Printer Friendly

Flow regime determination in horizontal hydrotransport using non-intrusive acoustic probes.


The oil sands located in northern Alberta are a valuable source of bitumen that can be recovered and processed into a synthetic crude oil used in the production of fuels, petro-chemicals, and lubricants (Valenti, 1998). Many advancements have been made in the oil sand industry, which includes a reduction in the number of draglines and belt conveyors in the mine, and an increase in the use of shovels and hydrotransport pipes to carry the raw materials to the processing plant (Fair et al., 1999).

Transporting solids through pipes is a practical and economic way to move large quantities of material between locations. It is important to know the flow regime of the slurry travelling through the system. Pipes are susceptible to erosion and wear, and require costly downtime for repairs. Knowledge of the flow regime and system monitoring can ensure that equipment meets the expected lifespan, saving money, and reducing downtime, as well as meeting product output expectations.

The objective of this paper is to present early detection methods for solid deposits at the bottom of a horizontal hydrotransport pipe. Since abrasive conditions in the pipe would damage invasive probes, non-invasive sensors are required. External microphones were used as the non-invasive sensors and their signals were analyzed using advanced techniques.


This section provides background information on hydrotransport and the flow regimes encountered in hydrotransport, acoustic emissions and their detection, and on the signal analysis techniques used in this study.

Horizontal Hydrotransport

Hydrotransport is the process of transporting a slurry, which is a mixture of water and solid particles. Slurries share many properties with liquids. However, the effective density and viscosity of the slurry are affected by the presence of solid particles (Shook et al., 2002).

There are two types of slurry systems: homogeneous and heterogeneous slurries. Very fine particles result in homogeneous slurries, where settling does not occur quickly and particles are distributed uniformly within the cross-section of the pipe. Heterogeneous slurries occur for particles with a high settling velocity, and segregation occurs along the bottom of the pipe (Gilies et al., 2004).

Figure 1 shows the different flow regimes encountered in hydrotransport. Figure la shows a homogeneous slurry where all solid particles are suspended, with a uniform concentration across the pipe. As the slurry velocity decreases or the concentration of solids increases, heterogeneous flows occur, which include a higher concentration of solids towards the bottom of the pipe (Figure lb), or moving (Figure lc), or settled (Figure Id) beds of particles on the bottom of the pipe.

Hydrotransport Flow Regime Characterization Methods

Pressure drop measurements have been used on hydrotransport pipes to identify flow regimes. Factors affecting the pressure gradient have been examined by several researchers such as Newitt et al. (1955), Maruyama et al. (1979), Lin (1982), Doron and Barnea (1993, 1996), Doron et al. (1997), Gillies et al. (1999), Skudarnov et al. (2004), Greenspan and Nigam (2001), Kaushal and Tomita (2002), Kaushal et al. (2005), and Matousek (2002).


The change in the pressure gradient with velocity can be used to identify the flow regime in a horizontal line. For a specific solids concentration, there is a corresponding velocity at which the pressure gradient is at a minimum, indicating the transition between flow with a stationary bed and fully suspended flow. Suspended particle flow results in an increased pressure gradient with velocity (Newitt et al., 1955). Skudarnov et al. (2004) determined that the critical velocity for the formation of moving and stationary beds are directly related to increased solids concentrations, as well as to increasing specific gravity of the transported material.

Maruyama et al. (1979) studied the hydrotransport of glass beads in horizontal pipes using visual observations and pressure measurements. A modified Fronde number was developed based on the depth of the solid deposit and the average slurry velocity to determine three flow regimes consisting of stable, unstable, and neutral flows. Pressure drop measurements were used to determine the flow regime as a function of the vortex frequency, which indicates flow stability. Many vortices resulted in suspended solids flow whereas few vortices led to the formation of solid deposits.

Greenspan and Nigam (2001) considered the effect of multi-sized particles in pipe flow. They determined that smaller particles have longer and random trajectories and move around the larger particles in a slurry. It was also determined that large particles migrate towards the centre of the pipe, whereas small particles are suspended to form a high viscosity fluid with settled fine particles at the wall. Kaushal et al. (2005) determined that fine particles experience a smaller pressure drop at low velocities and a larger pressure drop at high velocities compared to coarse particles. This was thought to be due to increased settling of coarse particles at low velocities, and a greater surface area of fine particles at high velocities leading to a higher frictional pressure drop in suspended flow. Matousek (2002) used pressure measurements to examine flow regimes using mixtures of fine, medium, and coarse sand particles. It was found that slurries with settled solids deposits had higher frictional pressure drops than slurries that contained shallow settled beds or suspended flows and the addition of finer sands to coarse sand reduced the frictional pressure drop.

Doron and Barnea (1996) used various models and visual observations of tracer particles to develop flow regime maps. Flow maps were developed for conditions including different solids densities, pipe diameters, and particle diameters. These maps show that the transition from settled bed flow to suspended flow occurred at higher flowrates for dense solids, and higher transition velocities result for larger pipe diameters and larger particle diameters. Doron and Bamea (1993, 1997) also examined flow regimes and critical deposition velocities in horizontal, upward, and downward inclined pipes using pressure drop measurements. It was determined that as the pipe inclination increased or decreased, the pressure drop increased or decreased, respectively.


Acoustic sensors can be used over a wide range of process conditions. They are inexpensive and provide reliable, on-line and non-intrusive monitoring. These sensors are easy to implement: microphones are attached to the outer surface of the pipe wall and can be easily relocated. There are two acoustic monitoring methods:

* active acoustics detect the effect of the process on a transmitted ultrasonic acoustic wave; and

* passive acoustics detect the acoustic emissions generated by the process.

Implementation of passive acoustic methods is generally much easier than active acoustic methods, and passive methods are preferred when the process acoustic emissions are strong, as is the case with hydrotransport.

Acoustics is the study of sound. Sound is created by pressure vibrations that can be detected by the human ear or microphones. In processes that involve solid particle motion, acoustic emissions are caused by particle collisions with each other, vessel walls or other objects (Boyd and Varley, 2001).

The microphones used in this study were prepolarized electret condenser microphones. Sound pressure fluctuations are transformed into capacitance variations, which are further converted into an electrical voltage signal. The microphones generate an oscillating voltage signal proportional to the original pressure oscillations (Valentino, 2005).

Signal Analysis Methods to Determine Flow Regimes

Hou et al. (1999) studied the power spectral density of passive acoustic signals acquired from a slurry system. It was determined that spectral peaks occurred at regular frequency intervals, implying that one basic frequency component existed in the spectrum and multiple peaks were harmonics of this basic frequency component. This frequency was determined to be the oscillation of the pipeline imposed by the slurry flow, which was distorted by the solids concentration and slurry velocity. Variations in spectral peak amplitudes provided a method to indirectly monitor the slurry flow regime.

Takahashi et al. (1989) used pressure fluctuations to examine dune formation in horizontal pipes. Pressure fluctuations were caused by dune movement, and the velocity of the dune increased with an increase in the slurry velocity. The power spectral density of the pressure fluctuation signals showed the existence of a dominant frequency, which was attributed to movement of the dunes.

Stienstra et al. (2005) successfully developed an early warning flow instability monitoring technique using pressure fluctuations and time series attractor comparison. An optimal operating condition time series of suspended solids flow was chosen as a reference signal to which measured time series were compared. Attractor comparison allowed for the S-Statistic value to be calculated; a S-value greater than 3 indicated deviation from the reference signal as solid deposits began to form.

Briens et al. (1999) and Briens and Briens (2002) developed a variety of methods for the detection of flow regime transitions in different multiphase systems based on the V-Statistic. Albion et al. (2007a,b) developed a method to determine the flow regime in horizontal and upward inclined pneumatic conveying pipelines based on the V-Statistic value. Acoustic probes were located along the length of the pipe and signals were acquired at 40 000 Hz. The acoustic signals were filtered using a wavelet residual filter and analyzed using the V-Statistic. At a specific subperiod length, there is a critical transition value; signals with a V-Statistic value greater than the transition value indicate suspended solids flow, whereas, a lower V-Statistic value indicate flow over settled solids flow regime. These methods are easy to implement, rapid, and tolerate signal contamination by electrical noise. They are ideally suited to monitor industrial processes.

Acoustic signals have been used to determine the particle size of ore being reduced in a ball-mill grinder based on the root mean square of the power (Breitenbach and Weber, 1999). In this study, the authors determined that as the particle size decreased, the viscosity of the slurry increased, decreasing the power of collisions between particles.

Advanced signal analysis techniques were examined in this work to identify flow regimes in horizontal hydrotransport, which are based on the standard deviation, cycle times, the power calculated through the Fourier Transform, and wavelet octave analysis.

Signal Analysis--Standard Deviation

The standard deviation of a signal is the second moment of the signal mean. Standard deviation is based on the amplitude of the signal fluctuations, and their distance from the signal mean.

Signal Analysis--Power Spectral Density

The power spectral density of a signal is derived from its Fourier transform. The transform decomposes components of a signal into its frequency ranges. The power spectral density reveals periodic cycles present in the signal and quantifies the relative strengths of the periodic components. The power of a signal is calculated from the area under the power spectral density curve.

Signal Analysis--the V-Statistic for the Detection and Characterization of Non-Periodic Cycles

Hurst (1951) developed the rescaled range technique for time series analysis. The time series is divided into time intervals of length [tau], and the data is rescaled to have a mean of 0, with a standard deviation of 1. The rescaled range is calculated over each time interval, and the average of the resealed ranges, [(R/S).sub.[tau]], is calculated for all time interval lengths. By varying the length of [tau] and plotting ln[(R/S).sub.[tau]] against ln([tau]), a curve is produced from which Hurst was able to detect and characterize non-periodic cycles. The ln[(R/S).sub.[tau]] against ln([tau]) curve can have two linear regions. The intersection of the linear segments corresponds to the cycle time of the measured time series. This cycle time, however, can be difficult to determine if the transition is gradual (Briens and Briens, 2002).

Peters (1994) extended Hurst's analysis using the V-Statistic to facilitate the detection of cyclic, non-periodic behaviour in the stock market. The V-Statistic is calculated through:

[V.sub.[tau]] = [(RIS).sub.[tau]]/[[tau].sup.0.5] (1)

and is plotted against ln([tau]). A peak corresponds to the maximum V-Statistic and occurs at the cycle time determined by the intersection of the linear Hurst curve regions. The V-Statistic provides an accurate measure of the cycle length (Briens and Briens, 2002).

Briens and Briens (2002) developed methods to characterize cycle times using the V-Statistic on non-periodic signals, which can consist of a range of cycle times and amplitudes. The regularity of the cycle time defines the width of the range of cycle times. It is calculated through:

[R.sub.T] = [V.sub.T][DELTA]t/[square root of T]. (2)


Signal Analysis--Wavelets

Wavelet transforms are applied to complex signals for denoising, data compression, and feature extraction. The original signal is decomposed into scaled and shifted versions of the original wavelet, resulting in a signal with multiple scales. For examination of low-frequency spectrum components, a stretched version of the wavelet function is used, whereas the opposite is true for higher frequency components (Trygg et al., 2001).

Very rapid signal analysis can be achieved using the discrete wavelet transform. For a signal of length [2.sup.j], filtering is performed j times, resulting in j levels of different scales, each separated by a factor of 2. The wavelet filter produces wavelet coefficients, whereas the scaling filter describes the signal, and the coefficients represent the signal in the next scale. This allows for exact reconstruction of the original signal through the use of the average value and wavelet coefficients. At the lowest octave number, corresponding to the lowest frequency, one coefficient is calculated, and is the average value of the original signal. The original signal can be reconstructed using the average value of the signal and the wavelet coefficients (Trygg et al., 2001).

Signal Analysis--Kurtosis

Kurtosis is a measure of the relative peakedness of a distribution; it is a dimensionless value based on the shape of a single peak (Flott, 1995). Kurtosis is calculated by:

K = [M.summation over (1) [(y - [bar.y].sup.4]/M[[sigma].sup.4] - 3 (3)

where y is the signal value, [bar.y] is the signal mean, [sigma] is the signal standard deviation and M is the number of points in the signal.

Intermittency can be calculated from the wavelet spectra at each octave, and is related to the kurtosis of the coefficients at each octave. Intermittency is determined by a ratio of moments at each octave, whereas kurtosis is the fourth moment (Equation 3) (Farge and Schneider, in press; Echim et al., 2007). Intermittency is calculated by (Farge and Schneider, 2006; Echim et al., 2007):

Q = [q.sub.4]/[([q.sub.4]).sup.2] (4)

Intermittency is a localized burst of low or high-frequency activity in physical and spectra space. A time series is intermittent when the intermittency value is large (Echim et al., 2007).


Hydrotransport System

The hydrotransport system consisted of a 0.05 m inside diameter stainless steel pipe, shown in Figure 2.

Water and solids were initially added to the slurrying tank through an opening at the top of the vessel. Air from a compressor powered the diaphragm pump used to pump the slurry through the pipeline. A magnetic flux flowmeter, located in the vertical pipe, was used to set and monitor the slurry velocity, which was controlled by the flowrate of air supplied to the pump. After exiting the pump, the slurry moved through a 90[degrees] bend and entered the top of the inclined pipe section. The inclined section had a length of 2.75 m at an angle of 30[degrees] from the horizontal. The slurry then travelled through a 1.7 m vertical section and a 3.6 m horizontal section before returning to the slurrying tank. Ball valves below the slurrying tank allowed for drainage of the slurry from the tank and pipeline.

The pump was a source of vibration on the pipeline system. To reduce the transmission of vibration from the pump, sections of pipe at the top of the inclined line and at the end of the horizontal line were replaced with a section of 0.05 m diameter reinforced flexible hose.

Slurry Characteristics

Silica sand, stones, and a 50 wt% sand-50 wt% stone mixture was used to develop the detection methods at various slurry concentrations and velocities. The Sauter-mean diameter of the silica sand was 180 [micro]m with a density of 2650 kg/[m.sup.3]. The terminal velocity of the sand, assuming spherical particles, was calculated to be 0.022 m/s (Shook et al., 2002). The volume-equivalent diameter of the small stones was 2250 [Lm, however, the longest axis ranged from 1000 to 3000 [micro]m. The small stones had a density of 2100 kg/[m.sup.3] and a terminal velocity of 0.172 m/s.

Slurry Flow Properties

Flow regimes in the pipeline are controlled by varying the solids concentration and the slurry velocity. The solids concentration was controlled by the amount of solids added to the slurrying tank. The slurry velocities examined were 0.5, 1, 2, and 3 m/s, with slurry concentrations of 10, 20, 30, 40, and 50 wt% for sand and concentrations of 10, 20, and 30 wt% for stones and the sand-stone mixture slurries.

Acoustic Sensors

Six prepolarized electret microphones model 130D10, equipped with a preamplifier (model 130P10), manufactured by PCB Piezotronics Inc. were used to simultaneously record acoustic signals at different locations along the horizontal pipeline. Data was acquired using a 12-bit National Instruments data acquisition card model NI PCI-6071E. Each measurement was recorded at a frequency of 40 000 Hz. Acoustic sensors were located directly on the top and bottom of the pipe, at each measurement location. Measurement locations were evenly spaced along the pipe at 0.03, 0.50, 1.00, 1.50, 2.00, and 2.50 m in the horizontal section measured from the vertical to horizontal elbow. The microphone locations are shown in Figure 2. Acoustic sensors were secured to the pipeline using shims formed to the shape of the sensor, and attached to a thin foam base. This base was held to the pipe with Velcro wrapped around the pipe circumference.


Visual Observations

A PVC section of pipe in the horizontal pipeline allowed for visual observations of the flow regimes. A video camera was used to visually record the flows at each operating condition. Video files were recorded at 10 frames per second. These files were used to observe particle motion and to visually determine operating conditions which resulted in suspended solids flow or flow over settled solids. The flow regimes determined from the video files were used to verify the results from the acoustic signals.

Experimental Method

For each slurry concentration, 45 L of water were added to the slurrying tank. The pump was started and the appropriate amounts of sand, stones, or sand-stone mixture were weighed and added to the water in the tank. Pumping the water prevented the sand from settling out of the slurry as it was added and kept the slurry mixed. Once the desired sand concentration was reached, the slurry velocity was set. After the flow stabilized, the acoustic measurements were started.


Raw Acoustic Signals

Visual methods were used to identify flow regimes and provided a basis to which the analyzed signals were compared. Figure 3 shows frames from video files recorded at 1.5 m from the vertical to horizontal elbow in the horizontal pipe. Figure 3a shows the flow over settled solids flow regime, whereas Figure 3b shows the suspended solids flow regime. A line was drawn on the video frames to help identify the settled bed from the suspended solids.

Figure 4a and b shows that solid deposits on the bottom of the horizontal pipe cannot be reliably detected by visually examining the raw signal. These figures show raw acoustic signals for silica sand, recorded for flow over settled solids (Figure 4a) and suspended solids (Figure 4b) flow regime conditions. The signals were recorded at 1.5 m from the elbow, on the bottom of the pipe.


Standard Deviation

Figure 5 shows the standard deviation values calculated from the raw acoustic signals for sand, recorded on the bottom of the pipe. There is a distinct separation between the standard deviation values for the water flow only and pump without water and solids in the system, and the values for the flow over settled solids and suspended solids flows. The flow of water through the pipe resulted in the largest standard deviation values. The standard deviation values decreased for suspended solids flow, and further decreased for a bed of settled particles on the bottom of the pipe. Solid particles in the pipe resulted in lower standard deviation values than the water flow, as solids damped the sound generated by the pumped water.



Figure 6 demonstrates that the flow regimes can be successfully determined using the standard deviation of the raw acoustic signal from measurements on the bottom of the pipe. A critical standard deviation value separates the suspended solids flow regime from the settled solids flow regime. The suspended solids flow regime is above the critical value, while the flow over settled solids flow regime is below the critical value.

Power Spectral Density

Figure 7 is a power spectral density plot of a sand slurry showing the flow over settled solids and suspended solids flow regimes from acoustic signals measured on the bottom of the pipe. As shown in this figure, many similar frequencies exist, however, the magnitude of specific peaks or shifts in frequencies cannot be used to reliably determine the flow regime. Additional methods are required for flow regime detection.

Figure 8 shows the total power values over the frequency range of 1-20 000 Hz obtained from the power spectral density for the flow over settled solids and suspended solids flow regimes, along with the flow of water in the pipe and the pump without solids or water in the system, measured along the bottom of the pipe. There is a large separation between the power values for the water flow only and pump without water and solids in the system, and the values for the flow over settled solids and suspended solids flows. The flow of water has the greatest power values, since the water gains momentum from the diaphragm pump as it travels though the pipeline. The pump alone has a lower power value, since there is no water or slurry in the pipe to pump through. The addition of solids to the system decreases the power values; suspended solids impede the movement of the water, and decrease the power values of the water and slurry. The values are further decreased for the flow over settled solids flow regime, since the particles settle to form a bed further damping the slurry flow. The dominant frequencies vary for each case; the dominant frequency for the water flow is approximately 1600 Hz, for the diaphragm pump is 3200 Hz, and for the suspended solids flow and flow over settled solids slurries is approximately 2000 Hz.




Figure 9 demonstrates that the flow regimes can be successfully determined using power calculated from the power spectrum from acoustic signals recorded on the bottom of the pipe. A critical power value separates the suspended solids flow regime from the settled solids flow regime. The suspended solids flow regime is above the critical value, while the flow over settled solids flow regime is below the critical value.

V-Statistic and Cycle Time Regularity

The V-Statistic and cycle time regularity were calculated for measurements recorded at all distances from the elbow, solids concentrations, and slurry velocities.

A maximum V-Statistic was not always present on the V-Statistic curve. However, the V-Statistic can successfully distinguish between the flow over settled solids and suspended solids flow regimes for sand slurries at a subperiod length of 0.000425 s, as shown in Figure 10. This shows the values of the V-Statistic at this subperiod length for all measurement locations along the bottom of the horizontal line, solids concentrations, and slurry velocities.

Cycle time regularity of a signal was then examined as a method to determine flow regimes. Figure 11 shows the cycle time regularity values at each measurement location along the bottom of the horizontal pipe for flow over settled solids and suspended solids flow regimes, as well as for the flow of water only in the system, and the pump only without water or solids in the system. When solid deposits exist on the bottom of the pipe, the signal becomes less regular (low cycle time regularity value) due to a large range of cycle times, caused by deposited solids, and the movement of suspended particles. The suspended slurry has a more regular cycle time (larger cycle time regularity value), since a settled bed does not exist in the pipe. The cycle time regularity values are greatest for the pump and water only flows showing that solids have the greatest effect on cycle times, through the change in the flow regime.



The regularity of the cycle time, therefore, can be used to determine flow regimes. Figure 12 shows the values of the cycle time regularity for sand slurries at all solids concentrations and slurry velocities measured along the bottom of the pipe. This figure demonstrates that the cycle time regularity of acoustic signals can reliably distinguish between the two flow regimes identified, flow over settled solids and suspended solids flow, at all distances from the elbow and all flow conditions. A critical value exists that separates the two flow regimes; above this critical value is the suspended solids flow regime, and below the critical value is the flow over settled solids flow regime.


Wavelet Spectra

The wavelet spectrum was used to determine an ideal octave for identifying flow regimes from the acoustic signals. The wavelet spectra of the acoustic signals consisted of 19 octaves. Figure 13 shows the normalized standard deviation of the coefficients plotted for each octave for the flow over settled solids and the suspended solids flow regimes, the flow of water only through the pipe, and the pump without water and solids in the system, measured on the bottom of the pipe. These values were obtained by dividing the standard deviation of the coefficients by the standard deviation of the original acoustic signal. The 17th octave (corresponding to 5000 Hz) was chosen as the optimum octave since it contained the largest difference in values between the two slurry flow regimes. As well, it is shown that the water and pump values are greater than the slurry values and therefore do not effect the slurry values at this octave.

Figure 14 shows the values of the normalized standard deviation of coefficients in the 17th octave for all measurement locations along the bottom of the horizontal pipeline. The flow of water only through the pipe has the largest normalized standard deviation of coefficients values, followed by the pump without water and solids in the system and then suspended solids flow and flow over settled solids. Water amplifies the values obtained from the pump, whereas when solids are added, the solids damp these values to below the pump values with further damping when deposits have formed on the bottom of the pipe.

Figure 15 shows that the normalized standard deviation of the coefficients can be used at all distances along bottom of the horizontal line to determine the flow regime. The data in Figure 15 is the same data shown in Figures 9 and 12. These plots contain data from all solids concentrations and slurry velocities tested. There is a critical normalized standard deviation of coefficients value where the flow regime changes: above this value is suspended solids flow, and below this value is the flow over settled solids flow regime.





Kurtosis was calculated from the wavelet spectra using the wavelet coefficients at each octave. Of the 19 octaves, the 17th octave provided the clearest distinction between each flow regime.

Figure 16 shows the kurtosis values for the flow over settled solids and suspended solids flow regimes, the flow of water and the pump only at each octave number, measured on the bottom of the pipe. At high octaves the water flow and pump do not affect the flow regimes since these values are significantly lower than when solids are present in the system.



Figure 17 shows the kurtosis values along the bottom of the horizontal pipe for the flow over settled solids and suspended solids flow regimes, for water only through the pipe and for the pump without water or solids in the system. The flow over settled solids has the largest kurtosis values, followed by suspended solids flow, the pump only values and the water only values. Again, these values are significantly lower and do not affect the analysis method.

Figure 18 shows that there is a critical kurtosis value that separates the flow regimes. This critical value is constant for all slurry velocities and concentrations at measurement locations along the bottom of the horizontal pipe. The data on which Figure 18 is based is the same data presented in Figures 9, 12, and 15. Values above the critical kurtosis value indicate the flow over settled solids flow regime, whereas kurtosis values below the critical value indicate the suspended solids flow regime.

Application to Other Slurries

In order to validate the reliability of the flow regime detection methods, each method was applied to slurries of multi-sized particles. The solids composition of these slurries were 100% small stones or a 50% small stones-50% sand mixture. Each method, with exception of the V-Statistic, was able to reliably determine the flow regimes for each slurry composition, on the top and bottom of the pipe. It was determined that the critical subperiod length cannot be used to detect the flow regimes for stone and sand-stone mixture slurries, since there is not a distinct separation between the V-Statistic curves for each flow regime at this time interval. The critical values for each detection method and slurry composition are shown in Table 1.


Discussion of Results

Many different advanced signal analysis techniques have been identified which allow for the detection of flow regimes in horizontal hydrotransport pipes. These methods allow for the identification of specific frequencies where the information regarding the flow behaviour is contained in the signal. Although all signal analysis techniques were applied to 30 s of acoustic signal, all the techniques presented can determine the flow regime using 1 s of data.

For sand slurries, the V-Statistic identifies 2000 Hz, corresponding to a cycle time of 0.000425 s, as critical for determining the flow regime. The octave number obtained from the wavelet spectrum indicates frequencies where flow regime behaviour is contained. In this case, all flow regimes can be identified at the 17th octave, corresponding to 5000 Hz, using the normalized standard deviation of coefficients and kurtosis. Similarly, the power spectral density plot identifies dominant frequencies between 1000 and 5000 Hz.

The cycle time regularity indicates that a range of cycle times exists in the signal, which was verified by the power spectra. The power spectral density shows that practically all the information contained in the signal is below 10 000 Hz.

Standard deviation is the deviation of the signal from the signal mean. The power determined from the power spectral density is the sum of the magnitude of all frequencies contained in the signal. The normalized standard deviation of wavelet coefficients is the deviation from the coefficient mean at a specific frequency. The standard deviation, power and normalized standard deviation of coefficient values at the 17th octave are higher for suspended solid flow compared to flow over settled solids. These techniques are sensitive to collisions and the sound of moving particles in the liquid which result in high standard deviation, power and normalized standard deviation of coefficient values. Fewer collisions with the pipe wall result when there is a settled bed of particles, and the settled bed of particles muffles sound.

Cycle time regularity indicates that a range of cycle times are present in the signal. The low cycle time regularity value for flow over settled solids indicates a wide range of cycle times. The wide range of cycle times is due to the varying particle trajectory lengths; this technique is sensitive to the movement of particles that are suspended in the flow and the movement of particles settled in the bed.

Kurtosis of the wavelet spectrum coefficients indicate the relative peakedness of the distribution of the coefficients at the 17th octave. For flow over settled solids, the kurtosis values are high, whereas the suspended solids flow kurtosis values are low. Kurtosis of the coefficients of the 17th octave is sensitive to the movement of particles. High kurtosis values result from collisions of suspended particles with the pipe wall and movement of the bed of settled particles, whereas the low kurtosis values result from collisions of the suspended particles with the pipe wall.

All methods indicate that the formation of a layer of settled solids at the bottom of the pipe reduces the amplitude of the sound emissions, which results in reduced standard deviation, power, and V-Statistic. The formation of the settled solids layer also increases the spread in the cycle times, reducing both the V-Statistic and cycle time regularity. Finally, the formation of the settled solids layer increases the regularity of the signal at the frequency of the 17th octave, since the normalized standard deviation of the corresponding wavelet coefficients is reduced and their kurtosis is increased.

Each of the techniques used to identify flow regimes were examined based on the Reynolds number. The plots for the V-Statistic and kurtosis are shown in Figures 19 and 20 with the flow regime transition value shown.

The V-Statistic and kurtosis can clearly detect changes in the flow regime. Figures 19 and 20 show an abrupt change in the V-Statistic and kurtosis values with an increase in the Reynolds number, which was calculated from the effective density and viscosity of the slurry. V-Statistic values increase significantly and become constant with an increase in the Reynolds number as the flow regime transition value is crossed. The kurtosis values decrease and become relatively constant with an increase in the Reynolds number as the flow regime transition value is crossed. The other methods identified show a gradual increase in their values as the Reynolds number is increased. This indicates that these methods are better at detecting other changes in the hydrodynamic properties of the system than at identifying flow regime transitions.



The differences in the critical values using each flow regime detection method for each slurry material is due to the size of the solids in the slurry and its influence on the slurry behaviour. Each slurry material has a different behaviour as settled particles move along the bottom of the bed. The settled bed formed by the sand slurry slides along the bottom of the pipe whereas stones in the deposited bed roll along the bottom of the pipe. The slurry of sand and stone exhibits a different flow behaviour than the other two slurries: the stones try to roll along the bottom of the pipe, however, the rolling is inhibited by the sand, and results in sliding as the dominant motion of the bed. Each of these bed behaviours affects the critical flow regime transition value.

The slurry viscosity depends on the solids within the system. The slurry of sand at a given solids concentration has a higher viscosity than the slurries of sand and stone and stones. Since the stones are significantly larger than the sand, they do not contribute to the viscosity of the slurry. Due to the large particle size of the small stones, the stones hit the pipe with more force and their trajectories will be shorter than for sand particles, resulting in more collisions with the wall. Sand particles travel further between collisions with the wall and do not collide with the same frequency as the stones. The sand-stone mixture slurry is more viscous than the slurry of stones, but less viscous than the sand slurry. It contains the stones that roll along the bottom of the pipe, however, since the rolling motion is impeded by the sand causing the overall deposited bed to slide.


Flow regimes in a horizontal hydrotransport pipe can be determined from acoustic probe measurements and frequency and cycle analysis techniques. Slurries with a uniform particle size are unrealistic in the oil sand industry, resulting in the need for techniques to account for a wide range of particle sizes. The signal analysis techniques used in this study were cycle time regularity, power derived from the power spectrum, normalized standard deviation of coefficients of the 17th octave, and kurtosis of the 17th octave. These detection methods are applicable to all the acoustic measurement locations, slurry velocities, slurry concentrations, and slurry materials studied in the horizontal line.

All analysis techniques were developed using the raw acoustic signals, and thus, did not require any filtering of the signal. This allows for rapid, on-line analysis of the signal for early detection of solid settling in the pipeline. Through monitoring and the use of a detection method, crossing of the critical transition region value indicates a change in the flow regime. Use of any of these methods allows for early detection of deposited solids and allows for rapid corrective action to ensure product quality, extended life of the pipeline and system efficiency.

[C.sub.m] sand concentration of the slurry (mass fraction)

K kurtosis

M number of points in the signal

Q intermittency ([V.sup.-2])

[q.sub.2] second moment of wavelet coefficients at an
 octave (V)

[q.sub.4] fourth moment of wavelet coefficients at an
 octave (V)

[(R/S).sub.[tau]] average rescaled range for a subperiod of length

[R.sub.T] regularity of cycle time

[DELTA]t time interval between data points (s)

T cycle time (s)

[U.sub.slurry,avg] average slurry velocity (m/s)

[U.sub.water] water velocity measured by flowmeter (m)

[V.sub.[tau]] value of V-Statistic for a subperiod of length
 [tau] ([s.sup.-0.5])

[V.sub.T] V-Statistic at cycle time ([s.sup.-0.5])

x distance of the probe from the elbow (m)

y signal value (V)

[bar.y] mean signal value (V)

Greek Symbols

[sigma] signal standard deviation (V)

[tau] subperiod length (s)


The authors would like to thank the Natural Sciences and Engineering Research Council of Canada and Syncrude Canada Ltd. for their financial support of this research. The authors would also like to thank the Chemical Engineering lab technicians at the University of Western Ontario, Mike Gaylard and Souheil Afara.

Manuscript received February 8, 2008; revised manuscript received August 12, 2008; accepted for publication August 28, 2008.


Albion, K., L. Briens, C. Briens and F. Berruti, "Flow Regime Detection in Horizontal Pneumatic Transport of Fine Powders using Non-Intrusive Acoustic Probes," Powder Technol. 172, 157-166 (2007x).

Albion, K., L. Briens, C. Briens, F. Berruti and G. Book, "Flow Regime Determination in Upward Inclined Pneumatic Transport of Particulates using Non-Intrusive Acoustic Probes," Chem. Eng. Process 46, 520-531 (2007b).

Boyd, J. W R. and J. Varley, "The Uses of Passive Measurement of Acoustic Emissions from Chemical Engineering Processes," Chem. Eng. Sci. 56, 1749-1767 (2001).

Breitenbach, J. and D. Weber, "On Acoustically Determining Ore Types and Particle Size in Tumbling Ball Mills," In: 5th Acricon Conference in Africa: Electrotechnological Services for Africa, IEEE, Capetown, South African September 28-October 1, 1999, IEEE pp. 331-334 (1999).

Briens, L. A. and C. L. Briens, "Cycle Detection and Characterization in Chemical Engineering," AIChE J. 48, 970-980 (2002).

Briens, C. L., L. A. Briens, E. Barthel, J. M. LeBlevec, A. Tedoldi and A. Margaritis, "Detection of Local Fluidization Characteristics using the V Statistic," Powder Technol. 102, 95-103 (1999).

Charles, M. E., "Transport of Solids by Pipeline", in "Proceedings of Hydrotransport 1: First International Conference on the Hydraulic Transport of Solids in Pipes," BHRA Association, Coventry, England, September 1-4, 1070, British Hydromechanics Research Association, Cranfield, A3, pp. 25-36. (1971).

Doron, P. and D. Barnea, "A Three-Layer Model for Solid-Liquid Flow in Horizontal Pipes," Int. J. Multiphase Flow 19, 1029-1043 (1993).

Doron, P. and D. Barnea, "Flow Pattern Maps for Solid-Liquid Flow in Pipes," Int. J. Multiphase Flow 22, 273-283 (1996).

Doron, P., M. Simkhis and D. Barnea, "Flow of Solid-Liquid Mixtures in Inclined Pipes," Int. J. Multiphas Flow 23, 313-323 (1997).

Echim, M. M., H. Lamy and T. Chang, "Multi-Point Observations of Intermittency in the Cusp Regions," Nonlin. Processes Geophys. 14, 525-534 (2007).

Fair, A., J. Oxenford, J. Coward and M. Lipsett, "Technology Visions for Mining at Syncrude," CIM Bull. 92, 113-121 (1999).

Farge, M. and K. Schneider, "Wavelets: Application to Turbulence," In: J. P., Francoise, G. Naber and T. S. Tsun, Eds., "Encyclopaedia of Mathematical Physics," Elsevier, Boston, pp. 408-420 (2006).

Flott, L. W., "Quality Control: Non-Normal Frequency Distributions," Met. Finish 93, 52-55 (1995).

Gilies, R. G., C. A. Shook and J. Xu, "Modelling Heterogeneous Slurry Flows at High Velocities," Can. J. Chem. Eng. 82, 1060-1065 (2004).

Gillies, R. G., K. B. Hill, M. J. McKibben and C. A. Shook, "Solids Transport by Laminar Newtonian Flows," Powder Technol. 104, 269-277 (1999).

Greenspan, H. P. and M. S. Nigam, "A Note on Separation of a Bimodal Mixture in Pipe Flow," Int. J. Multiphase Flow 27, 2015-2021 (2001).

Hou, R., A. Hunt and R. A. Williams, "Acoustic Monitoring of Pipeline Flows: Particulate Slurries," Powder Technol. 106, 30-36 (1999).

Hurst, H. E., "Long-Term Storage Capacity of Reservoirs," Trans. Am. Soc. Civil Eng. 116, 770-808 (1951).

Kaushal, D. R. and Y Tomita, "Solids Concentration Profiles and Pressure Drop in Pipeline Flow of Multisized Particulate Slurries," Int. J. Multiphase Flow 28, 1697-1717 (2002).

Kaushal, D. R., K. Sato, T. Toyota, K. Funatsu and Y Tomita, "Effect of Particle Size Distribution on Pressure Drop and Concentration Profile in Pipeline Flow of Highly Concentrated Slurry," Int. J. Multiphase Flow 31, 809-823 (2005).

Lin, S. H., "Pressure Drop for Slurry Transport," Chem. Eng. 89, 115-117 (1982).

Maruyama, T., K. Kojima and T. Mizushina, "The Flow Structure of Slurries in Horizontal Pipes," J. Chem. Eng. Jpn. 12, 177-182 (1979).

Matousek, V., "Pressure Drops and Flow Patterns in Sand-Mixture Pipes," Exp. Theom. Fluid Sci. 26, 693-702 (2002).

Newitt, D. M., J. F. Richardson, M. Abbott and R. B. Turtle, "Hydraulic Conveying of Solids in Horizontal Pipes," Trans. Inst. Chem. Eng. 33, 93-113 (1955).

Peters, E. E., "Fractal Market Analysis: Applying Chaos Theory to Investment and Economics," John Wiley & Sons, Inc, Toronto (1994).

Shook, C. A., R. G. Gillies and R. S. Sanders, "Pipeline Hydrotransport with Applications in the Oil Sand Industry," Saskatchewan Research Council: Saskatoon, SRC Publication No. 11508-1E02." (2002).

Skudarnov, P. V, C. X. Lin and M. A. Ebadian, "Double-Species Slurry Flow in a Horizontal Pipeline," J. Fluids Eng. 126, 125-132 (2004).

Stienstra, G. J., J. Nijenhuis, T. Kroezen, C. M. van den Bleek, and J. R. van Ommen, "Monitoring of Slurry-Loop Reactors for Early Detection of Hydrodynamic Instabilities," Chem. Eng. Process 44, 959-968 (2005).

Takahashi, H., T. Masuyama and K. Noda, "Unstable Flow of a Solid-Liquid Mixture in a Horizontal Pipe," Int. J. Multiphase Flow 15, 831-841 (1989).

Trygg, J., N. Kettaneh-Wold and L. Wallbacks, "2D Wavelet Analysis and Compression of On-Line Industrial Process Data," J. Chemometr. 15, 299-319 (2001).

Valenti, M., "Oil Sand Kicks into High Gear", Mechanical Engineering Magazine," backissues/december98/features/oilsand/oilsand.html (1998).

Valentino, M., "Microphone Handbook", PCB Piezotronics Vibration Division," Vibration/Microphone_Handbook.pdf (2005).

Katherine Albion, Lauren Briens, * Cedric Briens and Franco Berruti Institute for Chemicals and Fuels from Alternative Resources, The University of Western Ontario, London, Ontario, Canada N6A 589

*Author to whom correspondence may be addressed. E-mail address:
Table 1. Critical flow regime transition values for each detection
method and slurry composition for measurements recorded on the top
and bottom of the pipe


 Top Bottom

Standard deviation (V) 0.0125 0.0125

Power from power spectral 4.45 4.45
density ([V.sup.2])

Cycle time regularity 0.20 0.20

Normalized standard 0.62 0.70
deviation of coefficients

Kurtosis 11.91 11.64

 Sand-small stone mixture

 Top Bottom

Standard deviation (V) 0.0130 0.0130

Power from power spectral 4.72 4.58
density ([V.sup.2])

Cycle time regularity 0.24 0.25

Normalized standard 0.80 0.82
deviation of coefficients

Kurtosis 8.05 8.05

 Small stone

 Top Bottom

Standard deviation (V) 0.0175 0.0195

Power from power spectral 5.69 6.38
density ([V.sup.2])

Cycle time regularity 0.27 0.30

Normalized standard 0.65 0.64
deviation of coefficients

Kurtosis 4.67 4.59
COPYRIGHT 2008 Chemical Institute of Canada
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2008 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Albion, Katherine; Briens, Lauren; Briens, Cedric; Berruti, Franco
Publication:Canadian Journal of Chemical Engineering
Date:Dec 1, 2008
Previous Article:Henry's law constant for hydrogen in high temperature water with dissolved lithium and boron.
Next Article:A particle filter approach to identification of nonlinear processes under missing observations.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters