01 September 2004
Hear this: Acoustic sampling
Online viscosity sensor meets petrochemical quality standards.
By Jeffrey Andle and Douglas Korslund
Today's petroleum refining operations are striving for ever-increased productivity and higher quality standards.
One of the critical parameters that managers must measure to achieve better production results is the viscosity and viscosity index physical property. The importance of this measurement is to provide operations with the control feedback data to adjust the refining and blending process of petroleum products.
Considering the high throughput of product in such processes, the laboratory approach is not the most desired solution, due to the lag time between sampling and result reporting. Therefore, the continuous measurement approach is the most ideal, yielding instantaneous result reporting and quality control feedback.
The historical design of an online viscosity or viscosity index analyzer has resulted in a complex, highly mechanical device associated with a high cost/price. This is partially due to the design philosophy of replicating the laboratory measurement online. In the past, there were attempts to correlate the viscosity or viscosity index measurement against an alternate, surrogate measurement technique. However, those correlations have yielded limited success, since measurement linearity and measurement interferences are not commonly mutual.
A new approach and investigation is addressing a number of these concerns. Using highly sensitive acoustic viscosity sensors in an online configuration will provide the benefits of fast result reporting, simple fluidics management, and advantageous cost/benefit relation. The initial results clearly indicate the promise of an advanced online viscosity or viscosity index analyzer that can measure accurately in low ranges below 10 centistokes (cSt or mm2/second) as well as elevated ranges of up to 10,000 cSt. This sensor coupled with a solid process analyzer platform can install and exist in a harsh environment.
A proposed sensor measures viscous damping of a surface shear wave on a quartz crystal to obtain the viscosity–density product of the sample. The sensor offers solid-state reliability, rapid response time, and microscopic measurements directly online. The drawbacks are that the measurements take place at very high shear rates (approximately 2–20 million seconds-1 or 1–100 cSt oils and 0.2–2.0 million seconds-1 for 100–10,000 cSt oils). The sensor also measures the viscosity–density product rather than its ratio, and samples a very microscopic region (typically less than a micron thick).
The high shear rates lead to non-Newtonian behaviors even in mineral oils. The sensor calibration factors out the attendant shear thinning of the mineral oil standards, but at the expense of correlation to low shear methods for samples having a different degree of shear thinning (water, solvents, resins, plastics, etc.). The measurement of acoustic viscosity (ηρ) instead of kinematic viscosity (η/ρ) is conceptually foreign to many end users.
Finally, sample homogeneity is critical to a meaningful measurement. Nonetheless, the proposed system offers a unique set of advantages that will outweigh the new challenges that it imposes.
Slippage of the pumping fluid
When considering all applications of viscosity measurement, the most pervasive measurement method is the viscosity cup. A known volume of liquid flows through a known orifice in a measured period of time. The design of the cup and its orifice yield a cup factor, and measuring the viscosity takes only seconds.
A more refined method employs a capillary tube of known and controlled geometry.
One method measures flow rate or flow time under a known pressure while the other method measures pressure drop across a known capillary orifice under known flow rates. In either case, the unit of measure is the ratio of viscosity to density, η/ρ – the kinematic viscosity (mm2/s or cSt).
The latter method works in an online process-monitoring device; however, the instrumentation needed to control flow rates and measure pressures can be complex.
Typically the orifice geometry is small compared to the process piping, and the measurement must take place in a slipstream. Maintaining constant flow rate in liquids independent of pressure is difficult and requires positive displacement pumps.
With the introduction of positive displacement pumps, the construction of the pump as well as potential slippage of the pumping fluid can alter the desired flow rate.
The analytical method of choice outside the petrochemical industry has been the rotational viscometer, which measures the viscous drag on a rotating disk or cylinder having a known concentric spacing to a second stationary cylinder.
The relationship between torque and speed is interpreted as intrinsic viscosity (mPa*s or cP). While accurate, this laboratory method is highly susceptible to movement of the measurement platform and flow of the liquid other than that caused by the measurement device.
Another method of measurement is to time the fall of a piston, needle, or bearing in a sample. The method again measures kinematic viscosity (cSt). These methods are prone to clogging by soot and other solids. Sample flow can result in hydrodynamic lift of the falling object, altering the result.
Typical systems are National Institute of Standards and Technology (NIST)-traceable to the known viscosity of water at a reference temperature. Mineral or silicone oil standards mix together to provide known relative values to the NIST-traceable water measurement.
Existing physical effects
Non-acoustic methods offer the potential for on-process analysis. These measurement methods suffer from the employed technique. The technique employed alters the sample characteristics. The cost of implementing the non-acoustic methods are high due to sample preparation and sample handling and can have a compromising effect on the results.
Capillary tube viscometer: Capillary systems use a restriction to the flow of the sample and place challenging limitations on pressure and flow control. The requirement of measuring the liquid at a known shear rate places flow rate constraints on the system.
Rotary viscometers: The rotational viscometer is highly susceptible to the flow of the sample, and therefore is ideally suited for the static measurements. It also induces flow and mixing in the sample, and may cause incorporation of air and can accelerate the reaction between components of a mixture.
Falling piston: The automated versions of this method experience sample heating due to the mechanical and electrical systems that raise and lower the object. Typical systems lift a magnetic piston by using an electromagnet, and then detect the piston position as it falls through a variable reluctance coil. The heat generated by the lifting coil is substantial and is the source of the sample heating.
Properties of acoustic
A promising family of solid-state sensors is out now. The sensors platform on acoustic wave energy propagation, reflection, and dissipation. These methods generally overcome several limitations of the prior methods while introducing new and unique measurement opportunities. These solutions are, however, not without their own unique challenges and limitations.
The most promising methods measure the product of viscosity and density, termed acoustic viscosity (AV) or acoustic impedance. These readings typically come from the measurement of power reflectance from a solid-liquid interface or energy loss across a solid-liquid interface. Let's look at these while paying particular attention to the proposed method using acoustic waveguides in contact with the sample.
The physical method of measurement relies on the transfer of acoustic shear wave energy from a solid waveguide (a plate of quartz, for example) having characteristic material impedance . . .
ZW = (ρwµ)1/2
. . . into an adjacent liquid having characteristic material impedance . . .
ZL = (ωρLη)1/2
In these expressions,
ρW is the density of the waveguide material,
ρL is the density of the liquid,
µ is the shear elastic modulus of the solid waveguide,
ω is the radian frequency of the wave, and
η is the intrinsic viscosity.
Energy transfer is proportional to the ratio, ZL/ZW provided ZL << ZW. The square of the power loss is proportional to the product of frequency, density, and viscosity, ωρη. Since one knows the frequency, the viscosity-density product is measured. The acoustic waveguide is relatively immune to changing sample conditions. Regardless of the rate of flow of the bulk liquid, the sample appears stationary to the ultrasonic vibrations in the quartz crystal, even turbulent flows.
Acoustic waveguide sensors that measure changes in resonant frequency are susceptible to pressure-induced frequency changes; however, there are no reported mechanisms for pressure-induced power losses.
The available power to be dissipated into the liquid is 0.05 W. Assuming all of this power were dissipated into a high viscosity material in a (5mm)2 area with a 1µm penetration depth, the heat density would be 2W/(mm)3. While the total power dissipation is quite small, the power still can dissipate in a very small volume. Proper results require good circulation of the sample.
Unique challenges and limits
The candidate sensor is only responsive to the material properties of a microscopic region of the sample adjacent to the solid-liquid interface. Homogeneity and molecular weight/particle size are perhaps the two most significant issues, followed by the high, viscosity-dependent shear rate of the method. Since the penetration depth, d, of the acoustic wave is proportional to the square root of the viscosity and is 0.05 µm for water in the candidate sensor (5 µm for 10,000 cSt samples), the measurement method is unsuitable for extremely high molecular weight samples in which there is granularity of the material on the order of tenths of microns.
Furthermore, since the penetration depth is extremely small in low viscosity solvents, even moderate sized molecules (0.5 µm) begin to appear as distant, immovable objects that are beyond the reach of the acoustic wave in the lighter solvent.
Even for those materials meeting the homogeneity requirements, the local forces associated with the acoustic wave and the associated shear rates are sufficiently high that few, if any, liquids are truly Newtonian. A calibration procedure accounts for the varying degree of non-Newtonian behavior of mineral oils. A customer calibrated a batch of 20 pilot production sensors from –30°C to +145°C using a series of six standards. Evaluation of the calibrated sensors took place in N350 oil standard (a standard having 350 cSt viscosity at 40°C). The oil exhibits clouding below 20°C.
Exceeds the present accuracy
Historically, the online application of viscosity and viscosity index has been limited to the capillary approach, measuring cSt. The actual online, process monitoring implementation thereof requires numerous mechanical and electrical devices that can be complex to operate and of substantial physical size and weight.
With the advancements of measurement techniques such as the acoustic viscosity and the availability of refined mechanical and electrical components, the opportunity of designing and manufacturing an online analyzer that meets and exceeds the present accuracy and reliability requirements associated with a highly improved cost/benefit scenario is now feasible.
Applying a viscosity-sensing device inline poses a number of disadvantages, even if it seems to be a more cost-effective approach at first. Inserting and extracting such a device into a highly violent process stream goes along with a number of installation and safety issues. High-pressure valves with all the necessary safety interlocks have to install. Such configurations limit the possibility of quickly calibrating, validating, or cleaning the sensing device. The task on hand starts at the sampling point where the integrity and homogeneity of the sample must maintain throughout the entire analysis cycle. It's best to employ a bypass slipstream from which a small amount of sample diverts into the analyzer for analysis. Vital parameters like flow rate, pressure, and temperature must remain as constant as possible. The automated sample extraction for the analyzer will mimic the manual sample extraction for the laboratory analysis, whereas it is best to use the same sample port for either the process analysis or the laboratory analysis. This minimizes discrepancies between the two and yields higher and better correlation factors.
First, the sample enters the online analyzer where its vital parameters remain as constant as possible. Employing a highly durable micropiston pump allows for easy management of flow rate and pressure.
The addition of a highly precise and tightly controlled heat exchanger provides the thermal sample management required to provide consistent and reproducible results.
With the availability of a calibration/validation port, a questionable result can easily and quickly compare to and check against a known solution providing rapid diagnostics information to operations or maintenance.
Newer analyzers on the market have advanced decision logics already programmed, whereas the analyzer will automatically start a validation cycle when a measurement result falls outside of a preset percent deviation of its prior measurement. Only then will the analyzer report the result or the analyzer fault state. This type of technology works well for the viscosity index analysis for which the sample management poses additional challenges. Keeping the vital parameters under control is paramount.
This works by introducing strategic valve manifolds that minimize the fluidics passages as well as allow for simple and accurate thermo control of the sample measured. The sample amount required for the acoustic viscosity sensor is substantially smaller than what is required for conventional systems. This allows for the use of components that are smaller in size, weight, and complexity, as well as for reduced sizes in tubing diameter and tubing length.
Behind the byline
Jeffrey Andle (firstname.lastname@example.org) has a masters and a Ph.D. in electrical engineering. He has received several National Science Foundation awards and the Department of Energy recognized him for his work with acoustic sensors. He works for BIODE, Inc., focusing on liquid phase chemical and biochemical sensors. Doug Korslund (email@example.com) is an ISA member. He has three engineering degrees and is a PE in four states. He is the founder, president, and general manager of Orb Enterprises, Inc., an engineering consulting and contract manufacturing company.
Mineral oils, both the calibration standards and the samples evaluated in this test, are non-Newtonian at the shear rates of the acoustic viscometer (1–20 million seconds-1).
The spread in the data is in part due to variations (+/– 10% nominal) in the shear rate of the sensors and in part due to the discontinuities in the calibration set and the use of some calibration standards near their cloud points. 85W140 gear oil, SAE50 motor oil, 75W90 gear oil, ATF+3 transmission fluid, Type F transmission fluid, and brake fluid undergo measurement over temperature.
Each sample—about 200 ml in a beaker—covered the three sensors. Data is in AV units. To obtain cSt values, it is necessary to divide by the square of the density.
Some liquids exhibit excellent sensor-to-sensor reproducibility (e.g., all but 85W140). A lack of suitable oil standards in the upper right quadrant (high viscosity at high temperature) may be responsible for this.
It is also notable, however, that the acoustic viscometers exhibit some part-to-part variation in nominal shear rate and that high viscosity oils have non-Newtonian behavior.
Finally, as the most viscous sample and whereas the temperature gradients were largest in this region, it is also possible that mixing and thermal equilibrium were limited.
Two passes for certainty
Using a tandem heat exchanger system optimizes the sample handling where the very same physical sample is undergoing analysis using the very same acoustic sensor. This approach eliminates the potential errors introduced when splitting the sample and/or running the sample(s) by two sensors. The sample flows through all the normally open passages with the result loading into the second heat exchanger for subsequent measurement.
Then valve activation occurs and the sample previously measured at the lower temperature, now present in the heat exchanger, passes the sensor again, this time at its elevated or differential temperature. This cycle runs continuously without interruption, avoiding sample alteration, precipitation, or any other undesired effects.
Vacuum gas oil
Vacuum gas oil samples warmed to 75°C before pouring into beakers containing four sensors each.
The samples then went to 110°C, cooled to 35°C, heated to 90°C, and finally cooled to –10°C.
Data records over the specified calibration range of the sensors.
Sample 2008 (black) solidified at approximately 55°C and below 40°C the data was erratic as the solid mass separated from the sensors. Sample 3052 (red) also solidified but did not separate from the sensors and continued to read like a viscous fluid except with respect to one sensor below 10°C.
It is likely that some lighter fractions remained liquid and in contact with the sensor or that the acoustic vibration energy was sufficient to keep a small volume of material in the liquid state.
Additional data registered for a heating ramp from slightly below 0°C to 120°C. The process entailed agitation of samples at various temperatures in order to dissolve the waxy particles into the liquid phase.
Both data sets indicate that the sensor-to-sensor spread was larger than the difference between samples 2007 and 2051, whereas sample 2008 consistently read slightly lower than sample 2052 above the cloud point.
Return to Previous Page