Flow-Viz – An integrated digital in-line fluid characterization system for industrial applications

Valentino Meacci, Stefano Ricci Information Engineering Dept., University of Florence

Florence, Italy

Johan Wiklund SP-Technical Research Institute of Sweden,

Gothenburg, Sweden johan.wiklund@sp.se

Beat Birkhofer Sika Services AG,

Zürich, Switzerland

Reinhardt Kotzé FPRC, Cape Peninsula University of Technology,

Cape Town, South Africa

Abstract—The continuous monitoring of rheological parameters

of industrial fluids during production is of paramount importance for process and quality control. Up to now, no system capable of a complete and non-invasive in-line measurement is commercially available, so that only time discrete laboratory measurements on fluids specimens are possible. In this work a new, fully integrated ultrasound system for in-line fluid characterization, named Flow-Viz, is presented. The system measures the velocity profile of the fluid moving in a pipe through pulsed Doppler ultrasound, and combines it with the pressure drop. The electronics, featuring two ultrasound transmission/reception channels used alone or in pitch-catch configuration, includes powerful digital processing capabilities for real-time velocity profile calculation, and is fully programmable. Particular attention is paid to low-noise design for achieving the optimal performance in highly attenuating suspensions. An application is presented where the system, coupled to a non-invasive ultrasound sensor unit, performs in-line rheological measurements through the wall of a high-grade stainless steel pipe.

Keywords—Ultrasound, Doppler, Pulsed Ultrasound Velocimetry (PUV), rheometry, industrial process monitoring

I. INTRODUCTION

Typical industrial fluids are multiphase systems that are transported via pipes between process steps within a plant. For industries, understanding the fluid behaviour and flow dynamics is fundamental in optimizing such processes [1]. Moreover, flow properties such as the shear rate dependent viscosity, are directly linked to product quality and therefore represent important control parameters. The available measurement methods, such as rotational rheometers, are invasive and the measurement is limited to a specimen that is analyzed off-line. A few in-line rheometers have been tested for a small number of industrial applications, but since they are typically able to determine the viscosity only for a single shear rate, their success was limited. Until now, no practical in-line solution exists for non-Newtonian and opaque industrial fluid [2].

In this work we present an efficient, fully programmable and integrated system for in-line fluids characterization of a wide range of non-Newtonian and opaque fluids. This system, named

Flow-Viz exploits ultrasounds to detect the velocity profile of the flow moving in the pipe. The profile data, combined with pressure measurements, allows an accurate rheological characterization according to the PUV+PD method [3]. In particular, the system embeds all of the electronics required for the conditioning, acquisition and processing of the ultrasonic and pressure signals and includes the real-time calculation of the full flow velocity profile. It is specifically designed for industrial use, and represents the second generation of a previous experimental apparatus.

In the following sections, the measurement principle of the in-line rheometer is explained, The Flow-Viz system is described in detail and its successful application in a highly concentrated suspension is demonstrated. The main advantages with the integrated PUV+PD system are that it allows non-invasive, continuous measurements of the velocity profile and rheology of the investigated fluids. It is the only commercially available integrated system where signals from all sensors are processed on the hardware. This vital information can be used to detect process or product anomalies and to do rapid adjustments of the processing parameters. This can lead to improved product quality as well as cost savings, e.g. via reduced energy and water consumption.

II. METHOD

A. Pulsed Ultrasound Velocimetry

In Pulsed Ultrasound Velocimetry (PUV) a burst St(t) is transmitted every Pulse Repetition Interval (PRI) into the medium. When moving particles are encountered, the burst produces an echo affected by a frequency shift that correlates to the particle axial velocity component, according to the Doppler effect: = 2 (1)

= | | cos (2)

978-1-4799-7250-0/16/$31.00 ©2016 IEEE

This full text paper was peer-reviewed at the direction of IEEE Instrumentation and Measurement Society prior to the acceptance and publication.

where fd is the Doppler shift frequency, is the angle between the ultrasound beam and the flow direction, is the axial component of the velocity , ft is the transmitted ultrasound frequency and c is the sound velocity in the medium. The spatial velocity distributions along the axis of an emitted pulse beam can be obtained by measuring the Doppler shift at hundreds or thousands of depths aligned along the beam. For a single scatterer, which is moving at an axial velocity , the analytical description of the echo signal Se(t) is: ( ) = ( ) ( t− D) + 0 (3)

Where = 2 , = 2 , is the initial phase and ( ) depends on the transmission signal St(t), the attenuation of the medium and the system’s impulse response. The Phase-Quadrature (IQ) demodulation is applied to the analytical echo signal for shifting the frequency and to obtain the complex base band signal ( ): ( ) = ( ) ( ) = ( ) ( ∙ ) (4)

In each PRI, the in-phase (I) and quadrature (Q) components of ( ) are stored in a matrix by column. When a sufficient number of PRIs are stored, the data are read out by row, which represents the Doppler samples collected at same depth. The sum of the square of the spectrum analysis, is performed on the weighted read out data, in order to obtain the power spectrum ( ). A weighted mean is performed on every spectrum to obtain a single Doppler frequency from each depth: = − ( ) ∙ ∙ d( ) ∙ (5)

The velocity can be finally obtained by applying (1) to the measured fpm .

B. Rheology assesment

In a laminar pipe flow the shape of the velocity profile, i.e. the velocity distribution across the pipe diameter, depends on the rheological properties of the fluid. Newtonian fluids (viscosity independent of the shear rate) such as oil or water have a simple parabolic velocity profile (Fig. 1, left) while complex fluids (suspensions, etc.), with a shear rate dependent viscosity, develop different profile shapes. For example, shear

thinning fluids present a flattened profile (Fig. 1, middle) and fluids with a yield stress produce a plug flow (Fig. 1, right).

For simple Newtonian fluids, the rheology is often determined from the volumetric flow rate and the pressure difference over a fixed distance of pipe. This method is known as the tube viscometer concept and is capable of determining the shear viscosity at a single shear rate. The shear stress τ along the radial position , which is zero at the centerline of a pipe with the radius can be calculated with the pressure drop Δ measured over the distance L: ( ) = ∙2 ∙ ∙ (6)

For non-Newtonian liquids, which constitute 95% of all industrial fluids [2], the shear viscosity must be determined as a function of shear rate. In tube viscometry, this means that the measurements must be performed in several different pipe diameters or at several different flow rates. In the enhanced tube viscometer concept, such as the PUV+PD method used in this work, the flow profile shape is used to derive the shear rate dependent viscosity at the different shear rates present in the fluid. The shear rate along the radius is the derivative of the flow velocity . ( ) = − dd (7)

Combining the information from the measured pressure drop and flow velocity profile it is possible to calculate the shear rate dependent viscosity in real-time: ( ) = ( )( ) (8)

Some fluids, for example skin cream, show a more solid like behavior under very low shear. A corresponding flow velocity profile is shown in Fig. 1 (3). The yield stress , which is the minimum shear stress for the fluid to flow, can be determined from the plug radius ∗ also measured with the flow velocity profile.

∗ = 2 ∙ ∙∆ (9)

Fig. 1. Three different pipe flow velocity profiles. From left to right: (1) parabolic flow profile of a Newtonian fluid (viscosity independent of shear rate) (2) shear thinning fluid with flattened profile and (3) fluid with yield, resulting in a plug flow.

Rr

vr

P2P1 LΔP

R*

Fig. 2. The Flow-Viz operator’s panel.

1 2 3

III. THE SYSTEM

A. The Flow-Viz fluids characterization system

The Flow-Viz system consists of an operator’s panel and a remote sensor unit. The panel, shown in Fig. 2, houses the proprietary electronics (see next section) and an industrial PC (Beckhoff Automation, Germany) that displays the user interface. The sensor unit holds all of the sensors, is installed in the process network and makes up the measuring section. The measuring section is equipped with at least one pair of unique, custom made, non-invasive ultrasound transducer assemblies (Flow-Viz, Sweden, see Fig. 3), a differential pressure sensor with remote seals (ABB Automation Technology Products AB, Sollentuna, Sweden) and a non-invasive PT-100 sensor (Pentronic, Gunnebo, Sweden). The sensor unit typically comprises also a stainless steel cabinet protecting the heat-jacketed stainless steel pipe with diameters ranging from 10 to 200 mm inner diameter (larger sensor dimensions available upon request).

B. The electronics

The system includes all the electronics necessary for processing and conditioning the ultrasound signal. It is very compact (10 × 12 cm total dimension), low power (5 W max.) and made up of two boards: The Analog Front-End and the Digital Board. The system is connected to a board of the sbRIO family (National Instruments, Austin, TX), which includes an Ethernet network connection. The software running on the industrial PC sets the working parameters and can download the raw echo data, the demodulated data, the spectral matrices or the

final normalized frequency profiles. The main features of the system are listed in Tab. 1.

The analog front-end is subdivided in two equivalent channels, shown in Fig. 4. The transmission section (Tx) of each channel amplifies the transmission burst by a current feedback linear amplifier, which can reach 40 Vpp, and a transformer, that raises the signal up to 80 Vpp, and adapt the amplifier output to transducers impedance. The Tx devices are turned on only during the transmission to minimize the noise and optimize the power consumption.

The receiving section (Rx) amplifies the backscattered echoes and consists of an impedance matching transformer and a single chip, which integrates a Low Noise Amplifier (LNA) and the Programmable Gain Amplifier (PGA), which, together, produce a gain from 7 to 55 dB with bandwidth between 0.8 and 7 MHz. The selection of the Rx and Tx channels is managed by suitable switches, controlled by the Field Programmable Gate Array (FPGA), located on the Digital board.

The Digital Board, shown in Fig. 5, is based on the EP3C25F256 FPGA from the Cyclone family of Altera (San Jose, CA), which manages all of the digital devices present on the board. During the transmission, the burst is generated with programmable amplitude, frequency, number of cycles and tapering, by a Digital Direct Synthesizer (DDS) implemented on the FPGA. Transmission of coded excitations is possible as well. The burst is then converted by a 14-bit resolution, 100 MSPS Digital to Analog Converter (DAC) and sent to the

Fig. 4. Analog Board with ultrasound front-end. Switches are used in TX/RX to manage 2 different transducers.

TABLE I. MAIN FEATURES OF THE ON-BOARD ELECTRONICS

Parameter Value

TX/RX channels 2, multiplexed Dimension 10 × 12 cm Power consumption 5 W max Analog Gain 7 – 55 dB Tx voltage 10 – 80 Vpp TX/RX frequency range 1 – 7 MHz TX burst Arbritary waveform Internal buffer 64 MB Sampling Freq. 100 MHz Processing time 42 μs/depth Input noise on 50 Ohm 1.5 nV√Hz

On board Proc. RF and IQ data, Spectral Matrix, Fequency Profile

Fig. 5. Digital Board. A FPGA manages all of the devices

Fig. 3. The Flow-Viz non-invasive sensors.

analog front end. During the reception phase, the conditioned echoes are digital converted by an AD9265 (Analog Devices, Norwood, MA) at 100 MSPS with 16-bit resolution. The data are then processed in the FPGA like reported in Fig. 6. Samples are coherently demodulated by a multiplication to two, 16-bit resolution, quadrature-phase sinusoidal signals, which have the same frequency as the transmission burst. The resulting 32-bit, in-phase (I) and quadrature (Q) components are filtered by a Cascaded Integrator Comb (CIC) filter with a programmable cut off frequency and down-sampling factor. The filtered samples are stored in the 64 MB SDRAM buffer at 32+32 bit per complex sample. The data storage strategy is shown in Fig. 7, where DL,N is the sample collected at pulse repetition interval (PRI) N and depth L.

When enough data are stored, the multi-gate spectral analysis starts, reading from the SDRAM to the FPGA blocks of 128 complex samples acquired from the same depth. The Hanning window is applied to each block and a block-floating point complex FFT processes the results. The sums of the square of the FFT output, converted to 32-bit floating-point format, allow to obtain the power spectrum. All the available depths are processed and the results are stored in the rows of a Doppler spectral matrix. A programmable number of matrices are averaged for improving the signal to noise ratio before the frequency profile is extracted. From each row of the denoised spectral matrix, the normalized frequency profile nd is calculated with a discrete version of (5):

−

=

−

=

⋅=1

0

21

0

21 N

id,i

N

id,id CCi

Nn (10)

Where Cd,i is the matrix element corresponding to depth d and FFT bin i, and N is the FFT size. The frequency profile is obtained by dividing nd to TPRI, i.e. the temporal length of the PRI and. The frequency profile is moved to the PC where the velocity profile is finally calculated by applying (1) and the rheological parameters are extracted and shown in real-time results.

C. Accuracy of the FPGA calculations

The implementation on the FPGA of the processing chain reported in Fig. 6, was carefully checked for the achievable accuracy. This has been obtained by generating synthetic data in Matlab® (The MathWorks Inc., Natick, MA) that mimic the ultrasound signal received in a typical pipe investigation. The data refer to a 6 mm diameter pipe, insonated at 5 MHz, where a parabolic flow runs at 0.5 m/s. Such data were processed both in the FPGA and in Matlab®. Fig. 8 compares the velocity profile calculated by FPGA (top) to the reference obtained in in Matlab® in double precision maths (middle). The bottom of Fig. 8 reports the relative difference calculated as: ( ) = 20 log | ( ) − ( )|/ ( ) (11)where ( ) and ( ) are the profiles calculated in the FPGA and Matlab®, respectively. The error was always lower than -140 dB, which confirms the correct FPGA implementation of the processing chain.

Fig. 6. Signal processing chain for real-time profile calculation implemented

in the FPGA

Pulse Repetition Interval →

Depth →

D0,0 D0,1 D0,2 …. D0,n-2 D0,n-1

D1,0 D1,1 D1,2 …. D1,n-2 D1,n-1

…. …. …. DL,N …. ….

Dl-2,0 Dl-2,1 DL-2,2 …. Dl-2,n-2 Dl-2,n-1

Dl-1,0 Dl-1,1 Dl-1,2 …. Dl-1,n-2 Dl-1,n-1

Fig. 7. The 64 MB SDRAM of the system holds a matrix where acquired data are stored in columns and read by rows.

Fig. 8. Frequency profile calculated in the FPGA (top) and in Matlab® (middle) starting from the same RF data. The relative difference is

reported on bottom in dB scale.

0 2 4 60

1

2

3

Dop

pler

shi

ft (

kHz)

FPGA

0 2 4 60

1

2

3

Dop

pler

shi

ft (

kHz)

Matlab

0 2 4 6

-150

-100

Err

or (

dB)

Error

D. Experiment on industrial suspension

Unlike other commercially available process viscometers, the Flow-Viz system is designed specifically for non-Newtonian industrial fluids. These fluids cannot be characterized rheologically using conventional instruments as they provide point measurements at a single shear rate, In the Flow-Viz system the velocity data measured from the wall to the center of the pipe is differentiated and the resulting shear rate vector is used in combination with the pressure drop to determine the fluid rheology. For industrial applications, it is therefore important to verify that the Flow-Viz system is capable of measuring a complete velocity profile up to at least the pipe radius.

Anon-Newtonian industrial fluid (sludge) with a particle concentration of 26% (w/w) flowing in a SS316L stainless steel pipe with an inner diameter of 48.6 mm was measured with the Flow-Viz system in order to verify that the penetration depth and data quality is acceptable for rheological characterization of the fluid. The flow rate was set at 6 L/min. The system calculated in-line 32 profiles/s with a spatial resolution of a few µm.

Fig. 9 shows a calculated spectral matrix with the velocity profile overlapped. The penetration depth reached the full pipe diameter even in this highly attenuating (about 5 dB/cm) industrial fluid. The new hardware platform in combination with the non-invasive sensors thus provide improved spatial resolution, improved penetration depth and allows better possibilities to optimize the velocity resolution compared to other commercially available instruments. The measured profile is integrated and the determined flow rate is in general less than 5% from the flow rate determined using Coriolis flow meters.

The corresponding complete rheogram is shown in Fig. 10, where shear rate and shear stress are directly obtained from the measured data. The yield stress is for many industrial applications an important quality control parameter that is difficult to determine using conventional techniques. The yield stress, obtained directly from the measured plug radius, was equal to 8 Pa, which is in excellent agreement with the reference 8,1Pa determined using a stress controlled rheometer. The rheograms provided by the Flow-Viz system is in general within a 5% range when compared with rotational rheometry data. However, it should be noted that in the latter case a sample

must be removed from the process and analyzed in the laboratory. This introduces several sources of error including manual operations, time dependency and different measuring geometries and conditions.

The capability to determine a complete rheogram, continuously and non-invasively in-line thus makes it possible to monitor and control the fluid properties of industrial non-Newtonian fluids, which has also been demonstrated in our previous work, see e.g. [5,6,7,8].

IV. DISCUSSION AND CONCLUSION

This work presents Flow-Viz system, which is an integrated digital in-line fluids characterization system for industrial and research applications that is now commercially available. The electronics features an Analog Front-End and a Digital Board that offers on-board processing and velocity estimation capabilities, and a baseboard with 8 analog and 10 digital I/O DAQ channels for simultaneous acquisition and processing of e.g. pressure and temperature signals. The possibility to acquire data at 100 MHz allows a very high spatial resolution. Moreover, the low noise approach allows to detect the profile along the whole diameter, also in attenuating fluids. The electronics combined with the novel non-invasive sensor technology allows true in-line flow visualization and rheological characterization of opaque, non-Newtonian industrial fluids flowing in industrial grade stainless steel pipes. The main limitation with the technique is that it requires scattering particles to be present in the fluid and it can thus not be applied e.g. to distilled water. The strong attenuation of ultrasound in highly concentrated suspensions could be another limitation and the choice of sensor design thus depends on the application. Moreover, the system provides continuous data that allows the operator to make rapid adjustments of the processing parameters in the plant. However, today there is currently no support for handling the continuous feed of such quality control parameters in most production installations.

The system is a useful characterization tool for field use, pilot plant installations and academic research, and has been successfully tested in a wide range of industrial applications, e.g. oil, petroleum, food, minerals, chocolate, explosive emulsions, pharmaceutical industry.

Fig. 9. Spectral matrix and velocity profile (overlapped in red)

obtained from an industrial fluid, measured at a volumetric flow rate of 6L/min in a pipe with an inner diameter of 48.6 mm.

Fig. 10. Rheogram measured in-line by the system in a non-Newtonian fluid with 26% particle concentration. The yield stress was obtained directly from the measured plug radius and equal to 8 Pa.

0 100 200 300 400 5000

5

10

15

20

Shear Rate (1/s)

She

ar S

tres

s (P

a)

REFERENCES [1] J. M. Dealy and K. F. Wissbrun, “Melt Rheology and its Role in Plastics

Processing: Theory and Applications.” New York, NY: Van Nostrand, 1990

[2] I. Roberts, “In-line and online rheology measurement”, Kress-Rogers & Brimelow (eds), Instrumentation and sensors for the food industry. 2° ed. Woodhead Publishing Limited, Abington Hall, Cambridge (2001), page 403-419, ISBN: 978-1855735606

[3] M. Muller and P.O. Brunn and C. Harder, “New Rheometric Technique: The Gradient-Ultrasound Pulse {Doppler} Method”, Appl. Rheol, Vol.7, no.5, pp 204-210, 1997, http://dx.doi.org/10.3933/ApplRheol-7-204

[4] N. Dogan and M.J. McCarthy and R.L. Powell, “In-Line Measurement of Rheological Parameters and Modeling of Apparent Wall Slip in Diced Tomato Suspensions Using Ultrasonics”, Food Engineering and Physical

properties, Vol. 67, no, 6, pp. 2235-2240, 2002, http://dx.doi.org/10.1111/j.1365-2621.2002.tb09533.x

[5] J. A. Wiklund, B. Birkhofer, M. T. Stading, S.A.K. Jeelani, & E.W. Windhab, “In-line rheometry of particulate suspensions by pulsed ultrasound velocimetry combined with pressure difference method”, Applied Rheology, Vol. 22, no. 4 (2012) 42232. http://dx.doi.org/10.3933/ApplRheol-22-42232

[6] B. Birkhofer, A. Debacker, S. Russo, S. Ricci, D. Lootens, “In-line rheometry based on ultrasonic velocity profiles: comparison of data

processing methods”, Appl. Rheol. , Vol. 22, no. 4, 44701, 2012, http://dx.doi.org/10.3933/ApplRheol-22-44701

[7] S. Ricci, M. Liard, B. Birkhofer, D. Lootens, A. Brühwiler, and P. Tortoli. “Embedded Doppler system for industrial in-line rheometry”. IEEE Trans. Ultrason. Ferroelectr. Freq. Control, Vol. 59, no. 7, pp. 1395–1401, 2012, http://dx.doi.org/10.1109/TUFFc.2012.2340

[8] J. Wiklund, R. Kotzé, B. Birkhofer, S. Ricci, V. Meacci, R. Haldenwang and M. Stading. “Flow-VizTM – A fully integrated and commercial in-line fluid characterization system for industrial applications”. Proc. of 9th Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering (ISUD9), pp. 165-168, 2014, Strasbourg, http://www.isud-conference.org/proc/split/ISUD-09/Wiklund_ISUD-2014.pdf

[9] R. Kotzé, S. Ricci, B. Birkhofer, J. Wiklund, “Performance tests of a new non-invasive sensor unit and ultrasound electronics”, Flow Meas. Instrum., vol. 46, 2015, in press, http://dx.doi.org/10.1016/j.flowmeasinst.2015.08.013

[10] M. Pollakowski and H. Ermert, “Chirp signal matching and signal power optimization in pulse-echo mode ultrasonic non-destructive testing,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. , vol. 41, no. 5, pp. 655–659, 1994, http://dx.doi.org/10.1109/58.308500

[11] M. Vogt “Direct sampling and baseband conversion in Doppler systems for high-frequency ultrasound blood flow measurements.” , Vol. 41, no. 14, pp. 189-790, 2005, http://dx.doi.org/10.1049/el:20051619