In this section, a review of the properties in magnetic bearings and bearingless machines is given. A common practice in magnetic levitation is to use magnetic bearings to generate magnetic suspension. The so-called conventional magnetic bearings have four static electro-magnets placed around a rotating shaft; this enables radial control in the directions of x- and y-axes (Schweitzer, et al., 2009). Having no physical contact with the magnetic bearings makes the rotor have no wear, and therefore lubrication is not required. (Chiba, et al., 2005)
When separate radial bearings are placed at both ends of the rotor, the rotor needs to be longer than that in a bearingless machine, where integrated levitation coils are used. This can lead to problems such as high rotor elasticity. Even though the elasticity of a rotor can be reduced, the integrated levitation coils are often more preferred solution. (Majewski, et al., 2007; Chiba, et al., 2005)
In a rotor with rigid body, the distance between any two given points remain the same i.e., there is no bending regardless of any external forces exerted on it. (Chiba, et al., 2005;
Oshima, et al., 1996) Fig. 2.3 represents a traditional AMB supported high speed motor.
Figure 2.3. Rotor with high speed passive magnet brushless motor and magnetic bearings: 1. rotor, 2.
stator, 3. radial AMBs, 4. axial AMB. Adapted from (Gieras, 2010).
In a bearingless drive, the function of a magnetic bearing is integrated into the motor itself to create a bearingless unit, which generates radial forces as well as rotational torque. This integration makes the system more compact and allows the use of a shorter rotor. Bearingless machines also require fewer inverters than systems with magnetic bearings. It should be noted that bearingless machines function well in vacuums and sterile environments because of their characteristics. In addition, being able to dampen down vibrations is often sought
after property in these high frequency systems and the ability is more noticeable in systems with active suspension. (Schweitzer, et al., 2009; Chiba, et al., 1994)
The complexity of bearingless machines offers a multitude of freedoms in constructive de-sign. Different machine setups may have nearly the same dynamic behavior and may differ only slightly in their operation characteristics. The common bearingless machine consists of motors such as permanent magnet motor, switched reluctance motor, induction motor, sali-ent-pole synchronous motor, or cylindrical rotor synchronous motor accompanied by inte-grated levitation coils or external magnetic bearings. (Chiba, et al., 1994; Silber, et al., 2005;
Oshima, et al., 1996)
Permanent magnets
Permanent magnets offer strong attractive forces without the use of power. Permanent mag-nets can be utilized in variety of applications due to their compactness, reliability and very long lifetime. (Yonnet, 1978)
Fig. 2.4 represents radial passive magnetic bearing system, where a rotor is being attracted to both negative and positive directions of y-axis making it stable in perpendicular directions.
Figure 2.4. Radial permanent magnet bearing suspending a rotor with attractive forces.
The magnetic pull force of a single permanent magnet depends on their magnetic flux density B or their magnetizing field H, the cross-sectional area of the magnetic poles A and the per-meability of vacuum π0
πm =π΅2π2π΄
0 = π0π»22π΄. (2.1)
The (2.1) works also with electromagnets. (University of Surrey, 2010) Electromagnets
Fig. 2.5 represents a basic 1-DOF AMB control loop. The control loop consists: 1. rotor, 2.
sensor, 3. controller, 4. power amplifier and 5. electromagnet.
Figure 2.5. The basic AMB control loop and its elements. Adapted from (Schweitzer, et al., 2009).
The magnetic pull ππ generated by the electromagnet and earthβs gravitational pull ππ = ππ are tried to be kept in balance. The sensor measures the displacement of the rotor from in-tended air gap π₯0 and feeds the information to the controller. The controllerβs purpose is to maintain the rotorβs place at its intended position. In addition to balancing the affecting forces, the controller must stabilize the system i.e., dampen fast movements of the rotor and reduce steady state error. Finally, the controller sends out a current command signal to the power amplifier, which transforms this signal into control current for the coil, thus generat-ing desired magnetic force ππ. (Schweitzer, et al., 2009)
When modeling the system, several forces and laws of nature must be considered. To sim-plify the model of the system, neither dynamics of the sensor nor electronics of the power amplifier are to be taken into account. Although, these must be taken account in the final simulation model.
Despite the strong nonlinearities, the AMB systems can be controlled with a linear control system. Thus, the position and the current stiffness of the magnetic pull ππ must be linearized at the operating point (π₯0, π0, ππ) as displayed in Fig. 2.6. (Schweitzer, et al., 2009) Result-ing linearized model only works in the close proximity of the operatResult-ing point (Larjo, 2006).
Figure 2.6. Linearization at the operating point: (left) relationship between force and displacement i.e., position stiffness; (right) relationship between force and current i.e., current stiffness.
To eliminate all operating point quantities from the resulting equations new variables are introduced for the force π, the control current πc and the displacement x as follows
π = πmβ πg, (2.2)
πc = πβ²β π0, (2.3)
π₯ = π₯β²β π₯0, (2.4)
where πm is the magnetic pull of the electromagnet, πg is the earthβs gravitational pull, πβ² is the current in the coil, π0 is the constant bias current, π₯0 is the intended length of the air gap between the electromagnet and the rotor and π₯β² is the displacement from said π₯0. The value of π₯β² is limited to the length of the air gap at given time. (Schweitzer, et al., 2009)
Linearizing the system yields the following equation for produced force at the operating point
ππ₯Μ = πsπ₯ + πiπc, (2.5) where m is the mass affecting the bearing, π₯Μ is the acceleration of gravity, πs is the position stiffness and πi is the current stiffness. How the forces actually affect the rotor can be found in Fig. 2.7. (Schweitzer, et al., 2009)
Figure 2.7. Forces affecting a rotor: a) Magnetic pull produced in the electromagnet, b) Force aiding or resisting the magnetic pull depending on the displacement i.e. when π₯ < 0 the force is resist-ing the pull and when π₯ > 0 the force is aidresist-ing the magnetic pull, c) Constant force generated by the mass of the rotor and the acceleration of gravity.
To make an AMB system stable it needs to be actively controlled. To stabilize the system, a control law is needed, which purpose is to send out current command signals that result in force vectors πm and πg cancelling each other (Fig. 2.8).
Figure 2.8. Active magnetic bearing system with one degree of freedom, where the forces are caused by the opposite electromagnets and gravitation. Control current πc is the output of the control system; π0 is the constant biasing current.
Usually, the rotorβs radial axis is controlled with two electromagnets on opposite sides; this enables the production of both the positive and negative forces. The upper electromagnet is
controlled with the sum of constant bias current π0 and control current πc and the lower elec-tromagnet with their subtraction. (Kurvinen, 2016)
The electromagnetic pull for U-shaped electromagnet is calculated by
πm =14π0π2π΄πβ²π₯22, (2.6)
where π is the number of loops in coil and π΄ is the projected area by the electromagnet.
(Kurvinen, 2016)
In practice, radial magnetic bearings have magnetic forces affecting the rotor in an angle of Ξ±. This is due to the deviation in the angle of the magnetic force in relation to the gravitational pull. Sometimes it is intentional to distribute the weight of the rotor. Thus, the magnetic
The force πx is achieved in the direction of x-axis; it represents the difference between the forces pulling from opposite directions. Therefore, force πx can be evaluated with
πx = π+β πβ = π ((π(π₯0+πc)2
0βπ₯)2β(π(π₯0βπc)2
0+π₯)2) cos (πΌ), (2.9) Linearizing (2.9) with the presumption that π₯ βͺ π₯0 yields
πx =4πππ₯ 0
02 cos(πΌ) πc+4πππ₯ 02
03 cos(πΌ) π₯ = πsπ₯ + πiπc. (2.10) When assumed that the cross-sectional areas of the magnet π΄fe and the air gap π΄a are equal, the position stiffness and the current stiffness can be defined with (2.11) and (2.12)
πi =4πππ₯ 0
2-DOF system with radial active magnetic bearing
Fig. 2.9 represents a practical AMB control loop; it shares all the same elements as the sys-tem in Fig. 2.8, but with the addition of one axis and tilting of 45 degrees. The reasoning behind tilting the bearing is to distribute the weight of the rotor to both axes. When the bearing is tilted 45 degrees in respect to gravitational pull, it enables symmetrical control for both axes, with each axis supporting 1/β2nd part of the rotors weight.
Figure 2.9. Radial active magnetic bearing system with two degrees of freedom tilted in 45 degrees.
Resulting forces for πx and πy become
3 BEARINGLESS MACHINES
In this section, a review of control design, modeling and cross-coupling in bearingless ma-chines are given. In addition, the properties of the used test rig and the values used in simu-lations are presented. The examination focuses on 4-DOF magnetic levitation system with a rigid rotor. Even though the modelling of the system is examined as an AMB supported rotor and AMB terminology is used, the designed position controller can be used in the actual test rig, which uses bearingless motors instead of AMBs.
3.1 4-DOF SYSTEM WITH RADIAL ACTIVE MAGNETIC BEARINGS
Fig. 3.1 represents the principle behind the degrees of freedom that a system with two radial AMBs withholds when longitudinal movement is disregarded. The system acts like two sep-arate 2-DOF systems when the cross-coupling is not considered i.e., both radial AMBs act as separate units.
Figure 3.1. Radial active magnetic bearing system with four degrees of freedom and a rigid rotor.
In a typical symmetrical configuration, the rotor is supported with two radial magnetic bear-ings placed within the same distance of the rotorβs center of mass. Although, there are cases where asymmetry is preferable i.e., systems where the rotorβs center of mass is not in the middle. The electromagnets that generate forces in the direction of x-axis are controlled with currents πc,xA and πc,xB. The forces parallel to y-axis are controlled with currents πc,yA and πc,yB. (Schweitzer, et al., 2009)
The input for the model is bearing force
πb = [πxA, πyA, πxB, πyB]π, (3.1)
and the output is the position vector
πb = [π₯A, π¦A, π₯B, π¦B]π. (3.2) The equation of motion for the rotor can be written
πbπΜb = πb+ π g, (3.3)
where πb is the mass matrix of the rotor, πb is the bearing force vector and π g is the force vector caused by the rotorβs mass. It should be noted that the gyroscopic effect is considered insignificant, and therefore it is not taken into account. In the model, the x- and y-axes have been tilted 45 degrees in respect to gravitational pull, as seen in Fig. 3.1, which results in
πg = β ππ
β2(πβπ)[βπ βπ π π]π, (3.4)
where π is the rotorβs mass, π is the acceleration of gravity, π is the distance from bearing A to the rotorβs center of the mass and π is the distance from bearing B to the center of the mass (Fig. 3.1).
For state model presentation, a linearized vector ππ can be used
πb = πsπb+ πiπc, (3.5)
where πs is the position stiffness matrix, πi is the current stiffness matrix and πc is the con-trol current vector. The force caused by the rotorβs mass is invariable, so it can be compen-sated with constant current πcg
πππ = βπiβ1π g. (3.6)
By stating that πc = πc+ πcg the equation of motion (3.3) can be expressed as
πbπΜb = πsπb+ πiπc, (3.7) where πb is the mass matrix with respect to bearingsβ frame of reference. The πb can be found with the following
πb = πβTπcπβ1, (3.8)
where πc is the mass matrix with respect to the rotorβs center of mass
and π is the transformation matrix
π = [
The constant π½ found in (3.9) is the rotorβs inertia. (Larjo, 2006; Schweitzer, et al., 2009) State-space representation
In this section, a state-space representation of the system is given for the control design and the simulation implementation. The general equations for a continuous time-invariant state-space model are
πΜ(π‘) = ππ(π‘) + ππ(π‘), (3.11)
π(π‘) = ππ(π‘) + ππ(π‘), (3.12)
where π is the state vector, π the input vector and π the output vector of the system. π is the system matrix, π is the input matrix, π is the output matrix and π is the feedforward matrix.
(Larjo, 2006)
To make a state-space model, the differential (3.7) can be expressed as π = [πb
πΜb]. (3.13)
In addition, the second derivative πΜb needs to be solved from the (3.7) by multiplying it with πbβ1 which gives
πΜb = πbβ1πsπb+ πbβ1πiπc. (3.14) The state equations become
πΜ = [ 0 π
πbβ1πs 0] π + [ 0
πbβ1πi] πc, (3.15)
π = [π 0]π. (3.16)
(Larjo, 2006; Schweitzer, et al., 2009)