This article reviews flow controller that can drastically reduce the project costs. A flow controller has now been developed with an embedded microprocessor that is inexpensive to purchase, easy to program and reprogram for different applications, and simple to wire with the necessary inverters/drivers and analog-to-digital converter. With this system, project costs can be reduced by approximately 75 percent compared with systems using other control valves. Tests have shown that the response time to initiate flow correction is approximately .10 second, and that the valve is actuated to within the required control band in 2.17 seconds. A typical analog device used for valve position control includes float-type operators for liquid service, as in aircraft fuel-tank level control systems. The work shows that a miniature embedded microprocessor can be used effectively and economically for an intelligent control valve. Smart actuation through the microprocessor is not only efficient—it also provides the convenience of advanced modular electronics.
Historically, Controllers Used for flow control valves and pressure regulators have been analog devices manufactured for specific applications. Using off-the-shelf technology, however, a flow controller has now been developed with an embedded microprocessor that is inexpensive to purchase, easy to program and reprogram for different applications, and simple to wire with the necessary inverters/drivers and analog-to-digital converter. With this system, project costs can be reduced by approximately 75 percent compared with systems using other control valves. Tests have shown that the response time to initiate flow correction is approximately.10 second, and that the valve is actuated to within the required control band in 2.17 seconds.
A typical analog device used for valve position control includes float-type operators for liquid service, as in aircraft fuel-tank level control systems (the mechanics are similar to the water tank of a common household toilet). An analog control valve commonly incorporates a pressure sensor downstream from the valve to sense static or dynamic pressure. The valve then feeds the pressure back to the valve control mechanism to effect a desired motion of the poppet (or disc) if the required flow rate or downstream pressure is not sensed. This sensed pressure signal operates on a piston or diaphragm against a calibrated preload provided by a spring or a pressure-loaded cavity on the opposite side of the piston or diaphragm.
Increasingly, to achieve a versatile control strategy and reduce costs in short-run production applications, manufacturers are incorporating intelligent microprocessor- based controllers into the design of industrial valves (see "Improving Hydraulic Performance with Intelligent Valves," April 1996).
We have designed an intelligent controller by modifying a valve that is typically used for water, using an inexpensive stamp-sized microprocessor that we have designated Stamp C. The microprocessor we chose cost $69 with documentation and technical support. It uses an abbreviated instruction set that combines an 8088 assembler and Basic. We have tested our controller to determine realistic flow control and response times. Other points of interest have been the signal flow, the flow sensors, and the influence of the low and high set points of the software on the flow control bandwidth.
The flow control valve we chose is a gate valve that uses digital electronic feedback to control the gate position and thereby maintain a desired flow rate. The Stamp C micro-processor controller was embedded in the gate valve, which can flow in either direction. A flow system was constructed with the control valve integrated into. the system and an inlet pipe connected to a one-half-horsepower submerged flood pump. Our controller measures,. records, and transmits data such as required gate valve travel, including direction and flow rate (orifice differential pressure).
The analog output of two Amtek P-207 pressure transducers was fed to the digital electronics of the control system. The Amtek P-207 is a low-cost unit with an operating range of 0 to 30 pounds per square inch and an accuracy band of ±1 percent of the operating range at 25°C. Input/ output access was selected using the Stamp C controller with 32 bytes of RAM for the run-time variables. This controller is able to accept 500 sets of instructions with a clock speed of 20 megahertz. Program execution speed is 4000 instructions per second. A unipolar (four-phase), 24-volt (dc), 160-milliampere stepper motor with 1.8 degrees per step and with a connected gear mechanism was attached to the valve shaft. The stepper was run by an inverter! driver (Allegro P /N ULN2003), which is a high-current Darlington array. The inverter/ driver responded to pulses from the microcontroller, driving the stepper motor. The stepper motor provided a holding torque of 600 gram-centimeters.
To interface the differential pressure transducer (analog input) to the microprocessor, a serial analog-to-digital converter (ADC) was used. The ADC (National Semiconductor model ADC0831 CCN) required only three input/ output lines, two of which could be multiplexed with other functions. As originally set up, the program read the voltage at the ADC's input pin every 2 seconds and reported it, using a 5 volt reference. A 75-percent reduction in control valve response time was observed when the sampling time set by the software commands was reduced from 2 seconds to.001 second.
Warm water was pumped into the system at a rate of 14 gallons per minute. The flow rate was adjusted with the inlet valve until a differential pressure of 8 pounds per square inch was measured across a half-inch fixed orifice using the two pressure transducers, which are of the linear variable differential transformer (LVDT) type. The output of the transducers (1 to 5 volts) was fed into a type LM741 differential operational amplifier. The differential voltage was transmitted to the ADC and converted into binary code. Interfacing the ADC required three input/output lines, with the chip select line connected directly to the controller. The ADC's range of voltage input was established by the reference voltage (vref of 5 volts referenced to the negative input voltage pin (-Vin) equal to ground. Vref set the input voltage at a point where the ADC returned a full-scale output of 255, and -Vin set the voltage to return O. The digital output was transmitted to the microprocessor.
After a series of initial tests and evaluations with the specified set value, the processor commanded the motor to travel forward or backward or to maintain its position. The motor under control was affixed to a gear train (12:1) that amplified the mechanical capability. The 12:1 gear ratio was selected as a design tradeoff that permitted the use of the inexpensive stepper motor. The end of the gear was attached to a plug valve collar and to the driving shaft of the stepper motor. A 24-volt dc power supply was used to run the motor, and provided 5 volts dc to drive the ADC. The pressure transducer voltage, pressure readings, and flow rate data were collected and plotted to determine the transducer's linearity and the orifice differential pressure versus flow rate.
To check for proper operation, two procedures were used-starting with a 50-percent input flow and opening the inlet valve to 100 percent, and starting with 100-percent input flow and closing the inlet valve by 30 percent. Response times of 1.87 to 4.5 seconds were obtained after performing over 25 tests on the control valve, changing the flow characteristics of the inlet valve from 50 percent to 100 percent open for each test. The pressure drop across the fixed orifice was within 8 psi, and the flow rate at the 100-percent-open inlet position was 14 gpm. The variance associated with response time resulted from several factors: error associated with pressure transducer calibrations, relative error associated with the 8-bit ADC, undetected pressure drop across the fixed orifice, and human error in readings and measurements.
The inherent initial response of the valve to correct its position was within 200 milliseconds, and the selected dead band variation for this stage of testing was 0.03 volt, which equates to a flow change of 0.21 gpm. This means that the actual flow rate is no better than 14 gpm, ±0.21 gpm. Human and instrumental error in data sampling necessitated probability studies. The mean value of the response time to read the dead band was 3.02 seconds, the standard deviation was ±0.786, and the variance was equal to 0.6. Over 75 percent of the samples were distributed within the standard deviation.
Further testing was done on the control valve, starting with 100-percent input flow and closing the inlet valve 50 per percent, which reduced the flow rate to about 7 gpm. The pressure differential across the fixed orifice varied from 8 to 4 psi, and the control valve made a correction within 100 milliseconds. The response time of the control valve (to reach the dead band) varied from 1.66 to 3.44 seconds, with a mean of 2.17 seconds. The same procedure was used to determine the standard deviation of 0.4 72 and the variance of 0.234. Over 84 percent of the data were distributed within the standard deviation.
Another step in our testing was to examine the flow-versus- travel characteristics of the gate valve and then to compare the performance of the control valve at a later time. Ten consecutive tests of the control valve over a three-day period resulted in no necessary adjustments. A significant advantage of this control valve was that it al- lowed a personal computer to be used for troubleshooting. By examining the flow characteristics versus the gate travel and then comparing the performance at a later time, we could easily reprogram the microprocessor to modify the step rate of the stepper motor.
Our work shows that a miniature embedded microprocessor can be used effectively and economically for an intelligent control valve. Smart actuation through the microprocessor is not only efficient—it also provides the convenience of advanced modular electronics.
A 75-percent reduction in control valve response time was observed.