Edited by Jessica Shapiro
• Advances in electronics allow previously bulky controllers to mount on or near rotating machinery.
• Add-on software lets magnetic bearings act like expensive vibration-analysis units.
“Magnetic Bearings Come of Age,” Machine Design, Sept. 16, 2004,
“Wheel May Fly High,” Machine Design, Sept. 7, 2000,
What do you think of when you hear the word “bearings?” Whether you envision rolling-element, plain, or thrust bearings, odds are that power sources, control electronics, and magnetic fields don’t enter the picture. But active magnetic bearings are replacing oil-lubricated bearings in many applications.
Magnetic bearings support rotating machinery without contaminating or flammable lubricants and with high reliability, little or no maintenance, low frictional losses, less machine vibration, and improved monitoring and diagnostic capabilities. However, because they’re large, expensive. and require external controls — not to mention the complexity of integrating them into machinery — their applications have been limited to date.
Recent advances in magnetic-bearing technology, including miniaturization, simplification, and new integration techniques, have overcome many of these limitations. As a result, magnetic bearings are becoming more common in new machines designed for a variety of industries.
Where rolling-element bearings support shafts and other rotating components on balls or needles, magnetic bearings use stationary electromagnets positioned around the rotating components. Magnetic fields suspend the rotating parts within the bearings.
Typically, two radial magnetic bearings support and position the shaft in the lateral (radial) directions, and one thrust bearing supports and positions the shaft along the longitudinal (axial) direction. The bearings react against forces along the three translational axes and two angular axes to give a completely supported shaft with five axes of control. Ideally, the magnetic bearing offers little frictional resistance to rotational motion.
An active magnetic bearing consists of a stator and a rotor. The stator houses electromagnets and position sensors while the rotor turns with the shaft. In operation, the rotors are centered in corresponding stators so the two components do not contact each other.
A closed-loop feedback system controls shaft position. Position sensors detect local shaft displacements and signal a digital controller. The controller contains sensor-conditioning electronics, analog-to-digital (a/d) converters, digital processors, digital-to-analog (d/a) converters, power amplifiers, and a communications interface.
The digital processor calculates the current distribution in the electromagnets needed to restore the shaft to its centered position and instructs the power amplifiers to carry out the adjustments. The detection-calculation-adjustment cycle repeats approximately 15,000 times each second.
Like other kinds of bearings, magnetic bearings provide stiffness and damp vibration. But where other bearings’ stiffness and damping are constant, magnetic bearings’ stiffness and damping can be varied as a function of disturbance frequency. Engineers can change these two attributes simply by changing the input parameters to the algorithm the controller uses to redistribute the electrical current.
It is often convenient to describe the magnetic bearing as a transfer function with an amplitude and phase that vary with frequency. Engineers optimize this transfer function to ensure the bearing remains stable and resists force inputs over a range of frequencies.
The load capacity of a radial magnetic bearing is the product of the rotor diameter, the bearing’s active length, and the equivalent bearing pressure. Because magnetic bearings apply far less pressure than oil-lubricated fluid-film bearings, they need to be larger to provide the same load capacity as fluid-film bearings of the same diameter and length. End windings and position sensors further lengthen magnetic bearings beyond their active lengths.
Applying magnetic bearings
A small magnetic-bearing controller integrates into the drivetrain housing so the only required connections to the controller are dc power and an Ethernet network cable. A separate feedthrough at the top of the machine connects power leads to the motor/generator. Despite its small size, each power amplifier is rated at 7,500 VA. Machine-mounting the controller eliminates long cable runs for the coil and sensor wires, simplifies connectors, reduces EMI, and eliminates the need for special senor tuning.
The magnetic-bearing controller can also regulate the position of inlet-guide vanes and nozzles, perform machine-protection functions, and control other aspects of the machine using spare I/O and processing capability. Because the bearing controller gives the drivetrain built-in intelligence, external controllers are often not needed.
In some cases, control electronics can be completely integrated into radial or thrust magnetic bearings. For example, the housings of Synchrony’s Fusion radial bearings contain the control electronics, yet the bearings are smaller than previous generations of magnetic bearings that relied on an external controller.
For one line of Fusion bearings, the radial bearing has a load capacity of 300 lb. Thrust bearings from the same line handle loads to 1,000 lb. The Fusion bearings run on 48 Vdc and include dedicated Ethernet ports for high-speed communications and health monitoring.
The shaft and bearing end bells of a 250-hp industrial motor were modified to accommodate Fusion radial bearings. Thrust bearing integration is also possible for applications that need axial force capability, like vertical motors.
Despite being inherently larger, recent design innovations have let engineers shrink radial magnetic bearings by more than 30%. For one thing, adding electrical steel at the stator bore where the force originates ups the bearing pressure. At the same time, splitting the magnetic flux paths and isolating electromagnets from one another makes for a smaller stator outer diameter. Finally, new position sensors integrated with the electromagnets themselves have shortened bearing length.
Miniaturization and elimination of controller hardware has allowed engineers to shrink these components as well. Frequency-modulated (FM) sensing techniques make for smaller sensors and higher signal-to-noise ratios. FM sensing lets high-speed digital counters replace analog position sensors, eliminating the need for a/d converters.
Integrated electronics architecture that handles network communications, digital processing, and power-amplifier transistor timing signals further reduce the size of the digital-processing system. Direct connections between digital processors and power amplifiers means d/a converters aren’t needed. Finally, new control algorithms let power amplifiers provide stable magnetic-bearing performance with lower voltage and amperage ratings.
These innovations have shrunk the controller from the size of a household refrigerator to little more than the size of a DVD player. Where the bulky controller previously needed a separate enclosure, new smaller units can integrate into or mount on machines. Controller boxes purged with inert gas can even collocate in explosive-environment applications.
Carrying this miniaturization and integration one step further, it is now possible to buy magnetic bearings with the controller completely integrated into the bearing, eliminating separate controllers altogether.
Engineers encountering a magnetic-bearing system for the first time are often shocked by the quantity and complexity of the cables and connectors. The electromagnets require two wires for each coil and two coils/axis, so a five-axis suspension system had a total of 20 wires for the magnets alone. The sensors also require three to four wires/axis, adding another 15 to 20, not to mention any wires attached to temperature probes.
All these wires must route through the machine and exit to the controller. If the bearings operate in the process gas or fluid, the wires need a hermetic feedthrough.
The coil wires may carry currents of 20 to 50 A and have substantial high-frequency content from switching power amplifiers. Consequently, they are thick and must be properly shielded to cut electromagnetic interference (EMI). The sensor wires also have high-frequency voltages and are susceptible to noise from the coil wires and other sources.
Each electromagnet has a dedicated power amplifier, supplied with dc power, in the controller. In general, current flow to each electromagnet is an order of magnitude greater than the current supplied to each amplifier. This is because power supplied to the coils is reactive power, that is, the current lags the voltage by about 90° due to the inductance of the coil.
A small amount of real power is consumed by losses in the bearing. The power amplifiers should be located close to the coils so the wires carrying the largest currents are as short as possible. The wires carrying power to the amplifiers, on the other hand, can be long because the currents are fairly low and have a negligible high-frequency content.
Newer, more compact designs let controllers fit into the casing of the machine, mount on the exterior of the machine, or integrate into the magnetic bearings. Collocating the controller with the machine keeps wires between the controller and magnetic bearings short, simplifies cabling and connector, cuts EMI, and eliminates the need for special sensor tuning.
A power supply far from the machine can provide the 48 to 300-Vdc power the controller needs.
Better health monitoring
In the past, monitoring the health of a rotating system meant bringing in a large, expensive vibration-monitoring system with proximity probes, conditioning electronics, high-speed data acquisition, and alarms. However, a machine already equipped with a magnetic-bearing system can also perform health monitoring without additional hardware investment. High-resolution position sensors are inherent in magnetic bearings’ architecture, as are the digital processing and communication infrastructure to deal with the sensor signals.
The position sensors at each bearing can determine shaft orbits (X-Y trajectories) in the bearings without the need for additional sensor hardware. The magnetic-bearing controller itself can process most of the vibration data without a separate data-acquisition system and processor.
High-speed Ethernet networks then transmit the processed data to external computers which can visualize the orbits, perform advanced diagnostics, track data trends, maintain archives, and trigger alarms. The computer’s software simply extends the bearing controller’s existing functionality.
These extended capabilities, along with standardization, integration, and manufacturing advances have brought down the overall cost of implementing magnetic-bearing technology. Although more engineering effort goes into developing a new magnetic-bearing system, the ease of that integration has made magnetic bearings more economical for new and existing rotating machinery.