Design and prototyping of an optimised axial-flux permanent-magnet synchronous machine

Transcription

1 Published in IET Electric Power Applications Received on 2nd December 2012 Revised on 3rd February 2013 Accepted on 8th February 2013 Design and prototyping of an optimised axial-flux permanent-magnet synchronous machine Amin Mahmoudi 1, Solmaz Kahourzade 1, Nasrudin Abd Rahim 1, Hew Wooi Ping 1, Mohammad Nasir Uddin 2 1 UM Power Energy Dedicated Advanced Centre (UMPEDAC), University of Malaya, Kuala Lumpur, Malaysia 2 Department of Electrical Engineering, Lakehead University, Thunder Bay, Ontario, Canada P7B 5E1 amaminmahmoudi@gmail.com ISSN Abstract: This study presents the design and performance analysis of a prototype axial-flux permanent-magnet (AFPM) synchronous machine. First, the design of AFPM machine is optimised by genetic algorithm based sizing equation and finite element analysis. The design objectives of this machine are maximum power density, minimum total harmonic distortion (THD) of the sinusoidal back-electromotive force (back-emf) waveform and low cogging torque. Based on the optimised design of the machine a prototype 1 kw, three-phase, 50 Hz, four-pole AFPM synchronous machine is built. Then, the performance of the prototype machine is tested to see the cogging torque, torque speed characteristic, efficiency and the THD of the induced voltage. It is found that the prototype machine validates the design in terms of high-power density, lowest possible THD of the back-emf, low cogging torque while maintaining high efficiency. Nomenclature A B cr B cs B g B max c f D g D i D o D s D tot e(t) E pk f f α G i g G max G min G normal I i(t) I pk I rms K e k e k h K i K L total electrical loading rotor-disc flux density stator-core flux density air-gap flux density maximum flux density friction coefficient average diameter of machine machine inner diameter machine outer diameter slot depth machine outer diameter total phase-air-gap EMF peak value of phase-air-gap EMF electrical frequency angle frequency gene value air-gap length gene maximum value gene minimum value normalised chromosome current phase current peak value of phase current phase current rms value EMF factor eddy current constant hysteresis constant current waveform factor aspect ratio coefficient K p K w K φ l L cr L cs L e l e L pm L tot m m 1 n N ph N ph-p N ph-s N s N t p P cor P cu P den P e P h P i P m P nom P out P rot R s s cu T electrical power waveform factor winding distribution factor electrical loading ratio coil length rotor-core axial length stator-core axial length effective axial length of machine end winding length permanent-magnet length machine axial length total number of machine phases number of phases for each stator rotation speed number of winding turns per phase number of winding turns in parallel per phase number of winding turns in series per phase number of stator slots number of winding turns per phase number of machine pole pairs core loss copper loss power density eddy current loss hysteresis loss input power mutation probability nominal power output power rotational loss stator resistance cross-section area of wire period of one EMF cycle 338 IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

2 V LL V nom W s γ p η λ ɛ ρ c ρ r σ T t p ζ Ψ 1 Introduction line-to-line voltage nominal voltage slot width pole pitch motor efficiency diameter ratio fitness limit density of core material density of rotating part electric conductivity of wire permanent-magnet skew angle hysteresis coefficient flux linkage coupled by per pole Recently, the permanent-magnet motors are gaining popularity to the researchers because of their high efficiency, high-power density, high torque-to-inertia ratio and robustness [1]. The invent of high-energy permanent magnets, semiconductor technology, modern control algorithms and digital signal processors prices enable researchers to use permanent-magnet motors from domestic to high-performance industrial drive applications [2]. Permanent-magnet motors come in different geometries, among which is a disc-type or axial-flux permanent-magnet (AFPM) motor available in various configurations [3 7]. The AFPM motor s high torque-to-volume ratio, excellent efficiency and flat structure are especially suited to military and transport applications, and motivate researchers to develop new approaches for designing AFPM machines [8, 9]. The AFPM machines can be single- or double-sided, with or without armature slots/core, have internal/external permanent-magnet rotors, contain a surface-mounted or interior permanent magnet, and are single- or multi-staged [10]. The AFPM motor cogging torque is normally much higher as compared with the conventional motors [11]; however, they can still potentially be applied to high-torque applications such as ship propulsion or elevator direct drive [12, 13]. The double-sided AFPM motor type is the most promising and widely used as it needs less permanent magnets and windings. Topologies for double-sided AFPM machines are: one-stator-two-rotor, which is the type of TORUS structure, and two-stator-one-rotor, which is called axial-flux interior rotor (AFIR) [14]; whereas either of the two arrangements (external stator or external rotor) is practical. The external-stator arrangement uses fewer permanent magnets but at the expense of winding. However, the external-rotor arrangement is considered especially advantageous where the space is limited, mechanical robustness is required and torque-to-volume ratio is crucial [15]. The double-sided slotted TORUS AFPM motors are the most frequently applied among the other configurations, as they are mechanically stronger and have higher power density than the other configurations [16]. So, the slotted TORUS AFPM motor is used here for modelling and simulation. The genetic algorithm (GA) and finite element analysis (FEA) are used in the design process so that the machine s power density is maximised, the cogging torque is reduced and the undesired back-electromagnetic force (back-emf) harmonics are eliminated. Thus, the operational performance of the initial design is enhanced. Huang et al. [17] derived the general sizing and the power density equations for radial-flux permanent-magnet machines, which was a systematic method comparing the capabilities of various machine topologies. In another work, they developed the sizing equation for AFPM machines but did not present the machines optimised size [18]. A general optimisation process for an AFPM machine is possible with shape modification, via geometrical parameters, deterministic methods or soft computing methods. Aydin et al. [19, 20] has developed optimum-sized AFPM machines for both TORUS and AFIR topologies, but only two parameters (diameter ratio and air-gap flux density) were considered as optimisation variables, and the optimisation was done through the shape modification. In all shape-modification methods, there are trade-offs among the performance parameters, and the methods are not applicable to the multi-objective optimisation problems. In some studies the optimised value of 1/ for the ratio (λ) of inner diameter to the outer diameter was chosen in order to maximise the output power in AFPM machines [21, 22]. In [23], a method to reduce the free design parameters, in order to make a simple parametric study and to obtain an improved design for an AFPM machine equipped with concentrated winding was proposed and the most relevant figures of merit were theoretically analysed by means of some parametric analysis. In [24], Rostami et al. provided a design based on GA method for variable speed AFPM synchronous generator considering practical limitations. However, the design methodology is not clear and the analysis of the designed machine is very limited as they did not provide any results on power density, efficiency and total harmonic distortion (THD) of the induced voltage. Moreover, the prototype machine was not built to verify the design. Therefore in this paper a clear design methodology based on GA and FEA is developed for the double-sided AFPM synchronous machine. The GA is used as an optimisation tool to minimise the machine size considering practical limitations and various parameters such as winding turns, winding coefficient, electrical loading, air-gap length, diameter ratio, air-gap flux density, stator-slot number and permanent-magnet skew. The design objectives of the machine are maximum power density, minimum THD of the sinusoidal back-emf waveform and low cogging torque. In order to verify the design, a prototype 1 kw, three-phase, 50 Hz, four-pole synchronous machine is built. It is found from the experimental results that the optimised prototype machine exhibits low cogging torque, high-power density, low THD of the induced voltage which verifies the analytical results based on FEA. 2 Specific GA-based design optimisation This section presents the key elements of GA-based design optimisation incorporating practical limitations and the optimised dimensions of the machine. 2.1 Design restrictions and requirements Optimally, a design would include maximum power density incorporated with desired sinusoidal back-emf and would be maintained within design restrictions and requirements. Table 1 lists all the practical limitations and requirements for the design. The limitations are based on the typical 1 kw AFPM synchronous machine applied to scooter or any low-power application with similar rating, considering the materials available in the lab. IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

3 Table 1 Design restrictions and requirements Dimensional constraints machine outer diameter (D o ) D o 300 mm inner to outer diameter ratio (λ) 0.4 λ 0.75 effective axial-length of the motor (L e ) L e 500 mm air-gap length (g) 0.5mm g 2.5mm Material limitations stator and rotor core flux density B cs, B cr B max = 1.5 T (B cs, B cr ) permanent remanence 1.3 T Requirements rated line-to-line voltage (rms) V LL 100 V input phase current (rms) I rms 20 A air-gap flux density (B g ) 0.35 T B g 0.95 T electrical loading (A) 1000 A nominal power (P nom ) P nom = 1kW pole pairs ( p) p=2 motor efficiency (η) η 80% Frequency ( f ) f = 50 Hz number of phases (m) m=3 It is to be mentioned that in high-speed, high-torque and low-supply-voltage applications, sine-wave machines offer many advantages [25]. Mostly for small machines, the number of poles is limited because of the reduced space available for the windings. Nevertheless, the most restricting limitation for the number of poles is the motor operation speed. If the speed is high, a large number of poles will bring an increase in the frequency, which directly leads to higher stator core losses and higher convertor losses. In addition, the cost of permanent magnets increases. Therefore the final decision is made in favour of the four-pole machine, with the frequency limitedto50hz. 2.2 Key sizing equation The main dimensions of each electrical machine are determined by the electrical-machine-output power equation. Assuming negligible leakage inductance and resistance, the machine output power is expressed as [17] P out = h m T T 0 e(t) i(t) dt = mk p he pk I pk (1) where e(t) is the phase air-gap EMF, i(t) is the phase current, η is the machine efficiency, m is the number of machine phases and T is the period of one EMF cycle. E pk and I pk are peaks of phase air-gap EMF and current, respectively. K p is the electrical power waveform factor. A general-purpose sizing equation for AFPM machines is then extracted as [18] P out = 1 m p 1 + K f m 1 2 K ek i K p K L hb g A f ( p 1 ) l2 ( 1 + l ) D 2 2 ol e (2) where L e is the machine s effective axial length; K φ is the electrical loading ratio on rotor and stator; K i is the current-waveform factor (the ratio between the peak value and the root-mean-square (rms) value); K L is the aspect ratio coefficient with respect to a specific machine structure and keeping in view the consequence of loss, temperature rise and design efficiency requirements. m 1 is the number of phases of each stator; B g is the flux density in air-gap; f is the converter frequency; p is the machine pole pairs; λ is the AFPM machine s diameter ratio D i /D o ; D o is the machine s outer surface diameter; D i is the machine s inner surface diameter. A and K e are the total electrical loading and EMF factor, respectively A = 2m 1 N t I rms pd g (3) K e = E pk f a p C N t is the number of winding turns per phase; I rms is the rms value of phase current; D g is the average diameter of machine air-gap surface; Ψ is the flux linkage coupled by per pole; f α = np/60 is the angle frequency; n is the synchronous speed of the machine. The machine power density for total volume is defined as P den = (4) P out (p/4)d 2 totl tot (5) where D tot and L tot are the machine s total outer diameter and total length respectively, including the stack s outer diameter and end-winding protrusion from radial and axial iron stacks. 2.3 Real-coded genetic algorithm (RCGA) The GA includes operations such as reproduction, crossover and mutations. Reproduction is a process in which a new generation of population is formed by selecting the fitness individuals in the current population. Crossover is the most dominant operator in GA. It is responsible for producing new offspring by selecting two strings and exchanging portions of their structures. The new offspring may replace the weaker individuals in the population. Mutation is a local operator, which is applied with a very low probability. Its function is to alter the value of random position in a string. The RCGA is illustrated in Fig. 1; chromosome representation, crossover and mutation operators are described as in the following sections Chromosome representation: Fig. 1a illustrates each chromosome s 1 6-array for the proposed GA, while B g, λ, g, A, K w and N ph are air-gap flux density, inner to outer diameter ratio, air-gap length, electrical loading, winding coefficient and winding turns in each phase, respectively. Every generation has a chromosome population of 1400 and gets randomly selected from the first generation. It is to be noted that, the presented chromosome contains the genes g, K w and N ph, although they are not directly appeared in (3). The fitness function depends on all the mentioned genes implicitly as the ratios K e, K i, K p, and parameters D tot, L tot are functions of the genes. The authors only present the main results as the details are provided in [18]. Chromosome variables or genes have real values, and hence real coding is applied for normalising each gene as shown in Fig. 1b. Linear normalisation results from ( ) G normal = G G max G i G min (6) min where G i is the chromosome gene value varying between 340 IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

4 is randomly created and the chromosomes are normalised. Chromosomes evolve from generation to generation via successive iterations; a new generation is made by selection according to fitness value, parents and some offspring, whereas others are rejected to limit the population size. Half the genes from previous steps are omitted, and a new generation is created by performing crossover and mutation on selected genes. From every two selected genes, two children are created, replacing omitted genes, thus creating a new generation with an equal population as before (1400). The stopping criterion is then verified; upon validation, the algorithm stops and the final genes are selected, otherwise, new chromosomes or offspring are produced. The new generation undergoes all previous steps, and after several generations, the algorithm ends when the stopping criterion is fulfilled. Finally, appropriately selected genes optimise the motor dimensions or offer close to optimal dimensions with the highest power density. 2.4 GA-based computed results Fig. 1 Real-coding GA process a Chromosome representation (1 6 array) b Real-gene coding (linear normalisation) c Two-point crossover G min and G max. The normalised values are limited between upper and lower limits 0.8 and 0.2, respectively Crossover: For the present research, the elitist method is used as a selection operator for two-point crossover (Fig. 1c). Two random numbers between 1 and chromosome length 1 are first generated (1 random number chromosome length 1). Each chromosome is cut from the specified points in Fig. 1c, and the equivalent sections are then exchanged Mutation: In this research, mutation is executed with a probability P m (0.005 P m 0.05) and the outcome needs to be a valid chromosome. In real coding, for instance, genes are randomly chosen such that a random value is selected from the interval mentioned, after which it is added to, or removed from, the gene pool. Table 1 lists all genes permitted optimisation variations. A key issue in GA programming is the selection of a fitness function for obtaining the best solution to a problem [26]. An inappropriate fitness function may lead to the wrong answer. Another potential problem may arise when the produced genes are relatively better than other genes [27], and the answer may lead towards a local solution. The AFPM machine power density ((3)) is chosen as the fitness function and is calculated for each step and chromosome Flowchart: GA starts with a population which is the initial set of random solutions. The population consists of chromosomes that are string-structured concatenated lists of digits which code the problem s control parameters. In this paper, a 1400-string population For a three-phase, two-pole-pair AFPM motor, the potential number of stator slots is assumed to be 9, 12, 15, 18, 21 and 24; the GA program is then executed based on these stator-slot numbers. The present algorithm stops when the fitness function value ((3)) for the best current-population point is less than, or equals, the fitness limit (G n+1 G n ɛ). An AFPM machine may have any even number of permanent-magnet poles (2p) and any number of stator slots (N s ). From this infinite set, only a few permanent-magnet pole and stator-slot count combinations can maximize stator-slot utilisation and result in efficient production of torque. The number of stator slots in each pole, per pole pair, for 9, 15, 18 and 21 stator-slot counts, is fractional. The fractional slot-pitch winding configuration is more complicated than full slot-pitch, but all values are considered important because they reduce current and voltage harmonics, and also cogging torque. Table 2 lists various motor design parameters, with various stator-slot numbers optimised through GA optimisation. The optimised winding configurations for different stator-slot machines are also found using FEA-based simulation [28]. As a sample, the winding configurations for 15-stator-slot machine are shown in Table 3. As a sample, Fig. 2 shows the MATLAB-programming fitness function variations for 120 generation (which are not fully optimised) used for optimising the various stator-slot counts. 3 Finite element analysis GA facilitates getting the maximum power density, so dimensions obtained via GA are considered raw data, thus further analysis is needed for sufficiently mature final design. Three-dimensional (3D)-FEA is employed for analysing the double-sided TORUS AFPM motor s magnetic circuit and power density evaluation, providing an overall picture of different parts of the proposed motors saturation levels and extracting their characteristics. The AFPM motors have a unique construction; its lack of symmetry makes 3D-FEA a design requisite. An advantage of 3D-FEA is that various components of flux density can be calculated highly accurately. The design was simulated using commercial Vector Field Opera D software [29]. IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

5 Table 2 No. of slots Dimensions of the motor, with highest power density obtained via GA for different number of stator-slot counts P den, D o,mm N ph, turns A, A/m g, mm L pm,mm L cs,mm L cr,mm B g,t λ D s,mm W/cm Table 3 Stator winding optimised configuration for 15-stator-slot machine Slot No up + A B +C +C A +B C +A +A B +C A +B +B C down B B +C A +B C C +A B +C A A +B C +A Phase A Phase B Phase C in out in out in out Usually, permanent-magnet skewing is beneficial for reducing the cogging torque in electric machinery. It also eliminates some undesired harmonics reducing the back-emf THD. It should be noted that the back-emf amplitude is also reduced slightly with skewing. Skewing angle should be less or equal slot pitch. Through GA analysis, motor dimensions are obtained for each stator-slot count. The FEA then provides the THD of back-emf at various skew angles for the design candidates presented in Table 2. Fig. 3 shows THD variation against permanent-magnet skew angles. Minimum THD is clear to see for the motor with 15-stator slots and 9 permanent-magnet skew. So, the optimum selected chromosome is the one that represents a motor with 15-stator slots per pole pair in Table 2. Itistobe mentioned that the adopted fractional winding q=5/4 (slots per pole per phase) considered for back-emf waveform analysis includes the phase coils of one entire stator side, suitably series connected. 4 Final design and motor construction The best motor design dimensions are selected based on the proposed candidates from GA and FEA simulation. However, the final optimised design is made possible by minute changes effectuated by the powerful FEA with the strenuous task of changing permanent-magnet thickness, air-gap length and length of stator yoke and rotor yoke several times. Table 4 lists the machine design s final dimensions and specifications. It is to be noted that the outer bearing option is chosen as it provides better balance operation as compared with conventional inner bearing option. Fig. 4 shows the snapshot of rotor and stator of the prototype AFPM synchronous motor. The design challenge in manufacturing the AFPM motor is maintaining the air-gap between stator and rotor. Electromagnetic interaction between the rotor permanent magnet and the stator slots is quite large. (1000 N simulated value for this motor). The Fig. 2 Fitness function variation during GA optimisation 342 IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

6 Fig. 3 Back-EMF THD variation against permanent-magnet skew angles for machines with different number of slots air-gap needs to be as small as 1 mm. Fig. 4a shows the single disc of rotor with 9 skewed permanent magnets. The machine rotors were constructed by using mild-steel. In each rotor disc, four axially magnetided Nd Fe B permanent magnets are mounted on the disc surface facing the stator. The permanent magnets used in the machine prototype have 1.3 T remanence and 900 ka/m demagnetising field. Table 4 Motor s final design dimensions and specifications nominal voltage (rms line-to-line) V nom 90 V nominal power P nom 1kW number of poles 2 p 4 number of phases m 3 drive frequency f 50 Hz efficiency η 90.5% outer diameter D o 170 mm inner diameter D i 80 mm inner to outer diameter s ratio λ 0.47 magnet s axial length L pm 2.5 mm pole pitch γ p 118 permanent-magnet skew angle t p 9 stator-yoke thickness 2 L cs 30 mm rotor-yoke thickness L cr 11 mm slot width W s 10 mm slot depth 2 D s 16 mm number of stator slots 2 N s 30 number of winding turns per phase N ph 2 (15 18)/3 air-gap flux density B g 0.47 T air-gap length g 1mm Windings are hand-made professionally as shown in Fig. 4b. They are placed on slotted-stator surface with star connection. The fractional slot-pitch winding configuration for the 15-stator-slot counts (as shown in Table 3) is implemented. To prevent the windings from missing their position and from vibration during motor operation and to increase the insulation capability of the winding, a type of resin is applied giving the windings characteristics such as stiffness in working temperature, original dimensions and good thermal conductivity for heat release. An axial-flux motor s stator is theoretically either laminated spirally or axially. The spiral lamination is well known; however, the axial lamination of the stator, creating the slots, and to maintain the stator mechanically integral is too difficult. In this paper, spiral lamination silicon steel paper with thickness of 0.5 mm is utilised, which is quite fair as the supply frequency is 50 Hz. It is worth mention that the thinner paper (e.g. 0.1 mm sheet) may lead to poor stacking factor. Fig. 5 is a snapshot of the prototype 1 kw, three-phase, 50 Hz, four-pole AFPM synchronous machine. Fig. 4 Snapshot of stator and rotor of the prototype AFPM synchronous motor a Rotor disc with 9 skewed permanent magnets mounted b Winding configuration for the 15-stator-slot counts Fig. 5 machine Snapshot of the prototype 1 kw AFPM synchronous IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

7 5 Results and discussion 5.1 Simulation results The optimised AFPM synchronous machine is extensively simulated using FEA software. Sample simulation results are presented below. One of the objectives of this work is to design the AFPM motor with sinusoidal back-emf waveform; in other words, the back-emfs should be as sinusoidal as possible. Fig. 6a shows the three-phase back-emfs at rated speed (1500 rpm) for 15-stator-slot AFPM synchronous machine for both with and without permanent-magnet skewing; also FEA-calculated maximum and rms value of back-emf are displayed. The adoption of the fractional winding (q =5/ 4) implies a beneficial filtering effect on the back-emf waveform and avoid high distortion. This fact is confirmed in Fig. 6b (Fourier transform analysis of the back-emf waveforms) by the amplitudes of fifth and seventh harmonics which are rather low, the most important harmonics well known as the teeth harmonics of a q=1 winding. It is also found that the THD drastically decreases from 8.1 to 2.5% with 9 optimised permanent-magnet skewing for 15-stator-slot AFPM synchronous machine. Fig. 7 shows the FEA simulation based torque characteristics of the proposed 15-stator-slot machine with and without PM skewing. It is also found that the cogging torque is significantly reduced with 9 permanent-magnet skewing, which is shown in Fig. 7a. Thus, the torque ripple of the designed machine is reduced, which satisfies the design criteria. The 9 skew may not be the optimal skew angle for cogging torque and it may be further reduced using a different permanent-magnet skew angle or other techniques but the THD of the back-emf will not be maintained minimum. Fig. 7b shows the comparison of the speed torque characteristic for the motor designs with and without permanent-magnet skewing. It shows that skewing design also decreases the output torque. It is found that in various speeds the amount of torque for skew design is lower than that of without skewing; however, the torque difference is insignificant near rated conditions. Fig. 6 Voltage and its harmonic spectrum obtained from FEA for the proposed 15-stator-slot machine, with and without, PM skewing a Three-phase back-emfs at 1500 rpm, with and without, PM skewing b Back-EMF harmonic components, with and without, PM skewing 344 IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

8 Fig. 7 Torque characteristics obtained from FEA for the proposed 15-stator-slot machine, with and without, PM skewing a Cogging torque, with and without, PM skewing b Speed torque characteristic, with and without, PM skewing 5.2 Experimental results In order to test the performance of the prototype motor, an experimental setup is built. The hardware schematic for the experimental setup is shown in Fig. 8. Since the speed is not very high, in-line torque transducer with suspended installation and single-element coupling to create shorter drive train are used. Hysteresis brake on the motor shaft provides the desired load torque. The DC machine is used as the prime mover for the prototype motor during open-circuit test and permanent-magnet braking torque measurement. It should be noted that during cruising-speed test, secondary measurement such as temperature rise in the motor s critical sections are also monitored. Temperature of the motor during operation is found within acceptable range. The back-emf and cogging torque are considered as the main performance parameters to obtain. Fig. 9 shows the voltage and its harmonic spectrum obtained experimentally in open-circuit test. It is seen from the figure that the THD of the back-emf is only 2.6% which is verifying the simulation results. Further, three-phase back-emfs ensure the balanced operation of the machine. Fig. 8 Hardware schematic for experimental tests of prototype 15-stator-slot synchronous AFPM machine IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

9 Fig. 9 Voltage and its harmonic spectrum obtained experimentally from open-circuit test a Three-phase back-emfs of the prototype machine b Back-EMF harmonic spectrum (THD is 2.6%) The comparisons between FEA-based computed and experimental results are tabulated in Table 5 for two different speed conditions (1500 and 750 rpm). It is seen that the experimental results almost agree with those of the FEA-based computed results at different speed conditions. Fig. 10a shows the experimental cogging torque measured from open-circuit test conducted at rated speed (1500 rpm), which is in close agreement with the ones predicted via FEA simulation results. Fig. 10b shows the experimental speed torque characteristic for prototype AFPM motor. 5.3 Efficiency To accurately assess the machine efficiency, it is vital to calculate the losses. The machine efficiency is given by h = P out P out + P cu + P cor + P rot (7) where P out, P cu, P cor, P rot are output power, copper loss, core loss and rotational loss components, respectively. Core loss and copper loss are calculated from the equations below P cor = P h + P e, and P cu = R s I 2 (8) where P h and P e are the hysteresis and eddy current losses, respectively. Copper loss is responsible for most of the total losses. Stator resistance (R s ) depends on load and winding temperature which is calculated from [30] R s = 2N ( ) ph s l + l e (9) s T N ph p s cu σ T is the wire s electric conductivity at temperature T, N ph-s is Table 5 results Back-EMF comparison between experimental and FEA Back-EMF V max V rms THD% 1500 rpm experimental FEA-based computed rpm experimental FEA-based computed IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

10 Fig. 10 Experimental torque characteristics for the prototype AFPM machine a Cogging torque over one cycle b Speed torque characteristic the number of winding turns in series per phase, N ph-p is the number of winding turns in parallel per phase and s cu is the wire s cross-section. Thin parallel wires minimised the skin effect, therefore it is not considered in (7). l and l e are coil length and end-winding length, respectively. Hysteresis loss (P h ) and eddy current loss (P e ) comprise the motor core loss (P cor ) and can be calculated in terms of the Steinmetz equation as P h = k h B j max f r c, and P e = k e B 2 max f 2 r c (10) k h, k e, B max, and ρ c are hysteresis constant, eddy current constant, maximum flux density and core material density, respectively. ζ is the hysteresis coefficient which depend on the lamination material, thickness and conductivity. The power loss data of the 0.5 mm silicon steel paper used to fit the Steinmetz equation describes the specific loss in W/kg as P cor = B 1.8 maxf B 2 maxf 2 (11) As it is seen, open slots are used in the stator design because of simplicity and cheapness. They cause slot harmonics causing eddy current in this rotational speed and air-gap length. These losses are also taken into account to compute the efficiency. For a fine calculation of stator core losses, finite element-alternate current (FE-AC) analysis is done repeatedly for each space harmonic component (up to the 49th order) in combination with the current waveform s simulated time harmonic components, to obtain the laminated-stator eddy current losses. Rotational loss (including windage and friction losses) for analytical calculations is estimated from [31] P rot = 1 2 c ( f r r pn 3 )( D 5 o D 5 ) i (12) where c f is the friction coefficient, ρ r is the density of rotating part and n is the rotation speed (in rps). Fig. 11 shows the motor s efficiency for both predicted values via FEA simulation and experimental test which are in close agreement; however, the slight differences of simulation and experimental results are because of little deviation in coils winding and stator lamination during fabrication as these processes were hand-made by a professional. Fig. 11a shows the efficiency at various speeds for full-load and low-load (25% of full load) conditions. It shows the rise in efficiency with the increase in speed. Fig. 11b shows the efficiency against various loading conditions (from no-load to full-load) at rated speed. It is found that the motor maintains high efficiency even at low-load condition. The final optimal designed motor efficiency at full loading and rated condition reaches 90.5% at 1500 rpm. IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

11 Fig. 11 Motor s efficiency comparison between predicted values via FEA simulation and experimental test a Efficiency against speed for full-load and low-load conditions b Efficiency against various loading conditions at rated speed 6 Conclusion The design, simulation and real-time performance analysis of a prototype AFPM synchronous machine have been presented in this paper. The design optimisation using GA and FEA based on the machine sizing equation has also been discussed. The design objectives were maximum power density, minimum THD of the back-emf and low cogging torque for the machine. Based on the optimised design a prototype 1 kw, three-phase, 50 Hz, four-pole AFPM synchronous machine has been successfully built. The performance of the prototype machine has been tested and compared with the FEA based computed results in terms of THD of induced EMF, cogging torque, torque speed characteristic curve and efficiency. It is found that the prototype machine meets the design criteria. Therefore the proposed design technique could be utilised for designing any arbitrary capacity double-sided industrial AFPM synchronous machine. 7 Acknowledgment The authors thank the University of Malaya for the High Impact Research Grant No. D that funds the Hybrid Solar Energy Research Suitable for Rural Electrification. 8 Reference 1 Chen, J.L., Liu, T.H.: Implementation of a predictive controller for a sensorless interior permanent-magnet synchronous motor drive system, IET Electr. Power Appl., 2012, 6, (8), pp Jurca, F.N., Martis, C.: Theoretical and experimental analysis of a three-phase permanent magnet claw-pole synchronous generator, IET Electr. Power Appl., 2012, 6, (8), pp Vansompel, H., Sergeant, P., Dupre, I., Van den Bossche, A.: A combined wye-delta connection to increase the performance of axial-flux PM machines with concentrated windings, IEEE Trans. Energy Convers., 2012, 27, (2), pp Xu, W., Zhu, J., Zhang, Y., Guo, Y., Lei, G.: New axial laminated-structure flux-switching permanent magnet machine with 6/7 poles, IEEE Trans. Magn., 2011, 47, (10), pp Alberti, L., Fornasiero, E., Bianchi, N., Bolognani, S.: Rotor losses measurements in an axial flux permanent magnet machine, IEEE Trans. Energy Convers., 2011, 26, (2), pp Sharkh, S.M.A., Mohammad, M.T.N.: Axial field permanent magnet DC motor with powder iron armature, IEEE Trans. Energy Convers., 2007, 22, (3), pp Chan, T.F., Wang, W., Lai, L.L.: Performance of an axial-flux permanent magnet synchronous generator from 3-D finite-element analysis, IEEE Trans. Energy Convers., 2010, 25, (3), pp Vansompel, H., Sergeant, P., Dupre, L., Van den Bossche, A.: Evaluation of a simple lamination stacking method for the teeth of an axial flux permanent-magnet synchronous machine with concentrated stator windings, IEEE Trans. Magn., 2012, 48, (2), pp Caricchi, F., Maradei, F., De Donato, G., Capponi, F.G.: Axial-flux permanent-magnet generator for induction heating Gensets, IEEE Trans. Ind. Electron., 2010, 57, (1), pp IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013

12 10 Gieras, J.F., Wang, R., Kamper, M.J.: Axial flux permanent magnet brushless machines (Springer-Verlag, 2008, 2nd edn.) 11 Hwang, C.C., Li, P.L., Chuang, F.C., Liu, C.T., Huang, K.H.: Optimization for reduction of torque ripple in an axial flux permanent magnet machine, IEEE Trans. Magn., 2009, 45, (3), pp Di Stefano, R., Marignetti, F.: Electromagnetic analysis of axial-flux permanent magnet synchronous machines with fractional windings with experimental validation, IEEE Trans. Ind. Electron., 2012, 59, (6), pp Nguyen, T.D., Tseng, K.J., Zhang, S., Nguyen, H.T.: A novel axial flux permanent-magnet machine for flywheel energy storage system: design and analysis, IEEE Trans. Ind. Electron., 2011, 58, (9), pp Mahmoudi, A., Kahourzade, S., Rahim, N.A., Ping, H.W.: Improvement to performance of solid-rotor-ringed line-start axial-flux permanent-magnet motor, Prog. Electromagn. Res., 2012, 124, pp Choi, J.Y., Lee, S.H., Ko, K.J., Jang, S.M.: Improved analytical model for electromagnetic analysis of axial flux machines with double-sided permanent magnet rotor and coreless stator windings, IEEE Trans. Magn., 2011, 47, (10), pp De Donato, G., Capponi, F.G., Caricchi, F.: Fractional-slot concentrated-winding axial-flux permanent-magnet machine with core-wound coils, IEEE Trans. Ind. Appl., 2012, 48, (2), pp Huang, S., Luo, J., Leonardi, F., Lipo, T.A.: A general approach to sizing and power density equations for comparison of electrical machines, IEEE Trans. Ind. Appl., 1998, 34, (1), pp Huang, S., Luo, J., Leonardi, F., Lipo, T.A.: A comparison of power density for axial flux machines based on the general purpose sizing equation, IEEE Trans. Energy Convers., 1999, 14, (2), pp Aydin, M., Huang, S., Lipo, T.A.: Design and 3D electromagnetic field analysis of non-slotted and slotted TORUS type axial flux surface mounted permanent magnet disc machines. IEEE Int. Electric Machines and Drives Conf., January 2001, pp Aydin, M., Huang, S., Lipo, T.A.: Optimum design and 3D finite element analysis of non-slotted and slotted internal rotor type axial flux PM disc machines. IEEE Power Engineering Society Summer Meeting, July 2001, pp Upadhyay, P.R., Rajagopa, K.R.: FE analysis and computer-aided design of a Sandwiched axial-flux permanent magnet brushless DC motor, IEEE Trans. Magn., 2006, 42, (10), pp Chan, T.F., Lai, L.L.: An axial-flux permanent-magnet synchronous generator for a direct-coupled wind-turbine system, IEEE Trans. Energy Convers., 2007, 22, (1), pp Di Gerlando, A., Foglia, G., Iacchetti, M.F., Perini, R.: Axial flux PM machines with concentrated armature windings: design analysis and test validation of wind energy generators, IEEE Trans. Ind. Electron., 2011, 58, (9), pp Rostami, N., Feyzi, M.R., Pyrhonen, J., Parviainen, A., Behjat, V.: Genetic algorithm approach for improved design of a variable speed axial-flux permanent-magnet synchronous generator, IEEE Tran. Magn., 2012, 48, (12), pp Friedrich, G., Kant, M.: Choice of drives for electric vehicles: a comparison between two permanent magnet AC machines, IEE Proc. Electr. Power Appl., 1998, 145, (3), pp Rashtchi, V., Rahimpour, E., Fotoohabadi, H.: Parameter identification of transformer detailed model based on chaos optimisation algorithm, IET Electr. Power Appl., 2011, 5, (2), pp Acikbas, S., Soylemez, M.T.: Coasting point optimisation for mass rail transit lines using artificial neural networks and genetic algorithms, IET Electr. Power Appl., 2008, 2, (3), pp Hanselman, D.C.: Brushless permanent magnet motor design (New York: McGraw-Hill, 1994) 29 Opera Version 14.0 User Guide, Vector Fields, 2011, cobham.com 30 Saari, J.: Thermal analysis of high-speed induction machines. PhD thesis, Helsinki University Technology, Helsinki, Finland, January Wang, R.J., Kamper, M.J., Westhuizen, K.V.D.: Optimal design of a coreless stator axial flux permanent magnet generator, IEEE Trans. Magn., 2005, 41, (1), pp IET Electr. Power Appl., 2013, Vol. 7, Iss. 5, pp & The Institution of Engineering and Technology 2013