Institute of Physics Publishing Journal of Physics: Conference Series 15 (2005) 300–305
doi:10.1088/1742-6596/15/1/050 Sensors & their Applications XIII
Design and investigation of high-speed, large-force and long-
lifetime electromagnetic actuators by finite element modelling
S H Khan1, M Cai1, K T V Grattan1, K Kajan1, M Honeywood2 and S Mills2
1
Measurement and Instrumentation Centre, School of Engineering and Mathematical
Sciences, City University, Northampton Square, London EC1V 0HB, UK
2
Sortex Limited, Pudding Mill Lane, London E15 2PJ, UK
[email protected]
Abstract: Electromagnetic (EM) solenoid actuators are widely used in many applications such
as the automobile, aerospace, printing and food industries where repetitive, often high-speed
linear or rotating motions are required. In some of these applications they are used as high-
speed ‘switching’ valves for switching pneumatic channels. This paper describes the finite
element (FE) modelling and design of high-speed solenoid actuators. Operating at frequencies
between 150-300 Hz, these actuators are unique in terms of the large force they produce (8-15
N) and the requirement for very long lifetime (2-5 billion cycles). The complex nature of
electromagnetic, motional and thermal problems is discussed. The methodologies for FE
modelling of such high-performance actuators are developed and discussed. These are used for
modelling, design, performance evaluation and prediction of the above high-speed actuators.
Modelling results showing some of the key design features of the actuators are presented in
terms of force produced as a function of various design parameters.
1. Introduction
Electromagnetic (EM) solenoid actuators of wide variety of sizes, shapes, power outputs and
technological realizations are used in many applications where discrete cyclic motions are required.
Compared to other actuating mechanisms based on, for example, piezoelectric and hydraulic principles
EM actuators are simpler, cheaper, repairable, robust, and easier to manufacture. Although the analysis
and design of such actuators are well covered in the literature [1-3], there are, however, very few
situations which involve commercial EM actuators that normally operate under continuous duty cycles
at frequencies between 150-300 Hz, producing a relatively large force (8-10 N). Combined with a
stroke length of 0.05-0.1 mm, ‘on’ and ‘off’ times of 0.2 ms and 0.46 ms, and a requirement for
multibillion cycle operation (in excess of 5 billion cycles) without maintenance, these actuators
operate at the limit of what can be achieved by solenoid-based EM actuator technology.
This paper investigates one of the designs of actuators of this type, used as pneumatic ejector
valves in high-speed optical food sorting machines [4]. The whole area of systematic research into
high frequency EM valve-based ejector technology is new. To our knowledge, apart from previous
‘trial and error’ methods used in industry, no comprehensive research has been done so far in this area.
Given the demanding fast duty cycle, high frequency of operation, high reliability and robustness
needed for the ejector technology, the EM ejector valve being developed should be unique, meeting
very tight design and rigorous performance specifications. Some of the novel aspects of the design
being pursued include the minimisation of the overall size of the valve, and meeting the constraints of
© 2005 IOP Publishing Ltd 300
301
the particular geometric and material parameters of the magnetic circuit and the dedicated control
circuitry used.
2. Modelling of High-Speed Actuators
Figure 1 shows the simplified schematic of one of the designs of a high-speed solenoid actuator used
as ejector valves mentioned above. It is essentially an on/off valve actuator whose active components
comprise an excitation coil wound around a magnetic core that attracts or releases a movable valve
plate depending on the excitation state of the coil. Very tight design and rigorous performance
specifications make these ejectors used in high-speed sorting applications unique in terms of their
design, manufacture and reliable exploitation. Typically, an individual ejector fits into an array of 32,
64, 96 or 128 ejectors and to maintain the maximum packing density they are often packaged in
serviceable modules of two, four or more ejectors. All these factors put severe constraints on a number
of the geometric dimensions (e.g. TCL, TV shown in figure 1) of individual ejectors, some of which
are directly linked to their overall lifetime and the ‘resolution’ of the optical sorting to be carried out.
Figure 1. Simplified schematic showing the main
constructive features of a high-speed EM actuator used as
an ejector valve in an optical sorting machine.
EM actuators rarely operate in the steady state and various operational factors like start-stop duty,
operating frequency, response time and damping have a significant influence on their design. The EM
part of the system is represented by electric and magnetic circuits with self-inductance, resistance and
reluctance which are subject to variations, in general, due to eddy currents, saturation conditions,
motional electromotive force (e.m.f.), demagnetisation and hysteresis. The mechanical part is
represented by friction, damping, elasticity and inertia as well as external forces. Together these two
parts form an equivalent electromechanical system which has to be optimised against the performance
requirements and the complex behaviour of which is subject to static and dynamic analyses of these
actuators. For high frequency EM ejectors, the thermal problem of temperature distribution and heat
dissipation is of vital importance. Like all other electrical devices, they generate losses (e.g. ohmic
loss in the winding, core and any circulating-current losses, etc.) manifested in the form of heat and
temperature rise which may be considered to be one of the dominant factors in limiting the life of most
high-speed actuators. Temperature-rise may significantly increase the winding resistance, impairing
control. The frequency response may be altered because both the electrical and the mechanical time
constants are temperature sensitive. Often, when tightly packed, ejectors may create serious problems
of heat sinking from individual units. Also the diverse combination of materials used and the very
extensive duty cycles may give rise to a myriad of thermal problems, unique only to high-speed
actuators. The solutions to these problems will have a direct impact on the cost, size, reliability and
feasibility of a given design. Compared to conventional solenoid actuators, the nonlinear and transient
302
EM, thermal, and motional problems being solved in the above ejector valves pose substantial
challenges because of their high frequency of operation and the requirement for a continuous and fail-
safe duty cycle.
In general, the mathematical model of the above electromechanical system can be adequately
represented by the following four differential equations given below. In summary these represent (1)
an electrical circuit equation for the excitation coil and control circuitry, (2) a nonlinear magnetic field
equation (Poisson’s equation) for the flux, the change of which changes the EM energy storage in the
system and produces the magnetic force, (3) a mechanical equation for this force, load (e.g. pneumatic
force), friction, inertia, acceleration, speed and displacement, and (4) a nonlinear thermal diffusion
equation for the conduction of heat produced by electrical power losses.
dΨ (i, z )
u (t ) = iR + N (1)
dt
∂A
curl(ν curl A) = J − σ + σV × (curl A) (2)
∂t
d 2z dz
Fm (i, z ) = m +B + Kz + Fe (3)
2 dt
dt
∂T
ρC − ∇ ⋅ [k (T )∇T ] = q B (4)
∂t
In the above equations u(t), i and Ψ(i, z), and z are the applied voltage, coil current, flux linkage with
the coil, and the displacement of the valve plate respectively, R and N are the coil resistance and the
number of turns in the coil, J, A, V are the coil current density, magnetic vector potential, and the
plunger (valve plate) velocity; m, B, K, Fm and Fe are the mass of the valve plate, viscous damping
coefficient, spring constant, magnetic force and the load force respectively; and T, and qB are the
temperature and the internal rate of heat generated per unit volume respectively. The material
parameters ν, σ, ρ, C and k denote the magnetic reluctivity (ν=1/µ, µ is the permeability), the electric
conductivity, density, specific heat and the thermal conductivity respectively. In general, these
equations are nonlinear and inseparable. The current produced by equation (1) creates the magnetic
field given by equation (2) and produces the magnetic force which causes the displacement, speed and
acceleration of the actuator obtained from equation (3). The current also generates the heat (per unit
volume) and the resulting temperature distribution given by equation (4). There are two main
approaches to the coupled solution of these equations: the direct coupled approach and the indirect
coupled approach, neither of which alone is suitable to incorporate the whole array of factors which
are expected to be encountered in the practical exploitation of high-speed ejector valves. These assume
the necessity for qualitative and quantitative assessments of those factors that introduce nonlinearities
in the system, in order to justify the use of one (or several) of these methods for modelling and
simulation purposes. For example, if the motional and eddy-current effects are negligible, the winding
inductance can be taken as constant and the decoupling of the equations would be adequate for
dynamic analysis. On the other hand, the direct coupled approach would be more appropriate if the
eddy-current effects are negligible, but saturation and motional nonlinearities are prominent. Thus a
methodology which includes the provision for using both coupled or/and decoupled solutions of
electromechanical equations by a priori qualitative and quantitative justification would be most
appropriate. The methodologies for modelling and design of EM ejector valves are based upon the
modelling and computation of the 2D/3D nonlinear magnetic field distribution using the numerical
finite element (FE) technique. This involves the steady-state and transient solutions of the nonlinear
Poisson’s equation for which there are no analytical solutions. The results are used for design
optimisation and for investigating the effects of various geometric, material, EM and mechanical
parameters on the output performance of ejectors. The thermal modelling involves the development of
303
2D/3D thermal models and the FE solution of the steady-state and/or transient heat transfer equations
given by equation (4) above. The heat sources needed for this are given by the various losses
mentioned above. The coupling of the magnetic field and the thermal equations (owing to the
dependence of the power density on the magnetic vector potential and the temperature dependence of
the magnetic permeability and electric conductivity) may be realised either by indirect coupling (in
which the equations are solved separately and coupled by means of power density and an iterative
process is used to compute the power density and the temperature distribution) or by direct coupling in
which the equations are solved simultaneously. The prime aim here is to obtain a vital insight into the
thermal behaviour of the ejector valves and to enable quantification of the effects of various factors
that affect such behaviour. The FE models also enable the simulation of possible modes of thermal
failure and create an essential basis for the design of a predictable, thermally stable and reliable
actuator sub-system.
3. Results of Analysis and Discussion
3.1. Effects of material and geometric parameters
In the design of an EM actuator of the aforementioned performance requirements, the material and
geometric parameters play a very important role. The need for delivering a relatively large force and a
very fast response time under multibillion cycle operational regimes at 300 Hz puts an added
importance on some of the crucial design parameters. Some of the effects, which would otherwise play
a minor role in the design of low-frequency actuators operating under duty cycles far below that is
required of the actuator discussed here, are likely to be magnified under multibillion cycle operations.
Figure 2. Schematic of the EM actuator
given in figure 1 showing the lines ab,
cd and ef along which magnetic flux
densities are calculated in figures 3, 4,
and 5. (not drawn to scale)
One of the crucial issues here is the mechanical wear of the moving part of the actuator i.e. the
valve plate. It has been established by extensive testing of the ejector valves mentioned above that the
mechanical wear of the valve plate is the weakest ‘link’ that contributes to the failure mode of these
Figure 3. Effects of the valve plate thickness TV on magnetic field distributions in
the yoke and the valve plate (a) along the lines ab and ef respectively and (b) along
the line cd shown in figure 2; TV=1.5 mm.
304
valves under multibillion cycle operation. The rate and magnitude of this wear seems to be dependent
upon the material and geometric parameters, some of which (e.g. mass, size) affect the dynamic
performance parameters (e.g. velocity, acceleration, mechanical response time). To quantify this, a
number of magnetic materials (such as Radiometal 4550, Armco, Hyperm 0, Hyperm Co50, etc.) have
been investigated. Also, extensive modelling studies have been carried out to investigate the effects of
geometric parameters. One of the aims here is minimise the thickness of the valve plate, TV (shown in
figure 1).
Figure 4. Effects of the valve plate thickness TV on magnetic field distributions in
the yoke and the valve plate (a) along the lines ab and ef respectively and (b) along
the line cd shown in figure 2; TV=2.5 mm.
Some of the modelling results are presented in figures 3-5, which show the effects of the valve
plate thickness TV on the distribution of magnetic flux in the yoke and in the valve plate along the
lines ab, cd and ef, shown in figure 2. These, together with the magnetic force versus valve plate
thickness, F=ƒ(TV) curve, presented in figure 5 show that the force (an important design parameter) is
not only strongly dependent upon the thickness of the valve plate but also on the relative levels of
saturation in the yoke and the valve plate. The maximum force is obtained for the valve plate thickness
values between 2.5-2.75 mm, which correspond to almost identical levels of saturation in the yoke and
in the valve plate (figure 4).
Figure 5. Effects of the valve plate thickness TV on magnetic field distributions in
the yoke and the valve plate (a) along the lines ab and ef respectively and (b) along
the line cd shown in figure 2; TV=4 mm.
3.2. Thermal modelling and its validation
As discussed earlier in Section 2, the EM ejector valve investigated in this work poses a unique set of
thermal problems because of its small size, high frequency of operation, continuous duty cycle and the
very compact packaging in a linear array with many other ejector valves. In addition, the primary
305
mechanism of sinking of heat produced in the ejector valve by the high pressure compressed air which
the valve is meant to ‘switch’, also constitutes a challenge for adequate thermal modelling. Various
modelling investigations were carried out to investigate the thermal behaviour of the above ejector
valve using 2D FE models. Some of the results have been compared with the corresponding
experimental data, which in general, shows good agreement (figure 7). The experimental data
presented in figure 7 were obtained for those cases when there was no air cooling by the compressed
air. Although this does not constitute a normal mode of operation, at higher currents the loss of air
cooling causes significant over-heating of the excitation coil, resulting in its thermal failure after
several minutes of operation (curve 5 in figure 7). Since all the modelling and experimental studies
were carried out on single isolated ejector valves, this situation is likely to worsen for valves which are
compactly packaged together.
Figure 6. Variation of magnetic Figure 7. Experimental validation of
force F with valve plate thickness thermal modelling results for various
TV for a given excitation current. excitation currents I: 1, 3 and 5 –
experiment, 2, 4, 6 – modelling; 1, 2
– I=1 A, 3, 4 – I=1,5 A, 5, 6 – I=2 A.
4. Conclusions
Finite element modelling methodologies have been developed for the design and investigation of
electromagnetic and thermal behaviours of high-speed, large-force and long-lifetime solenoid
actuators used as pneumatic ejector valves in optical sorting machines for bulk food sorting. The
actuator designed so far has been shown to have performed between 1-2 billion cycles under life and
field trials. To our knowledge no other EM actuator with similar requirement specifications is capable
of matching this performance. It is believed that the above design is pushing the limits of performance
that can be achieved by actuators based on the EM solenoid principle. Work is continuing on the
further development of the models discussed and their use in actuator design and optimisation.
5. Acknowledgements
The authors would like to thank the Department for Environment, Food and Rural Affairs (DEFRA),
UK for financing the work under the Advanced Food Manufacturing (AFM) LINK Programme (AFM
189). The authors would also like to thank the project partners Sortex Limited, UK and Vogan and
Company, UK for their active collaboration and support.
References
[1] Pawlak A M and Nehl T W 1988 IEEE Trans. Magn. 24 270-73
[2] Furlani E P and O’Brien M 1994 IEEE Trans. Magn. 30 4323-25
[3] Yamaguchi T, Kawase Y, Shiomoto H and Hirata K 2002 IEEE Trans. Magn. 38 361-64
[4] Low J M, Maughan W S, Bee S C and Honeywood M J 2001 Instrumentation and Sensors for
the Food Industry, ed. E Kress-Rogers and C J B Brimelow (Woodhead Publishing Limited)
chapter 5
