The millennium pump, shown in combined schematic-block diagram form in Fig. 1, consists of the basic electronically stroked pump kernel, plus several sensors used for control and diagnostics. The pump kernel for bus use is nothing more than the universal, electronically controlled, variable-displacement pump itself. We will examine the pump not as a component, but in the context of the system to which it must adapt. The elements in Fig. 1 form the framework for that look.
The communications interface performs the connection between the pump assembly and the enterprise-wide communications bus. The pump has a single input from the control computer that will allow the pump to be configured in any of several ways. By selecting different control algorithms in the control computer, the pump can act as a
- simple variable-displacement device,
- a conventional pressure-compensated pump,
- a speed-control loop,
- a position-control loop,
- a load-sensing pump,
- or any other desired performance, provided that the performance can be programmed into the control software. By monitoring the sensors on a periodic basis, other analytical software can compare the existing performance against the original performance and make assessments as to any degradation in performance.
The electromechanical interface will undoubtedly be a coil of some type (torque motor, force motor, or proportional solenoid) that requires a substantial current to provide appropriate mechanical motion of the valve, flapper, or whatever. That current will range between a low of 10 or 15 mA for certain torque motors, to a maximum of 2.5 to 3.5 A for certain proportional solenoids. This requires that the coil be driven by an analog servo or proportional power amplifier (S/P amp), shown as A in Fig. 2.
1. Modifications to the universal prototype pump allow greater interaction between the pump and electronic controls and sensors. This configuration lends Itself to two-way communication between devices and controls through the data bus.
Issues to Confront
There are several issues for the designer to consider. First, is the issue of where the amplifier will be located. Most manufacturers have moved control electronics to the valve bodies of servo and proportional valves. This move has significantly reduced headaches with cabling and wiring, which are the least reliable components, mostly because they are external and subject to severe environments and physical abuse. The proximity of the electronic module to the valve allows delicate wiring and cabling to be safely tucked within rigid, sealed enclosures, in addition to reducing conductor length.
More of the electronic adjustments can be done at the factory, then set and locked, reducing the need for tuning at application time. Effects of severe environments can be minimized by eliminating exposure of the electronic circuit boards to such destructive conditions as mud, dirt, dust, steam cleaning, chemical cleaning, or perhaps worse.
On the other hand, experience with valves has shown that in high-vibration applications, the mechanical integrity of the circuit board and its mounting are critical. Field reports of circuit board structure and mounting failures are rare, but in severe vibratory applications, they do occur. It is incumbent upon the electromechanical designers that the electronic devices onboard the pump be industrially hardened to withstand the rigors of the most abusive applications.
This requires designing for structural integrity, vibration and environmental testing and, finally, qualification of the entire pump subsystem. Having done so, the advantages of placing the electronic circuits directly on the pump far outweigh the problems that accompany remote siting. The final package will have the SIP amp safely enclosed and sealed, as well as equipped with any necessary cooling fins.
The second issue that must be faced involves the need for dc power, which is required by the local electronics such as the SIP amp. The dc power supply must also provide the power for the coil. Certain bus systems have been standardized to include two or three wires in the bus cable for carrying dc power to the devices. As long as the valve coil is at the low end of current requirements, (i.e., 10 to 20 ma), the bus can probably carry sufficient current for at least one of the millennium pumps.
If the pump is going to use a proportional solenoid requiring 2 or 3 amps, then a special cable with dc power will have to strung for that purpose; it is unlikely the bus will be able to supply that much current. Special tee connectors at the bus nodes permit a remote power supply to be added for special needs within a local area.
Alternatively, low-voltage ac power (probably 12- to 20-V RMS) could be connected to the pump. However, dc power would then have to be generated onboard the pump—requiring rectifier, filter capacitor, and voltage regulator circuits. Note that this external cabling defeats the advantage of buses—that is, reducing the number of all those accursed cables!
The Prototype Pump
The prototype pump of the future is drawn with seven different sensors to monitor the critical variables associated with the pump (Fig. 2). For any given application, this may be overkill, but the aim here is to look at possibilities and to discuss the issues—even in the extreme, if need be. Specific choices must be made with the application in mind. Refer to Fig. 2 regarding the following sensors:
- HPO is the pump outlet pressure sensor HN is the pump input shaft speed sensor,
- HN is the pump input torque sensor HD is the pump displacement or cam angle sensor,
- HPI is the pump inlet pressure sensor,
- HP, CD is the pump case drain pressure sensor, and
- HCD is the case drain temperature sensor.
Each sensor has issues associated with it that will affect the success of the pump in the marketplace—from the standpoint of utility, performance, and cost. There are three pressure sensors, and it behooves the designers to make them as identical as possible in order to reduce inventory and related costs.
2. The prototype millennium pump is shown here with seven different sensors, which are responsible for monitoring critical pump variables. These variables include outlet pressure, input shaft speed, Input torque, displacement, Inlet pressure, case drain pressure, and case drain temperature.
Issues with Pressure Sensors
The first issue that must be resolved is the question of pressure sensor range. The pump symbol used in Fig. 2 implies that the pump can stroke over center and, thus, reverse its output flow without reversing direction of shaft rotation. It is, therefore, the classic variable-displacement pump used in a hydrostatic transmission. As such, there is no real inlet or outlet port, because the two power flow ports change roles from inlet to outlet and vice-versa.
Either port can be at high pressure and either can be low. The ranges of the two sensors, therefore, would most likely be the same, requiring at least several thousand psi capability in each. The case drain would, most likely, have a very low range requirement (perhaps no more than 100 or 200 psi), necessitating at least two different ranges and two different sensors.
Other applications, such as pressurecompensated pumps and load-sensing pumps, do not require over-center operation. In these cases, the inlet pressure sensor need only be capable of reading relatively low pressure, such as the pressures expected at the case drain port. Again, two sensors could be identical, but have a low pressure range. The third would have to be capable of measuring high pressure.
The cost of sensors is cannot be ignored. The proliferation of pressure sensors used in cars—at least for low-pressure applications—has driven the price of semiconductor strain gage varieties into the range of $15 to $30 in OEM quantities. The cost of higher pressure sensors has yet to break the $100 barrier, even in large quantities. For this reason alone, the pump must be tailored to the application, and the number and kinds of sensors will be adjusted to meet specific needs. However, as the need for high-pressure sensors in hydraulics grows, costs will fall.
The specific type of pressure sensor—strain gauge, quartz crystal, or variable reluctance, to give examples—to use for a specific application also must be resolved. Semiconductor strain gauge types will likely be used when low cost is most important, but quartz crystal and variable reluctance types may get the nod for measuring high-frequency pressure ripple pressures.
Note that the prototype pump has no differential pressure sensors; all are gauge pressure sensors. This is not an oversight. It is just a realization that gauge pressure sensors will be the least expensive, regardless of type, create fewer problems in predicting performance, and provide the most information—especially when diagnostics is needed.
Challenges for Speed Sensors
A variety of issues surrounds the selection of the right speed sensor for pumps of the future. Foremost is the type of sensor. Probably not more than three different types are viable with today’s technology: rotating ferromagnetic protrusion with magnetic pickup, dc brush-type tachometer, and brushless dc tachometer.
The rotating protrusion-type sensor is very popular because of its simplicity. In the laboratory, it is common to mount a 60-tooth iron gear on the shaft and position a magnetic pickup near the OD of the gear. As the gear spins, each tooth generates a pulse as it passes by the magnet. The frequency of the generated voltage from the pickup is directly proportional to the speed of the shaft.
All that is needed is an electronic digital counter, which counts the number of pulses generated during a precise internal gate time. If the gate time is set to one second, and the gear bas 60 teeth, the frequency (in Hertz) is exactly equal to the speed in rpm. The gate time need not be exactly one second, and the number of teeth or protrusions need not be 60, because any combination simply requires a different scale factor to convert frequency to speed.
The first problem that arises with this type of speed sensor is that the digital circuitry must accumulate pulses over some finite amount of time. In the digital control world, the sequence typically goes something like this: The processor, by virtue of its programming, “knows'” when it needs speed information, and initiates the pulse accumulation process by setting the pulse counter to zero. If a 1-se.c gate is used, the data is one second old by the time it is ready.
To get around this, you might argue that the computer should start one second earlier; then the data will be ready when the computer asks for it. However, that does not work, because the data is still old when it is finally available. A workable ploy is to reduce the gate time. But doing so requires an increase in the number of teeth to maintain measurement resolution.
Perhaps a more workable plan would be to measure the period of the pulse train—that is, the time between two consecutive cycles. That way, the maximum time delay in the data is the time for one tooth to pass. But this is not without its own challenges. If the gear has is a great number of teeth, the time is short, and an extremely high frequency clock (several MHz) and high-speed counter are needed. In fact, period measurement argues in favor of fewer teeth, which now raises some practical possibilities.
The ferromagnetic protrusions need not be gear teeth, but any non-uniform iron body. Such a body could be the barrel in a piston pump, a key on a shaft, the vanes in a vane pump, or the teeth of the displacement gears in a gear pump. The major problem is figuring out how to mount the magnetic pickup to sense the protrusions magnetically without interfering with the normal hydraulic operation of the pump. Whatever is resolved regarding these issues, the electronic circuitry must either digitize the frequency data of the pulses streaming from the magnetic pickup or convert the frequency data into a proportional analog voltage.
Inexpensive frequency-to-voltage converters are readily available and easy for a competent electronic circuit designer to incorporate into a control circuit. But these devices are not without problems because they, too, have a considerable time delay associated with the conversion process. In the end, any speed-measuring method that uses frequency as its basis will suffer from time lag. And time lag can lead to instability if the speed data is used to close a feedback loop.
Another important issue with the magnetic pickup as a speed measuring device is that in their simplest manifestations, they will not work below some threshold minimum speed. The output voltage from the magnetic pickup varies not only in frequency, but also in amplitude. At a slow enough speed, the output voltage becomes so low that the digital counter cannot generate pulses from the teeth passing by.
In the specific case of pump speed, conditions near zero rpm normally are not very important, so magnetic pickups are viable. But if zero speed detection is needed, then the magnetic pickup must be replaced with a Hall-effect sensor. The Hall-effect sensor holds the advantage of not relying on a changing magnetic field to generate its voltage. It senses only the amplitude of the magnitude of the magnetic field and generates a proportional output voltage.
Thus, the output from the Hall-effect speed sensor varies only in frequency, not amplitude, right down to zero speed. Furthermore, the signal is very nearly a square wave, which the signal conditioning circuits can easily process.
But there is yet another method of speed measurement that relies on a pulse train for its base data source. If very high resolution and speed are needed—and frequency is the parameter of choice—optical encoders become a viable option. The optical encoder uses light emitting diodes, phototransistors, and a wheel with alternating opaque and transparent openings. The wheel must be mounted to the shaft whose speed is to be measured. The advantage of optical encoders is that there can be thousands of pulses per revolution of the shaft, giving very high resolution, even with frequency measurement and short gate times. Plus, they work down to zero speed, just like Hall-effect magnetic sensors. They are, however, the most expensive of the three types.