
The primary problem with ultrasonic measurement of gas or liquid flow at temperatures >260°C lies in the survival of the transducers or sensors that generate and detect the ultrasonic waves. A secondary problem is coupling those waves into and out of the fluid (when using wetted transducers), or coupling ultrasound into and out of the pipe (when using clamp-on transducers). Whether the transducers are wetted or clamp-on, it is also important not to disturb things too much, that is, not to overly disrupt the flow patterns or the temperature. For example, local cooling at the transducer sites could cause solids to precipitate there, clogging the port or blocking the sound beam.
The main focus of this article is on the contrapropagation method of ultrasonic flow measurement (see Figure 1).
In the contrapropagation method, transit times are measured in the direction of flow, and later (or sometimes simultaneously) against the direction of flow. From the two measured transit times we can calculate the flow velocity, V. At the low Mach numbers encountered in liquid flow, the time difference, t, is directly proportional to V. In steam and gas flows the Mach number often exceeds 0.1. Here, t is not exactly proportional to V but V is still calculable from the two transit times. The V is a line average along the ultrasonic path. In order to accurately take into account the flow profile, considerable care is exercised in the choice and weighting of the paths. The most general solutions available today use multipath sampling of the cross section. In custody transfer spoolpieces for large pipes subject to nonideal inlet conditions, it is not unusual to use four or five paths.
The High-Temperature Problem
Most types of ultrasonic flowmeters use piezoelectric transducers to generate and detect the waves. Just as special magnetic materials have been developed to retain their magnetic properties at high temperatures, so too have high Curie point piezoelectrics been developed that retain their dielectric and other properties. Lithium niobate, for example, remains piezoelectric to ~1210°C. But when the designer attempts to build a practical high-temperature transducer using this material, several technical problems emerge:
While other ferroelectric or piezoelectric ceramics exist or are under development that retain their oxygen to the 300°C 500°C range or even higher, one or both of the other problems remain.
Buffer Waveguides: One Solution
In seismology, workers concerned with earthquake waves refer to P and S waves (primary and secondary waves). The terminology dates back to the 1800s, perhaps earlier, and is understandable if we note that after an earthquake, at a recording site far away, two waves are received, separated in time. First comes the compressional wave, then the shear wave. If the Earth were made of steel, and if the effects of temperature and pressure are neglected, the P waves would travel at ~6 km/s and the slower S waves at 3 km/s. From the difference between arrival times, t, the distance to the source is calculable. In a hypothetical "steel Earth" example, if t were 100 s the range would be 300 km. Data at two or more stations allows us to triangulate and locate the source. In a few moments we will see why the ultrasonic equivalents of P and S waves are important in measuring fluid flow at high temperature. Briefly, the P waves correspond to the waves used in wetted buffers; the S waves, to the waves used in clamp-on buffers (see Figures 24 and Photos 16.)
More than 50 years ago, a relatively simple way to avoid having high temperatures at the transducer was already known to ultrasonic researchers. This solution was the buffer rod. The rod was usually solid metal, sometimes steel, sometimes stainless steel, and occasionally tungsten. The diameter was typically 1225 mm. The outside was threaded or grooved to break up unwanted mode conversions. This solution worked well at high ultrasonic frequencies, say 2
If the industrial flow problem concerns gases or other media at high temperature, where only low frequencies (0.11 MHz) have a chance of successfully penetrating the fluid, and if the application does not allow a big port, then we must find a small-diameter, low-frequency buffer solution. By "small diameter" we mean that the buffer outside diameter should be in the 12.733.4 mm range and preferably not exceed 25.4 mm [2,3].
In 1995, confronted with the problem of buffering ultrasonic transducers subject to the constraints of small-diameter nozzles, high temperatures (up to 450°C), and
The resulting bundle forms a pressure boundary suitable for ultrasonic flow measurements of hot corrosive gases. It also works for hot liquids such as molten asphalt, viscous liquids at room temperature requiring frequencies below 1 MHz, and cryogenic fluids that have a tendency to flash. The low frequency penetrates the fluids despite their two-phase character. The KEMA-approved transducer is removable from the buffer, without breaking the pressure seal (see Figure 3B). The portion of the BWT assembly within the insulated nozzle operates uncooled, at the fluid process temperature. The external portion can be as long as necessary to convection-cool down to a temperature such as 100°C, well within the range of ordinary piezoelectric materials and constructions.
The first industrial application of the BWT transducers was in the Shell Per Plus refinery near Rotterdam (see Photo 1). Buffers housed in stainless steel, titanium, or special alloys were used initially in about 15 locations. Some included startup complications where first nitrogen, then hydrogen-rich (low-density) gas had to be measured from as low a pressure as possible (~10 bar) up to an operating pressure of 209 bar, at temperatures >200°C, and furthermore withstand 450°C during upset conditions. The BWT bundle buffer worked well, so its use was expanded to many other locations in the plant, and then to the next plant Shell built in the same area, Moerdijk, for its MSPO-2 project. The BWT buffered transducers were subsequently used in steam tests conducted by Dow in Terneuzen, the Netherlands, at T = 340°C, 34 bar, ±45 m/s steam flow. By mid-1999 there were over 100 BWT systems installed in gas and liquid flow measuring applications around the world (see Table 1).
Around 1996, EDF (Électricité de France) initiated a project at Panametrics to develop high-temperature clamp-on transducers that could be used to measure the flow of superheated water in steel pipe (see Photo 2). The temperature was 260°C; the pressure, 60 bar. Clamp-on transducers for contrapropagation flowmeters generally introduce the ultrasonic waves obliquely. In theory, one way to do this could be to use the BWT bundle at an angle. However, as the compressional wave velocity was 5 km/s in the steel BWT buffer, the refracted angle in the hot water would be too small for most applications. The small refracted angle is a consequence of the Snell law of refraction, applied to hot water where the sound speed is only about 1 km/s.
Instead, suppose we select a shear wave as the incident wave, in order to have a slower incident velocity (3 km/s, such as in the seismology example) and a proportionately larger refracted angle in the hot water. The shear wave theoretically offers the possibility of a refracted angle larger by a factor of ~5/3 than that achievable using incident compressional waves at the same oblique angle, say 60°. Only one set of problems now remained: how to produce shear waves, obliquely incident, having a shape easily timed to nanosecond or subnanosecond precision, and at frequencies from 0.5 MHz to 4 MHz?
The answer turned out to be another waveguide. This time, instead of using many thin wires, we used one thin plate. The plate has certain topological features for generating reference echoes or for measuring temperatures at corresponding points, but overall it resembles a small hockey stick (see Figure 4). So "hockey stick" became its name and OKS its abbreviation. The OKS waveguide in Photos 36 has a length of ~250 mm and a thickness of 6.4 mm. Of this length, the 75 mm nearest the pipe is insulated. The rest, like the BWT system outside the nozzle, is cooled by convection so that the temperature to which the piezoceramic is exposed is <100° C.
Examples of appropriate applications for the OKS are given in Table 2, organized as a function of pipe diameter. The liquids were superheated water (260°C271°C, 60 bar) if not otherwise identified. Table 3 (page 46) summarizes BWT and OKS buffered waveguide transducers and offers guidelines on where to use them.
More than 20 years ago, H. Karplus at Argonne National Laboratories demonstrated soft metal foil as an effective long-term (>5 yr.) couplant in ultrasonic flowmeters, in his work to 600°C on liquid sodium. The best choice for the foil material for a particular application is dictated by technical and cost factors. If the foil's melting point is too close to the operating temperature, intergranular corrosion is a technical concern.
In tests of short duration (1 hr), satisfactory economical couplants might be aluminum foil or anti-seize grease. For permanent coupling (usually desirable because of the inconvenience of recoupling at high temperature and perhaps in a radioactive area), foil metals other than aluminum are generally preferred. A low melting point foil was found to be easy to install and use at cryogenic temperatures, taking advantage of the OKS to solve a clamp-on flow problem that previously had been quite troublesome. Until the foil and OKS were introduced, this job at a Korean liquefied gas facility (150°C) required recoupling at frequent intervals, especially following thermal cycling.
The metal foil is normally supplied as a strip 6.5 mm wide by 100 mm long, bent to conform to the OKS foot shape and secured with two small strips of electrical tape around the foot just before use. After installation, the tape strips are either removed or, at high temperatures, simply burn away, but they've done their job. The foil is already clamped in place, using a 1000+ kg force that is readily obtained by torquing the pressure screw that is part of the clamp in Photo 5.
Again, for quick trials, anti-seize couplant suffices; because coupling pressure can be modest, a "room temperature" clamp such as shown in Photo 3 suffices, aided by an aluminum block that centralizes the OKS. To minimize crosstalk we have found it best to place OKS transducers on opposite sides of the pipe. Sometimes this is an inconvenience, and sometimes there can be undue responsiveness to crossflow. If the latter is a problem, then two sets of transducers are used in an X configuration. The average is the desired V. The difference yields the crossflow velocity VX .
Sometimes the pair of OKS transducers is to be installed at room temperature and the metal foil couplant isn't soft enough to couple at reasonable coupling pressure. A solution in this case, first demonstrated by my colleagues P. Kucmas and P. Fredriksson in a 1998 startup at Forsmark Unit 1 in Sweden on 10 in. stainless steel heavy wall pipe, was to wet the metal foil with a drop of glycerine couplant. This conventional ultrasonic couplant works fine at room temperature and evaporates later, at high temperature, leaving no residue. The same glycerine solution was used during a scheduled shutdown in a 1999 "cold" startup on 3 in. schedule 40 carbon steel blowdown pipes at a U.S. nuclear plant. When the temperature later reached 271°C, coupling was provided by the soft metal foil.
As another approach to finding a permanent couplant, consider welding. The OKS, typically made of CS1018 but in principle manufacturable of any standard engineering material such as 316SS, can be held in place against the pipe with one of the standard clamps and tack welded, after which the clamp is removed. Full-penetration fillet welds along both sides of the OKS foot then provide complete coupling that is unlikely to decouple over the life of the pipe (see Figure 2C). Weld-on OKS transducers and welded-on yokes (instead of removable clamps) might make the OKS system easier to use in some applications, but welding to an operational pipe is not allowed in many cases. Code and safety issues must be addressed for the particular case at hand.
Hybrids--Calibratable Spoolpiece with Removable OKS Transducers
Hybrids combine features of wetted and clamp-on transducers. Hybrid spoolpieces could use welded-on OKS transducers or welded-on yokes. Recall that the BWT wetted buffers, originally developed for high-temperature flow problems, ended up solving some problems at room and cryogenic temperature as well.
First, consider that the KEMA-approved transducer portion of the BWT system is removable while the buffer itself remains in the pipe nozzle as the permanent pressure boundary. This is not unlike the PanAdapta precision plug concept shown on page 168 in . Some 1800 were in use by mid-1999, mostly for ordinary temperature. Referring again to the BWT buffer system in Figure 3B, the KEMA portion is sort of a clamp-on. (In fact it is normally epoxied in place at the end of the BWT buffer, but can be removed using two wrenches and later reinstalled or replaced.)
Second, consider the introductory remarks about multipath for a more general solution under bad flow conditions. Let us now observe that each pair of thin OKS transducers shown in Figure 2A and Figure 4 can be arranged to interrogate a narrow, well-defined region. Four OKS pairs can therefore provide a Gauss-Chebyshev (GC) quadrature solution to integrating flow over the cross section of a calibratable spoolpiece. This concept, illustrated in Figure 5 and Photo 7, was tested in water at room temperature by my colleague B. Dean last year and yielded an error band of ±0.33% in its first test over the Reynolds number range of 0.47 million to 1.3 million .
This accuracy by itself is not remarkable. But it was achieved with no holes for the multipath transducers. All the transducers were removable clamp-on OKS types. Given that the internal notches for this design are triangular and penetrate radially only about 5 mm or less, depending on location, it seems reasonable to surmise that they disturb the flow less than the conventional ports of 30 mm or so associated with transducers of 25.4 mm, for example. Although this hybrid spoolpiece was tested only at room temperature, its design is not limited to room temperature. In those tests, incidentally, B. Dean used another form of the OKS (see Photo 4, center), clamped in a plane perpendicular to the spoolpiece axis, to measure swirl. In that configuration the OKS transducers are referred to as swirl sensors. Recognizing that the swirl sensors interrogate in a thin sheet perpendicular to the pipe axis, much as the OKS hybrids interrogate in the thin GC planes, there is a possibility for sensing liquid level in the swirl sensor plane. Laboratory swirl and liquid level test results are shown on page 380 in .
Of these three spinoffs--hybrid GC spoolpiece, swirl sensor, and liquid level--R&D activity is currently aimed at the first. Although the OKS transducers are released and on the market, and work with commercially available flowmeter consoles such as the Panametrics PT868, DF868, and SMT868, unreleased R&D items like the swirl sensor or the notched GC spoolpiece with the conventional four parallel planes of measurement and Vee paths, but no holes, are not yet available except on special order. (The '868 instruments are the work of my colleague S.A. Jacobson and his staff.)
To obtain the best results with clamp-on, it is recommended but not absolutely required that the pipe be flat and smooth under the OKS foot. As the OKS transducers are normally manufactured with flat feet, the pipe would ideally have two parallel flats at 180°, say at 3 and 9 o'clock. Flats, however, are not absolutely necessary, as the metal foil can be supplied thick enough to make up the theoretical gap. Those who have used ultrasonic clamp-on flowmeters at room temperature know that the gel or grease couplant supplied by the manufacturer similarly fills up the air gap between transducer and pipe. What's different about the metal foil is that is doesn't dry out, doesn't get washed away, and doesn't run away when it gets hot. Early tests at EDF, for example, during an eight-month high-temperature trial on one pipe, showed no change in coupling. Similar results have been reported for the cryogenic Korean application, which has been running with no maintenance for six months since it was started up early in 1999.
BWT, PanAdapta, and Transflection are trademarks of Panametrics, Inc.
Contributors whose transducer design or field test work is reported here include: A. Arden, D. Bruinzeel, M. Capobianco, E. Chérifi, C. Frail, J. Freeke, G. Jossinet, O. Khrakovsky, R. Koch, Yi Liu, E. Machado, J. Marchi, J. McGregor, J. Matson, R. McCarey, W. Mellish, T. Nguyen, J.K. Park, R. Richard, T. A. Russell, J. Slaughter, D. Thomson, D. Xiao, and M. Zimmerman. The author acknowledges Panametrics's permission to use excerpts from its copyrighted report UR-242S.
1. C.K. Jen et al. 1997. "Clad Buffer Rods for In-situ Process Monitoring," Proc 1997 IEEE Ultrasonics Symp:801-806.
2. Y. Liu et al. 1998. "Buffer Waveguides for Flow Measurement in Hot Fluids," Ultrasonics, 36 (1-5):305-315.
3. L.C. Lynnworth and V. Mágori. 1999. "Industrial Process Control Sensors and Systems," E. P. Papadakis (Guest Ed.), Ultrasonic Instruments and Devices: Reference for Modern Instrumentation, Techniques, and Technology, 23 in the series Physical Acoustics, Academic Press:275-470.
4. L.C. Lynnworth. 1989. Ultrasonic Measurements for Process Control--Theory, Techniques, Applications, Academic Press.
5. B.J. Dean and L.C. Lynnworth. 1999. "Hockey Stick Developments Extend Range, Improve Accuracy of Clamp-On Flow Measurements," Proc 1999 IEEE Intl Ultrasonics Sym, Paper PG-7.
6. R.C. Asher. 1997. Ultrasonic Sensors for Chemical Process Plant, Institute of Physics Publishing, London.
7. L.C. Lynnworth. 1979. "Ultrasonic Flowmeters," Physical Acoustics--Principles and Methods, W.P. Mason and R.N. Thurston (Eds.), Academic Press 14:407-525.
Larry Lynnworth is a VP & PCI R&D General Manager, Panametrics, Inc., 221 Crescent St., Ste. 1, Waltham, MA 02453-3497; 781-899-2719, x-146/307, fax 781-894-5785, lynn firstname.lastname@example.org