Pressure regulators are a major source of noise in gas pipelines. In recent years, continuous optimization of the regulator design has led to a noticeable noise reduction, particularly in the audible range. However, the remaining noise generated may still be problematic for ultrasonic gas flow meters. The reliability and accuracy of the signal transit time determination, and thus the quality of the measured value provided by an ultrasonic gas meter, are defined by the minimum signal level differential required between the sensor sound pulse to be analyzed and the signal interfering with the sound pulse. The spectral distribution of noise, and its dependence on the pressure difference and flow rate at the regulator, is of particular interest. Figure 1 shows a generalized schematic of an installation with regulation.

The impact noise originating from pressure regulators has on the performance of ultrasonic meters is a well-known issue. The installation of expensive noise-attenuating piping configurations is often used to solve the problem. Extensive research and development has taken place to address this problem resulting in a simpler, more cost effective solution.

The issue has been examined systematically both in field tests in the measuring station of a transportation pipeline, and in the E.ON Ruhrgas high-pressure test facility in Lintorf. The objective was to determine the limits of use and potential applications of an ultrasonic gas meter with chordal path layout in combination with a regulator. The tests, utilizing a 16-in. meter in the measuring station, confirmed that accurate measurement can be achieved even under the most extreme conditions.

For further systematic testing at the Lintorf high-pressure test facility, an 8-in. ultrasonic meter was equipped with two four-path systems working independently of each other. This approach made it possible to directly compare a measurement system utilizing 210 kHz ultrasonic sensors with one utilizing the latest technology of 350 kHz sensors. It was found that the 350 kHz sensors were clearly less sensitive to interfering noise signals and therefore improved measurement reliability under worst case conditions. Based on auto-diagnosis parameters, such as signal-to-noise ratio and performance, it was confirmed that the meter was capable of clearly detecting and rejecting disturbed signals. The goal here is to describe the test results and the information derived with respect to an expanded use of ultrasonic technology.

Noise from gas pressure regulators

Energy loss, and the consequential noise generated during pressure reduction, are proportional to the flow rate and pressure difference. These relationships of noise sound pressure are:

(1)

The following approximation of the noise level produced by a sonic nozzle can be found by:

(2)

For verification of this equation, the noise level emitted by a compressed air gun was recorded and the frequency spectra of the real signal and the theoretical model were compared. Figure 2 shows the result of this test. A sufficient degree of congruence in the frequency range in question could be found.

A pressure regulator generates sound waves over a wide frequency range that may well be in the typical working frequency ranges of ultrasonic gas transducers (80-200 kHz). These sound waves travel through the gas and superimpose the ultrasonic sound pulses emitted by an ultrasonic gas meter. It should be noted in this context that the gas industry uses various types of regulators that differ with respect to noise emission behavior. The rough approximations contained herein are only intended to assess the nature of noise generation.

Signal-to-noise ratio

To fully understand this issue, an overview of useful signal (sound burst emitted) to the interfering signal (broadband sound signal of pressure regulator) is required. This is more commonly known as the Signal to Noise ratio (SNR). Basically, sound waves in a gaseous medium always propagate in a directional fashion from their source. The sound pressure at a certain point is proportional to the amplitude of the sound-emitting source and decreases exponentially with the distance l from the source of sound. During its propagation, the sound wave is weakened as a result of interactions with the medium (attenuation a). Sound energy is transformed into thermal energy due to the viscosity of, and heat conduction in, the medium. The attenuation is very dependent on the medium and on the frequency f used. For similar mediums, this relation can be simplified as follows:

(3)

As the sound wave f hits an interface, its energy will be distributed into a different direction. The ratio of wavelength and dimensions of the disturbing object play a major role here. On the one hand, there will be diffraction effects, which is why “one can hear around a corner.” On the other hand, the sound wave can be reflected. The ratio of reflector to transmitter surface area defines the resulting reflection signal loss. Taking into account attenuation, geometric distances and reflection signal losses, the ratio of useful signal to interfering signal can thus be expressed as follows:

(4)

Equation (5) defines the ratio in equation (4) as a logarithmic measure in dB units. If a logarithmic measure is also used for attenuation a, the signal-to-noise ratio can be expressed as follows using equations (4) and (5):

(5)

(6)

Possible solutions

From equation (6) it can be seen that two options exist for the optimization of the SNR, and thereby, signal detection by the meter. These are to minimize the interfering signal, or to maximize the sound level of the useful signal. The interfering noise sound level depends on the type of pressure regulator used. The noise produced can be attenuated by appropriate acoustic measures. To date, engineers have attempted to solve the problem by installing complex and costly spatial pipe arrangements to reduce interfering noise levels. A flow straightener installed upstream also provides considerable attenuation of the noise level. For the PTB-type flow conditioner, an attenuation of 6 dB was recorded across the entire frequency range.

A further improvement of the noise attenuation can be achieved if the flow straightener is combined with metal foam panels. Because of the different thickness and structural density of the metal foam panels, an acoustically selective attenuation system can be created which is adapted to the working frequency of the ultrasonic transducers. This leads to further attenuation amounting to 3–6 dB. Because these structures are always symmetrical, this type of muffler can be used in conjunction with ultrasonic gas flow meters in bidirectional operation.

Path layout

The sound burst emitted from the transmitting ultrasonic sensor is attenuated in the same way as the interfering signal. The geometric distances between the ultrasonic sensors of a measurement path should therefore be as short as possible to ensure maximum useful signal levels at the receiving ultrasonic sensor. It is also obvious that each point of reflection in the measuring path normally further weakens the useful signal level.

The signal level chart in Figure 3 may serve as an exemplary illustration. It shows the signal level passing from the transmitter to the receiver of the ultrasonic measuring path for a single-reflection arrangement in contrast to a direct arrangement. This is a theoretical illustration based on equation (3). The sensor frequency and the angle between measuring path and flow axis is assumed to be the same in both cases. The signal emitted at the position of the transmitter (level A) is attenuated on its way to the receiver. While the direct signal still has about 70% (level B) of its original level in this example when it arrives at the receiver, the bounced signal is further attenuated because it travels twice the distance, and because there is an additional loss at the point of reflection.

The noise level in the received signal consists of both electric noise caused by the signal amplifiers and additive noise signals collected by the receiving sensor. Modern, closed-loop amplifier electronics modules (automatic gain control, or AGC) allow dynamic amplification ranges of 86 dB (1-20,000) to be processed without any limitation through electronic noise.

Ultrasonic sensors

Transducers for ultrasonic gas metering are usually of a piezo-ceramic type. The piezoelectric transducer itself is basically a thin disc. Two different vibration modes can be distinguished:

• the radial vibration mode, and

• the thickness vibration mode.

If an alternating voltage is applied to the electrodes of the piezo-ceramic elements, their geometry will change. This generates a mechanical oscillation with the frequency of the alternating voltage. The maximum usable electric energy is limited because of the intrinsically safe design of the sensor circuits which is required in this specific application. Further, because of the acoustic impedance jump between the oscillating surface and the gaseous medium, only a small portion of the energy is transmitted into the medium. In order to achieve the necessary efficiency of the energy transformation, and to increase the sound pressure transmitted into the gas, the mechanical oscillation amplitude is amplified by a coupled mechanical oscillator.

Due to their simple design, bimorph transducers are widely used. These transducers have an acoustic matching layer which adheres to the ceramic element and performs this energy transformation. This layer is made of epoxy resins using hollow glass spheres and its thickness is dependent on the working frequency of the ultrasonic sensor.

The alternating electrical field excites the piezoelectric disc so that it starts oscillating radially. The radial movement is transformed into an axial movement by the adhering matching layer. Great shear forces must be transmitted by the adhesive layer. In order to protect the epoxy resin of the matching layer from the detrimental effects of gaseous components such as hydrogen sulfide, the layer may be covered by a thin metal foil. However, this leads to a reduction in the amplitude of the transmitted acoustic signal and in the reception sensitivity.

This sensor type is characterized by a spectral feature that consists of a mixture of pure tones which are close to each other. This can be seen by looking at the characteristic of the received signal in the frequency domain. These sensors are therefore also often referred to as broadband sensors. In order to be able to generate maximum sound energy, the sensor is run in the resonance region with the greatest amplitudes (transmitter side). On the reception side, the additional neighboring resonance regions are problematic, and can cause possible noise signals to superimpose the received measuring signal.

The acoustic matching layer could be left out if it were possible to achieve sufficient vibration amplitudes at the sound emitting surface. This concept leads to a stacked piezoelectric transducer in the form of a resonance converter. A metallic spring-mass-system is used to increase the amplitude at resonance. Utilizing numerical optimization of mechanical and electrical parameters it is possible to produce sensors which exhibit:

• Sufficient bandwidth for short signals at great amplitude, and

• A maximum acoustic efficiency.

This sensor concept is characterized by pure tone resonance mode and a well-defined working range. There are several advantages:

• The energy is efficiently transformed into acoustic energy,

• The transducer is hermetically sealed and has a full metal housing, and

• The bandwidth allows relatively short pulse signals.

Figure 4 shows examples of the two different transducer designs.

Signal processing

Generally, the SNR may be improved with the help of signal averaging methods or signal coding. However, specifically in gas flow metering applications, the problem is that the signal path is modulated due to turbulence in the flowing gas. This limits the efficiency of the averaging and encoding methods. According to the signal theory, correlation methods provide optimum results in signal transit time measurements, but they cause great computational load during the digital signal processing.

If the SNR falls below a minimum threshold defined by the signal processing algorithm, faulty measurements of the signal transit time may occur. This must be prevented through adequate monitoring and analysis of the received signal quality, otherwise significant measuring errors of the gas velocity would occur.

Optimization criteria

Based on the previous general explanations, optimization criteria applicable to ultrasonic gas flow meters near pressure regulators can easily be derived:

1. Selection of ultrasonic sensors with a working frequency which is as high as possible because:

• The noise signals emitted by the pressure regulator are significantly weakened at frequencies greater than 100 kHz, and

• The frequency-dependent attenuation of the noise signals at a given distance to the pressure regulator causes lower noise levels compared with lower frequencies.

2. Selection of ultrasonic sensors which work within a very defined frequency range minimizes the collection of undesirable noise signal components.

3. Selection of a suitable path layout in order to ensure a maximum ultrasonic burst signal level.

4. Selection of a signal processing method which:

• Makes only minimum demands on the required SNR, and

• Securely avoids faulty triggering and thus prevents biased measurement results.

The aforementioned requirements were considered in the development of a noise-insensitive ultrasonic gas flow meter (FLOWSIC600). The ultrasonic transducers mounted in the meter are stacked type transducers, which work according to the thickness vibration principle, and are available with working frequencies of 210 kHz and 350 kHz (Figure 5). The path layout is the chordal direct path design with four independent paths which are configured in parallel in one plane so as to cover the entire cross-section of the pipe. This layout also boasts the advantage that it is very insensitive to turbulent flow profiles.

Further robustness is achieved by the signal processing technology in the investigated ultrasonic gas flow meter. A model-based correlation method is used in combination with several plausibility criteria, so that even at a minimum SNR of just 6 dB the position of the ultrasonic signal burst is clearly detected in the received signal.

In the received signal, the signal processing algorithm determines the signal portion which comes closest to the signal model. Thanks to an extensive array of plausibility checks, it can be ensured that the measured value is correct even at a performance of as low as 5% (i.e., 95 out of 100 received signals being rejected). The signal is evaluated with respect to:

• The position in a time frame (not too early or too late),

• The amplitude (not too small or overloaded),

• The SNR (above the minimum required level), and

• The degree of congruence with the model signal.

Only if all of these criteria are met will a threefold transit time calculation be conducted according to various criteria in the signal. At least two of the three calculated transit times must be identical for the result to be acceptable.

Test results

The testing was done on the high-pressure test loop of E.ON Ruhrgas in Lintorf and at an M&R (measurement and regulation) station. On the high-pressure test loop, the already proven 210 kHz sensors were directly compared with the newly developed 350 kHz sensors under conditions very similar to those in the field. At an M&R station, a FLOWSIC600 fitted with 210 kHz sensors was tested under extreme conditions with a regulator installed downstream of a flow meter.

E.ON Ruhrgas operates a high-pressure test facility used for testing and optimizing large volume gas metering instruments. While the Pigsar test rig of E.ON Ruhrgas is used for high-precision calibration and verification of meters with natural gas under high pressure, the Lintorf facility (Figure 6) serves to:

• Test new measurement instruments under near-field conditions,

• Investigate special factors influencing measurement behavior,

• Optimize measurement instruments and other components,

• Solve operational problems, and

• Examine new measurement
technologies.

The configuration of the test facility is shown in Figure 7. The pressure is controlled at the inlet to the test facility while the volumetric flow rate can be adjusted at the outlet using a flow control valve. The working standards (test rig standards) used are five parallel meter runs, four of which are orifice plate meter runs (DN 200) built according to ISO 5167 and calibrated with high accuracy. The other is a DN 150 meter run fitted with a turbine meter and an ultrasonic meter. The working standards provide reference values for the meters and pressure regulators to be tested. A turbine flow meter (DN 300), which is permanently installed upstream of the working standards, and an ultrasonic flow meter (DN 300) permanently installed downstream of the test run are used for investigating long-term stability and for quality control purposes.

On the test rig, an RMG regulator was installed upstream of the 8-in. ultrasonic meter tested. The distance between the regulator and ultrasonic meter was 15D. The regulator used was fitted with a sound damper to reduce audible sound. This regulator is normally always fitted with a sound damper. For test purposes, the systems were examined both with and without the sound damper.

The ultrasonic meter tested had two independent measurement systems installed in one meter body. The measurements were made with a four-path 210 kHz sensor system as well as with a four-path 350 kHz sensor system. The tests with the 350 kHz system were completely independent of the 210 kHz system tests to allow direct comparison of the two systems. This arrangement ensured that both systems were operating under identical conditions.

A FLOWSIC600 ultrasonic gas meter with standard sensors (210 kHz) was installed near a regulator in a measurement and regulation station. The risk
was that the pressure regulator would
produce noise interference with the ultrasonic meter because of the very high flow velocities and pressure differentials across the regulator.

The ultrasonic meter was 16-in. and the downstream regulator was 20-in. A flow rate of up to 17.6 MMACF (500,000 m3 (n)/h) could be set for the meter run at an operating pressure of 725 to 1230 psig (50 bar to 85 bar) with a pressure reduction across the regulator of 0 to 435 psig (0 bar to 30 bar).

Several flow rates and pressure differentials across the regulator were set for the tests. The deviation between the vortex meter and the ultrasonic meter were noted. The diagnostic parameters of the ultrasonic meter were also recorded to better analyze the influencing effect of the regulator on the ultrasonic meter. The parameters

• Relative number of faulty signals
(performance), and

• Calculated signal-to-noise ratio (SNR)

proved very useful for determining and assessing regulator influence.

A microphone was installed between the regulator and the ultrasonic meter to measure the noise in the gas line. It could be seen that flowing velocity has a significant effect on the noise level produced by the regulator. In the frequency range of the 210 kHz ultrasonic meter sensors, the interfering noise measured is between 128 dB and 136 dB.

It should be noted that different regulators have different characteristics with respect to interfering noise emission, in particular in the ultrasonic range (100 - 400 kHz). Prior to use, the interfering effects of a regulator should be investigated. The results described here only apply to the regulator examined here and the FLOWSIC600. The tests did not examine the significance of meter size (nominal diameter). The tests focused on the more critical application where the regulator is installed upstream of the meter. Under these conditions, the meter was pushed to its limits. Compared to the proven 210 kHz sensors, the newly developed 350 kHz sensors improved the meter’s ability to deal with interfering noise in the pipelines and good results were obtained under the extreme conditions tested.

The 350 kHz system did not reach its limits until it was installed and tested downstream of a regulator without a sound damper. However, this situation is very unusual as the regulator is normally always used together with a sound damper. The 350 kHz sensors will be optimized further and tested again on the Lintorf test rig. The solution with the 350 kHz sensors is an attractive alternative to the use of noise-reducing devices such as sound dampers, flow straighteners or complex piping. Develop-ment work will therefore be continued. n

Acknowledgment

Based on a paper presented at American Gas Association Operations Conference held in Phoenix, Arizona, May 14-16, 2008.