25 September 2001
Fluid Power System Design
by Peter Nachtwey
Taking advantage of fluid power may yield higher-quality output with lower life-cycle costs.
Many engineers use electric motors when fluid poweróeither hydraulics or pneumaticsówould be a better power choice. Traditionally, however, hydraulics and pneumatics haven't been thought of as power sources for precise motion. In the past, many hydraulically or pneumatically driven machines used two-position actuators, with motion bound only by limit switches. Valve controls for such systems earned the nickname "bang bang" due to the actuators' sound and shock. Owing in large part to advances in motion control technologies, fluid power control systems have come a long way and have opened up new opportunities in both improving quality and cutting life-cycle costs for machinery designers.
Fluid vs. Electromechanics
Electric motors must be sized for the maximum applied load, whereas fluid power sources (pumps) need to be sized only for the average load. Fluid power actuators are comparatively small, even for applications involving heavy loads. Fluid power's advantage is greatest when the motion duty cycle isn't 100%. This is because a fluid power system's accumulator stores energy while the system isn't moving. Conversely, electric motors make sense in applications with continuous motion (e.g., conveyor applications).
In electromechanical power systems, the electric motor is typically located close to or directly on the motion axis. Fluid power, however, may remotely locate the pump (along with its associated noise and weight). Only the accumulator (the fluid pressure storage tank) must be located near the actuators.
In a system employing fluid power, a single pump can provide the energy for multiple actuators. This makes it an ideal motive force for multiaxis robotics applications. In such an example, mounting the pumps in a base location keeps the weight on the arms as low as possible.
Fluid power has an additional advantage in that its pressure can be held constant without applying significant additional energy. Compare this with driving an electric motor, where applying constant torque (as required in a press type of application) could cause the motor to overheat.
Fluid power is also much easier and more cost effective in applications that require both pressure and position control. Machinery such as presses, injection molding machines, and material transport systems that involve grabbing things (e.g., log positioning systems in sawmills) often requires position control to move an actuator into place, and then pressure control to fine-tune that motion and ensure a controlled amount of "grip." Other systems, such as press rolls in steel rolling mills or sawmills, position heavy rollers above the work in process, then apply a controlled amount of pressure on the roll. Electric motors apply a pressure force by generating torque, but holding torque requires continuous power. If the control system doesn't limit that torque, the motor might burn out.
If the application involves only rotary motion and that motion is continuous (as in a conveyor system), then there may be no advantage to using fluid power because the pump may be running all the time. On the other hand, if there are multiple conveyers in the system, then hydraulics may prove less expensive because the motors are small and may require only a single power source. In addition, with material transfer applications prone to binding due to material mishandling, fluid power, with its more compressible power transport medium, may be more forgiving of "jams" than electromechanical power.
Unique Design Issues
Designers developing fluid power systems for the first time will have to manage some new issues. Fluid power is most commonly used in linear motion, and the most important factor in planning linear motion systems is sizing the actuator cylinders. Clearly, the cylinder selected needs to be long enough for the stroke required. Where mistakes are sometimes made is in specifying the cylinder's diameter. The cylinder choice is crucial, as the system's natural frequency is roughly proportional to the cylinder's diameter. This natural frequency determines the system's maximum controlled acceleration rate. Therefore, if a system needs to accelerate twice as quickly, its natural frequency must be twice as high; this requires a cylinder diameter that's twice as large.
A common error is to use small-diameter cylinders that move very quickly. A significant amount of hydraulic force goes to compressing the "hydraulic spring" associated with the smaller cylinder. This causes the system to oscillate when it's no longer accelerating, and the "hydraulic spring" returns to its uncompressed state. Furthermore, there isn't much surface area for the hydraulic pressure to push against to provide the required force. Consequently, the system may not get to the desired speed in the distance required.
Because fluid is compressible, a system with a large-diameter piston is much "stiffer" than one using a small piston in a long, thin cylinder. Hence, systems with larger cylinder diameters won't compress as much when accelerating and more quickly accelerate and decelerate because there's more surface area against which to push. Because of the fluid medium's compressibility, it's harder to keep long, thin cylinders under precise control than shorter, wider ones. In general, the diameter must double to decrease the acceleration time by half.
Selecting and Sizing Components
After choosing the piston diameter for the desired acceleration, you'll need a pump that provides the fluid flow for the speed and acceleration you require. If the pump is too large, however, both fluid and the power that pumps it may be wasted. Fortunately, the calculation is relatively simple: If you want to travel at 60 inches (in) per second and you have a 3-in-diameter cylinder, you can compute the minimum fluid flow needed (which in this case is 424 in3 per second):
The speed of motion varies proportionately with the speed of oil flow, but if the goal is to make the system move in half the time, accelerations and decelerations must double to make a move of similar quality. Achieving twice the acceleration requires doubling the diameter (a fourfold increase in surface area). Quadrupling the area and doubling the speed requires the oil flow to be eight times higher.
A fluid power system's accumulator serves two purposes. First, it acts as a buffer, allowing the pump's power requirements to be time averaged. Second, it allows the system pressure to remain relatively constant so that the effects of motion control inputs remain relatively constant. This avoids the need to continually change the control input response relationships the motion controller uses to maintain precise control. A good rule of thumb is to make the accumulator large enough to ensure that the pressure doesn't change by more than 10% during the system's operating cycle. Further, in order to minimize system pressure losses, it's important to locate the accumulator closer to the valve than to the pump.
Fluid power systems use two types of valves: servo and proportional. With servo valves, a linear increase in the current through the valve coil directly moves the spool, causing a linear increase in the flow of oil through the valve. Proportional valves, on the other hand, have position feedback on the spool, which the valve amplifier uses to linearize the valve. Proportional valves are generally less expensive and more tolerant of contaminants than their servo counterparts, but these benefits often come at the expense of performance.
Motion control requires servo-quality proportional valves. Valves often have an overlap, or "dead band," in the center where the flow is blocked. This causes a nonlinearity in the system response, for which the motion controller must compensate. Zero-overlap valves are often necessary for optimum performance.
Proportional valves have another problem: At high pressures, the spool is harder to move because it must push against the oil pressure. Sometimes dual-staged valves are used to get around this problem.
Servo systems function similarly to an auto's power steering. Turning the wheel slightly diverts a little oil, which ports to either end of the valve spool, causing it to move.
For maximum system responsiveness to control inputs, valves should be sized to provide the required flow plus another 10% to 20% percent. On the other hand, if the valve is too large compared with the size of the cylinder, valve control will be coarse, as only a small part of the control range is being used. In a position/pressure application, this is critical because the system gain, when controlling pressure, is very high. Compensating requires very low controller gains, making the system harder to control.
If you need a larger valve for position control, then you might need two valves in the system: a coarse valve and a fine valve. You'll use the larger coarse valve for gross travel and the smaller fine valve for pressure control. Another way to solve this problem is to use jack rams. A jack ram is a smaller parallel cylinder that's used while traveling. While in transit, a small valve moves the smaller diameter cylinder at the required speed because there's no load. The smaller cylinder also moves the larger cylinder, which is filling itself from the tank. Just before the system goes into pressure controller mode, the valve that lets the larger cylinder draw oil from the tank is shut, and the small valve takes over, filling both cylinders. This reduces the system gain, allowing the small valve to easily control the pressure.
In laying out the system topology, mount the valves as close to the cylinder as possible and use tubing instead of hoses. This reduces both the volume of the trapped oil and the compressibility. In addition, the valve should be on top of the cylinder, so that any air in the system will automatically be carried back to the fluid reservoir.
To monitor pressure, place sensors in the bottom of the cylinders at either end, where they aren't affected by trapped air and there's less oil motion. A common mistake is to mount them in the manifold, where the venturi effects of moving oil decrease pressure readings. Turbulence in the oil flow may reduce the venturi effect, but in any event, the manifold pressure may not be the same as the pressure in the cylinder.
Electric motor systems typically use quadrature encoders connected to the motor shaft. Although this is convenient, it can lead to imprecise motion if backlash exists in the system. You can avoid this by using linear transducers (e.g., magneto- strictive displacement transducers, or MDTs). Unlike quadrature encoders, MDTs measure absolute position and don't require homing. MDTs also have pressure and temperature specifications that allow them to be inserted directly into hydraulic cylinders.
The same types of encoders measure rotation in rotary hydraulic applications as in electromechanical applications. Some of the newest encoders and MDTs provide synchronous serial interfaces (SSIs) to connect to the motion controller. Direct SSIs allow for very precise control, providing 24 bits of position information, which permits position resolution down to 2 microns.
Using Motion Controllers
Smooth motion with linearly changing acceleration extends machine life and improves product quality. Fluid power systems are capable of very smooth, precise motion when controlled by the correct motion controller.
This motion controller should perform both pressure and position control and interface directly with servo valves and transducers, without requiring additional translators or interface elements.
The motion controller provides gearing and complex motion control for fluid power systems with splines. Spline functions make implementing smoothly curving motion profiles as easy as providing the motion controller with end-point coordinates and instructing it to connect the dots.
The motion curve defined by the spline represents an axis's position as a function of time or another axis's position. Velocities and accelerations are determined by differentiating the spline equation at each axis position. Splines permit the easy graphical specification of complex motion profiles. The machine designer defines only the positions; the spline algorithm computes the acceleration and velocity necessary to get smoothly from one point to another. Under ideal situations, these points can be defined graphically with a computer-aided design type of tool, and the machine designer is relieved of the tedious calculations for each segment between the defining points.
Tuning fluid power systems is similar to tuning electromechanical systems. Electric servos have two main modes of operation. In velocity mode, the speed is proportional to the control output from the motion controller to the drive amplifier. In torque mode, the servo's torque or acceleration is roughly proportional to the amplifier's control output. Hydraulic systems operate only in a velocity mode, as oil flow is ideally proportional to control output from the motion controller. Velocity mode is more intuitive and easier to set up by running the system with open-loop controls (i.e., without feedback). In torque mode, the system must always be on closed-loop control because a constant open-loop voltage will cause the servomotor to accelerate and keep accelerating. Sending a zero-control output doesn't cause the servo to stop ó it just allows the servo to coast to a stop.
Tuning the proportional, integral, and derivative terms (P, I, and D) is similar to tuning a velocity-mode or torque-mode controller. However, the importance of the differential term is much greater when controlling an electric motor in torque mode. Torque mode requires the differential term to provide speed stability. In contrast, electric servo velocity-mode systems are easier to set up and usually don't require a differentiator because the drive amplifier provides this function. The downside is that both the drive amplifier and the motion controller must be properly tuned, increasing the effort required to ensure proper system operation. It's often easier to fix the drive amplifier's gain to a constant value and let the motion controller manage the motion profile solely in relation to its internal PID values so that all the gains are in one place. Because fluid power systems always operate in velocity mode, they share the advantages of simpler tuning with velocity-mode electric motor controls.
Another difference between the torque and velocity modes is how their feed-forward gains are set up. In addition to P, I, and D control parameters, many motion controllers also provide feed-forward parameters. Feed-forward terms in the control algorithm enable a control system to anticipate and proactively drive the motion, rather than react to transducer stimuli.
In velocity mode, the velocity feed forward is the most important term in a correctly designed algorithm. It provides a component to the control output proportional to the velocity. This means that if the motion profile is a trapezoid, control output will also look like a trapezoid (Figure 1). Acceleration feed forwards are required only to give the control signal an extra boost while accelerating and braking while decelerating, but they have no effect while the system is moving at a constant velocity.
In torque mode, the acceleration feed forwards are the most important term after the differentiator, and the velocity feed forwards play a minor part of overcoming frictional forces proportional to speed. In torque mode, the control out will rise proportionately to the acceleration and then drop to almost zero while in the constant velocity part of the profile (Figure 2). During the profile's constant velocity portion, the velocity feed forward supplies the output necessary to overcome friction as a function of speed. During the deceleration portion, the control output goes negative because the acceleration is negative. This causes a braking action that stops the motor rather than just letting it coast to a stop.
After the feed forwards are set up, the designer typically tweaks the P, I, and D gains to get the desired control. PID tuning is similar in both torque and velocity modes; however, it takes a certain amount of output to make a system move. The controller generates this output using five terms, generated by the acceleration feed forward, velocity feed forward, proportional gain, integral gain, and differential gain. The goal is to make the feed forwards do most of the work. This way, the PID contribution to the control output is small, and thus the error between the target and actual position is small.
Another difference between electric servos and fluid power actuators is that electric servos are rotary, whereas hydraulics may be either rotary or linear. Rotary systems require only one set of gains. Fluid power motion controllers-for example, hydraulic controllers-require two sets for linear cylinder applications. The surface area on either side of the cylinder piston is different because of the cylinder rod. This difference in area causes the maximum force ó and therefore system gain ó to be greater when extending than when retracting. A typical electric servo controller will find it difficult to control a hydraulic system because it usually has only one set of gains. The electric servo controller can be tuned to work properly in one direction only. Ergo, hydraulic motion controllers should have two sets of gains: one for extending and one for retracting. Having two sets is also handy in vertical applications where the load changes greatly, depending on whether the system is rising or descending.
Making Advantages Possible
Designers who understand ó and take advantage of ó the differences between fluid power and traditional electromechanical power can build machines that produce higher-quality output with lower life-cycle costs, especially in applications where precise control of large forces and smooth motion are required. In order to deliver fluid power's control benefits, however, you must take care in selecting and sizing the hydraulic system elements and in tuning the motion controller for optimal performance. MC
Peter Nachtwey, president of Delta Computer Systems, has more than 19 years of experience developing hydraulic, pneumatic, and vision systems for industrial applications. He graduated from Oregon State University in 1975 with a BSEE and served in the U.S. Navy until 1980. He has worked for I.E.C.C. and Applied Theory Inc. as a systems engineer. He became president of Delta Computer Systems in 1992. Contact him at 11719 NE 95th Street, Suite D, Vancouver WA 98682-2444; tel: (360) 254-8688; fax: (360) 254-5435; www.deltacompsys.com.