Soil Science Society of America Journal 67:107-111 (2003)
© 2003 Soil Science Society of America
DIVISION S-1—NOTES
Rapid water flow instrumentation
Wallace Troyera and
J. Skopp*,b
a Department of Agronomy and Horticulture, University of Nebraska, Lincoln, NE 68583-0915
b School of Natural Resource Sciences, University of Nebraska, Lincoln, NE 68583-0915
* Corresponding author (jskopp1{at}unl.edu)
 |
ABSTRACT
|
|---|
Measuring high flow rates for small quantities of water is useful when determining soil hydraulic properties. A system that measures such flow rates precisely and accurately is not currently on the market. We present a flow measuring system, which uses pressure transducers and a data logger. This flow system has three advantages over a scale: (i) an automated interface allows precise initiation and measurement of system response, (ii) the ability to measure intervals as short as a tenth of a second, and (iii) quick response. Tests show that the system is highly repeatable, accurate, and responsive. A direct comparison of the flow system and scale output collected simultaneously shows that the flow system was more responsive than the scale in the early stages of the test, though the total output of both was similar. The instrumentation is adaptable for both laboratory and field methods. The flow rate can be adjusted simply by changing the diameter of the cylinder used. The flow system is built from off-the-shelf parts, except the control circuitry, and can be built for less than $1000.
|
INTRODUCTION
|
|---|
THE ABILITY to measure high flow rates (per area of soil) for small quantities of water is a limiting factor for many soil physical observations. This is true both in the field and in the laboratory. For example, measurement of rapid water flow through high permeability saturated soil or unsteady flow in unsaturated soil is a key to determining soil hydraulic properties (Klute and Dirksen, 1986; Green et al., 1986). Traditional methods to observe flow suffer from a number of limitations relating to the magnitude and rates that can be observed as well as the smoothness of the resulting data. The objective of this research was to develop an accurate and precise means of measuring high flow rates of water (yet small total quantities) that was usable both in the field and laboratory. We illustrate the principle by referring to one-step outflow methods described below.
Electronic scales have been used to weigh water from one-step outflow experiments (Borcher et al., 1987; Mu'azu et al., 1990). The liquid is captured in a beaker sitting on the scale. Electronic scales have been the primary equipment used in soil for the following reasons. First, scales that can be interfaced with a computer are readily available on the market. This allows output to be collected and saved to disk with relative ease. Second, the response time of the scale is relatively fast (approximately a quarter of a second). And finally scales are relatively stable, precise, and accurate for traditional laboratory measurements. Scales are designed for static rather than dynamic measurements and a displayed mass is actually a time weighted average. Thus the accuracy and precision of scales requires a stable or constant mass placed on the pan. However, the measurement of rapidly changing masses is a particular need when trying to develop rapid methods for soil hydraulic properties.
Electronic scales cannot readily fill the requirements for measuring rapidly changing masses. First, many scales are restricted to time intervals greater than a quarter of a second and report a time weighted average with no user control on the time period used for weighting. Second, there is no interface allowing a simultaneous start of the water flow and the data acquisition in a manner that allows an unbiased determination of the experiment start time relative to the scale time. Mechanical initiation of data collection is not simultaneous, but is often seconds apart. And third, a scale with a precision of 0.01 g may be sensitive to air currents within the laboratory.
The criteria set forth to develop our measuring system follow. First, the system should be economical and simple to build. Second, it should have the ability to interface with a data logger or computer so that data collection and the recorded start time are simultaneous. Third, the response time of the system should be no greater than a tenth of a second with an accuracy of 0.01 g. And last, the system should be physically robust and capable of use in the field.
|
MATERIALS AND METHODS
|
|---|
Flow rate system
The flow measuring system is sketched in Fig. 1
. The system consisted of an analog to digital converter I/O board connected to a relay (R) and to pressure transducers (P1 and P2) mounted on a cylinder. Assembly included mounting the I/O board in a easily carried protective case, mounting the transducer(s) on a cylinder, wiring the I/O board, transducers, control valve, and control relay, and developing a 8-V DC power supply to power the transducers. Total cost of the items required to build this system was approximately $1000.
View larger version (41K):
[in this window]
[in a new window]
|
Fig. 1. Apparatus used to test the flow system (includes measuring system, water supply system, and water delivery system but excludes core) for repeatability. Inclusion of core allows the apparatus to be used for falling head measurements. The measuring system is limited to the pressure transducers and the cylinder to which they are attached. Pressure transducers P1 and P2, mounted on cylinder a distance Hp apart, are used to measure water head in the cylinder. Maximum cylinder fill height was regulated by Mariotte device at H1. Minimum cylinder fill was regulated to H3. Manual valve V2 and automatic valve V1 regulated water flow. Switch SW energizes control relay R. SW simultaneously opens valve V1 and sends a signal to the data logger.  H is related to total water flow from cylinder.
|
|
The main component of the system is a Tattletale Model 6 analog to digital board (hereafter referred to as a data logger), purchased with a development kit (Onset Computer Corp., Pocasset, MA). The data logger has eight analog channels (allowing -5 to 5 V DC input) with a 12-bit A-D converter. It has a 9-V DC built-in power regulator that allows for a 9- to 12-V DC power connection. Its default interval reading time is 1 ms although the minimum allowable reading time is 416.6 µs.
The pressure transducers used were PX164-010D5V (Omega Engineering Inc., Stamford, CT). They have a rated differential pressure range of 0 to 25.4 cm of water, with a corresponding output range of 1 to 6 V DC. Their response time is 1 ms. They require an 8 V DC constant excitation voltage, which was provided through a control circuit developed in the laboratory, using a MCT7800 (Motorola, Newark Electronics, Omaha, NE) series voltage regulator as its principle component.
Although, the transducer and data logger both have a range of 5 V DC, the transducer begins at 1 V DC and the data logger at 0 V DC. The 1-V offset was adjusted with a bridge that split the signal from the transducer and sent its output to two analog channels of the board where they were digitally added.
The control valve used to start water flow is an OMEGA SV-310 (Omega Electronics Inc., Stamford, CT) (0.635-cm diam. orifice). The control valve was energized through the closing of a switching relay. This switching relay, DPST NO, was energized by a closing a switch. The second contact of the relay was connected to I/O Line 1 of the datalogger to start data collection. By this means both water flow and data collection started simultaneously.
Both a 120-V AC system for laboratory use and a 12-V DC system for field use were developed. In the AC system the control valve and relay were also 120 V AC. They were housed in a control box with a switch outside the case containing the data logger. In the DC system, the control valve and relay were 5 V DC. The power supply for the DC system was a 12-V lithium rechargeable battery.
The measuring cylinder had two transducers mounted on a plexiglass tube with a 0.95-mm i.d. (P1 and P2 of Fig. 1). Two transducers extended the operating range without sacrificing precision. The transducers were connected to a port by a short hose. Ports were inserted into the plexiglass cylinder by forming a hole and gluing in a connector. The hole was formed by inserting a slightly undersized red-hot metal rod and finished using a drill bit. The transducers were mounted slightly above their port, and adjusted a distance of Hp apart, which was <25 cm to provide overlap of their ranges.
The measuring cylinder was calibrated in the following manner (compare Fig. 1 noting that the core was removed for calibration). A small increment of water was allowed into a beaker (approximately 0.15 g of water passed each time Valve V1 was opened). A transducer reading was taken and recorded along with the beaker mass. The collected data were analyzed using SAS Proc Reg (SAS Institute, 1994) to calculate linear regression coefficients for total water mass as a function of pressure transducer response. System responsiveness was tested using a water-filled syringe along with repeated injection and withdrawal of water directly into the "Falling Head Cylinder" of Fig. 1. Both the response time and the baseline drift were evaluated with this procedure.
A slightly different arrangement was used to directly compare scale and transducer system transient response. A cylinder with one transducer, 0 to 25.4 cm was mounted on a 2.54-cm polyvinyl chloride (PVC) pipe (Fig. 2)
. A port was inserted into the cylinder by drilling a hole and gluing. The cylinder was mounted on a block of wood to keep it in an upright position while sitting on a scale pan. Transducer wires were taped to the scale pan and then to a block of wood level with the pan to minimize bias because of wire tension. A tube was added so that water entered below the sensor directly to eliminate oscillations because of falling drops.
View larger version (44K):
[in this window]
[in a new window]
|
Fig. 2. Modified measuring cylinder placed on scale pan. Other labeled components are the same as in Fig. 1. Wires coming from transducer are taped to scale pan to eliminate bias because of changes in wire tension as water level changes. The tubing inside the measuring cylinder allows water to flow without drops forming.
|
|
The cylinder was calibrated for filling, while sitting on the scale pan, using a syringe to inject about 1.5 g of water at a time. The actual mass injected was determined from the scale reading. Scale and data logger values were recorded. The data was analyzed using the SAS Proc REG Module to calculate linear regression coefficients for total water mass as a function of transducer response.
Scale
A scale was used for comparison purposes. The scale was a Thomas Scientific T2000S (Thomas Scientific, Swedesboro, NJ). This scale registers to 0.01 g and has a stable response time of approximately a quarter of a second. The scale was interfaced, with a 386-computer through a serial port set at 9600-baud rate.
The DOS computer program used to access the serial port was Telix, version 3.21 (deltaComm Development, Cary, NC). A program written in Script Application Language for Telix (SALT) was inserted into Telix to automate data acquisition. Data collection started when a key was pressed. Data written to the computer included scale mass in 0.01 g and time of reading. The program automatically increased the time between readings based on the rate of change. Data collection was stopped either at a preset time or manually.
|
RESULTS
|
|---|
Evaluation of the Measuring Cylinder
All pressure transducers were tested for linearity. Subsequent readings were always within the linear range. Linear regression used the following equation:
where Mo is mass outflow into beaker and R is output from the data logger. Examples of system calibration follow. For P1, the predicted values for a and b were 23.38 ± 0.044 and -28.99 x 10-5 ± 0.087 x 10-5 g, respectively. The R-square for the model was 0.9998 and the standard error of prediction ranged from 0.018 to 0.034 g. For P2 the predicted values for a and b were 22.39 ± 0.058 and -29.96 x 10-5 ± 0.119 x 10-5 g, respectively. The R-square value for the model was 0.9997 and the standard error of prediction ranged from 0.021 to 0.041 g.
The apparatus was evaluated for stability, repeatability, hysteresis, responsiveness, and accuracy. Stability was evaluated by keeping a constant water level and observing drift over a period of 3 min. at intervals of 0.1 s. In one test, the cylinder was filled in the upper range while in another test, the cylinder was filled in the lower range of the transducer. The slopes of these two responses were calculated by linear regression as 0.0015 and 0.00016 g s-1 for the upper and lower pressures respectively. Both slopes were found to be significantly different from zero at the 99.99% confidence level. Readings <1 min in duration can be expected to show a total drift of <0.1 g.
The repeatability of the flow system was evaluated. The measuring cylinder was filled to H1 (Fig. 1). Switch SW was opened allowing water flow and data collection to start simultaneously. Once the water level in the cylinder dropped to H3, SW was closed. The procedure was repeated six times. The output from this test was used to calculate a root mean square deviation among the replicates. The largest observed deviation was 0.20 g per observation while the mean deviation was 0.10 g. This represented <2% error after 0.7 s and never more than 9% of flow. Hysteresis in the calibration of the flow system was evaluated. Calibration of the system was performed by filling as well as by draining. Any differences were within the reported tolerances of the pressure transducers.
The system responsiveness was tested by repeated injection and withdrawal of water. The response time for the system was consistently <1 s and appeared to be because of oscillations in water level. The overall baseline drift was 0.000027 g s-1 based on eight baseline segments over a 4-min period. This is an order of magnitude less than observed during the static drift test.
Comparison with Balance
Accuracy of the flow measuring system can be examined in several ways. Bias can occur in the start time as well as in the mass. A direct comparison with a scale is only possible with simultaneous measurements as was done with a modified measurement cylinder (Fig. 2). A separate calibration was performed for the single transducer used. The equation was:
where MI is mass inflow and R is the transducer reading (increasing values). The regression coefficients for the intercept and slope were -35.80 ± 0.155 g and 26.44 x 10-4 ± 0.032 x 10-4 g, respectively. The R-square value for the model was 0.9999 and the standard error of prediction ranged from 0.059 to 0.117 g. (The standard error of prediction was larger than for the previous measuring cylinder because the larger diameter of the cylinder resulted in a larger mass of water representing a similar change in R).
Time accuracy was ensured for both the pressure transducer and scale output as follows. At the start of a test the real time clocks of the computer and data logger were synchronized, and as both systems were started the initial real time was captured. However, the actual data alignment can only be to the nearest second. This is because the smallest value for the scale and datalogger real time is multiples of seconds.
Figure 3
shows the output of the scale and pressure transducer. In the initial 4 min, the slope of the scale response was smaller than that of the pressure transducer. This showed the bias of the scale because of time averaging and a slower responding system. As time elapsed the two curves came together and their slopes were similar (outside range of Fig. 3).
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 3. Output from simultaneous readings of scale and data logger. The inset shows data from the initial 400 s. Output from scale shows an initial lag in response.
|
|
Application to Soils
The apparatus shown in Fig. 1 was used with a sand core to conduct a falling head experiment. The falling head method is a standard technique (Klute and Dirksen, 1986) used for low conductivity soils. The technique has not been applied to high conductivity soils, in part because of the inability to measure the rapid drop in water level. Figure 4
shows an example of a falling head experiment where a rapid response can now be accurately determined.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 4. Application of instrumentation to the determination of hydraulic conductivity using the falling head method.
|
|
Additional applications to soils are possible. The tube labeled "outflow" in Fig. 1 can be directly connected to an unsaturated soil core. Opening the valve will result in an infiltration experiment. Alternatively, a saturated core can be positioned so the outflow flows into the falling head cylinder that receives the outflow. In this case the valve is superfluous.
|
DISCUSSION
|
|---|
The measurement of small quantities of water moving rapidly is a useful technique to develop new methods for determining soil hydraulic properties. The system is usable for either inflow or outflow of water to soils. The proposed method allows for more rapid measurement than is possible using scales. Further, the method is adaptable for both laboratory and field methods. The rates of flow observed can be adjusted simply by changing the diameter of the cylinders used with the pressure transducers. One important feature of this apparatus is the ability to respond accurately over a range of water flow rates. Figure 3 clearly shows the delayed response of a scale at a flow rate of about 15 g s-1. This is clearly a limitation of scales when developing fast response systems.
Drawbacks of the system include the need to calibrate each apparatus separately because of differences in cylinder size, control circuit parts, and pressure transducer. However, the calibration curves are linear. With this knowledge it is possible to develop simpler two or three point calibration procedures. A second concern is the extreme sensitivity of the transducers. The initial readings (<0.3 s) were usually censored because of spiking. These pressure fluctuations represent a limitation of the flow system and show the responsiveness of the pressure transducers.
The proposed method appears to meet the objectives as stated in the introduction. It can be used in the laboratory with one-step out flow methods to obtain initial data with confidence. Additional uses for this equipment include the development of falling head techniques for use with high conductivity materials.
|
NOTES
|
|---|
A contribution of the University of Nebraska Agric. Research Division, Lincoln, NE 68583. Journal series no 13334.
Received for publication June 25, 2001.
|
REFERENCES
|
|---|
- Borcher, C.A., J. Skopp, D. Watts, and J. Schepers. 1987. Unsaturated hydraulic conductivity determination for fine-textured soils. Trans. ASAE 30:1038–1042.
- Green, R.E., L.R. Ahuja, and S.K. Chong. 1986 Hydraulic Conductivity, Diffusivity, and Sorptivity of Unsaturated Soils: Field Methods. p. 771–798. In A. Klute (ed.) Methods of Soil Analysis. Part. 1. 2nd ed. Agron. Mongr. 9. ASA and SSSA, Madison, WI.
- Klute, A., and C. Dirksen. 1986 Hydraulic Conductivity and Diffusivity: Laboratory Methods. p. 687–734. In A. Klute (ed.) Methods of Soil Analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA and SSSA, Madison, WI.
- Mu'azu, S., J. Skopp, and D. Swartzendruber. 1990. Soil water diffusivity determination by a modified one-step outflow method. Soil Sci. Soc. Am. J. 54:1184–1186.[Abstract/Free Full Text]
- SAS Institute. 1994. The SAS system for Windows. Release 6.10. SAS Inst., Cary, NC.
TOP
ABSTRACT
INTRODUCTION
