Non-Salient Brushless Synchronous Generator Main Exciter Design for More Electric Aircraft

: This paper presents a prototype of high speed brushless synchronous generators (BSG) design for the application in autonomous electric power generation systems (e.g., airplane power grid). Commonly used salient pole ﬁeld of the main generator part of BSG was replaced with a prototype non-salient pole ﬁeld. The main objective of the research is an investigation into the advantages and disadvantages of a cylindrical ﬁeld of the main generator part of BSG over the original salient pole ﬁeld. The design process of the prototype generator is presented with a focus on the electromagnetic and mechanical ﬁnite element method (FEM) analysis. The measurements of prototype and commercial BSG were conducted for the nominal speed of 8 krpm. The advantages and disadvantages of the proposed solution were established based on measurements in load and no-load conditions.


Introduction
Modern commercial planes are designed according to the concept of more electric aircraft (MEA) [1][2][3][4][5][6][7]. This concept states that the electric power system in future aircraft should replace pneumatic and hydraulic systems supplied from main turbofan engines. The electrical power system should provide the ability to control the aircraft via electromechanical and electrohydraulic actuators and also provide deicing protection and control the pressurization of the cabin. The high increase in electric power demand is the main result of such an approach. High electric power demand, in turn, requires much more powerful generators to supply it. In addition, an important role of the generator is also the ability to work as a starter to accelerate a turbofan engine [7][8][9][10]. The increase in power output of the generator should not affect the volume and weight of the generators in a significant way. One of the obvious directions is the increase in the rotational speed of the machine [11][12][13]. However, one of the consequences of an increase in rotational speed is a higher centrifugal force acting on the rotor parts and the limitation in the rotor and shaft diameters.
To meet the requirements of the MEA concept considers various types of machines with high power densities defined as the power to weight ratio [7,[11][12][13][14]. In addition to standard, wound field brushless synchronous generator (BSG), it is also proposed to use a permanent magnet (PM) [15,16], a switched reluctance (SR) [17][18][19] or an induction (IM) [8,20] machine as a starter-generator for aircraft. Despite the diversity of machine types proposed in research and prototype solutions, the BSG with a wound rotor is still the most often used in commercial aircraft. BSG has been used for decades, so its from the armature winding of the exciter through a 6-phase diode rectifier. The exciter armature is on the rotor along with the field of the main generator and the exciter field winding is on the stator. The exciter field winding is supplied from the subexciter via a controlled rectifier. The generator control unit (GCU) controls the voltage and the protection systems for the generator. In the variable high-speed operation of the generator, the excitation system is designed to ensure the voltage control (RMS value) requirements of the power grid. Figure 1. The brushless synchronous generator-based on a three-stage electrical machine topology: the subexciter-permanent magnet generator (PMG); the brushless exciter-synchronous machine with stationary field winding, rotating armature winding and rotating diode rectifier; the main generator-synchronous machine with rotating field winding; GCU-generator control unit.
In modern aircraft such as the Boeing 787 or the Airbus A350XWB the variable frequency generation system is used, where generators are directly driven from the turbofan engines. The system voltage is regulated at 230 V and the bus frequency varies from 350 to 800 Hz. Changing the way in MEA electric power generation requires adaptation of the distribution system and the use of more power electronic converters supplying loads [4]. This affects the power quality, which is also an important issue, as it has a direct effect on the reliability and efficiency of the power system [35]. The harmonic content of produced voltage and current of the main generator can also have a significant influence on the system power quality. In the salient pole machines, the effect of unsymmetrical pole shoe saturation during loaded conditions can increase the harmonic content in induced voltage [36].

Objectives and Scope
The main objective of the research is an investigation into the advantages and disadvantages of a cylindrical field of the main generator part of BSG over the original salient pole field. The main motivation for the research was the analysis of the possibility of introducing a cylindrical main filed construction without any detriment to the generated power quality in reference to the original construction. The proposed construction should allow for: • more sinusoidal back EMF of the generated power due to more sinusoidal excitation winding magneto-motive force (MMF) spatial distribution, • the possibility of reaching higher rotational velocities of modern MEA variable frequency power systems in future developments of BSG.
The development of a modern aircraft power system is based on MEA concept. However, only repetitively proven and reliable solutions can be used in commercial avionic applications.
The MOET FP6 [37] project investigated the component level power system development. The practical result of the investigation was an evolutional change in the power generation system components and particularly the BSG. The focus of those changes was the improvement of the generator structure for higher efficiency without any detriment to its relatability. That meant the introduction of a variable frequency system with a higher rotational velocity of the generator and improved power quality. The proposal in this paper solution is aiming to investigate the possible advantages in this regard of a cylindrical structure of the main exciter of the BSG. The literature on the subject of the construction of BSG main exciter is very limited and only investigates the possibility of using claw pole design [38] for the main exciter of the machine or the possibility of using a cylindrical structure for operation with higher rotational velocity [39].
The comparative study is conducted on a commercial and redesigned generator construction. Both machines are three stage machines consisting of a permanent magnet generator (PMG) subexciter, an inverted synchronous generator exciter and the main synchronous generator. The only difference between the two machines is the construction of the main generator field ( Figure 2). The commercially available machine has a salient pole main generator field construction and the proposed prototype has a cylindrical (non-salient pole) construction of the main generator field. The proposed prototype was designed as a cylindrical rotor machine in order to obtain sinusoidal distributed excitation winding MMF with the possibility of achieving adequate mechanical strength when working at a higher rotational speed. The reverse engineering approach was used on the commercial salient pole generator to determine the electromagnetic parameters of the machine. The generator performance in the steady state was measured and analyzed. The main factor for the analysis was the minimization of the high-order harmonic content of currents and voltages waveforms during the no-load and load operation.
The design of the prototype generator is based on the salient pole brushless synchronous generator type GT40PCz8 (S n = 40 kVA, U n = 208 V, p. f . = 0.8, n n = 8000 rpm) used in the Russian MI-28 helicopter and is a typical construction for avionic application. The commercial generator has a salient pole field of the main generator part and the proposed and constructed prototype uses a non-salient pole field ( Figure 2). The commercial GT40PCz8 generator was the subject of the reverse engineering process using analytical calculations and FEM simulations. This approach was used to determine the value of electromagnetic parameters such as current densities in the armature and field windings and magnetic flux density in various parts of the machine. Based on those values, the new main generator field was designed and developed.
The main contribution of the research is the design of the prototype of the main generator part cylindrical exciter that provides better power quality in no-load and under load conditions. This will allow for future developments for variable frequency power system higher maximum frequency.

Solution
The development and the implementation of a new design of the electrical machine requires a comprehensive approach to the physical phenomena taking part in the electromechanical energy conversion process in the drive system.
The first step of the design and analysis process is an electric machine design based on analytical calculations ( Figure 3). The dimensions of a machine are calculated based on the sizing equations approach [22,34]. The sizing equation describes the relationship between the output power of the machine (P) and its main dimensions design features, material parameters, and rotational speed: where: f s -the stator voltage frequency, p-the number of pole pairs; the machine main dimensions: D s -the inside stator diameter, l s -the length of stator core; the machine design features: K I -the current waveform factor, K P -the electrical power waveform factor, K E -the EMF (electromotive force or induced voltage) factor which incorporates the winding distribution factor and the air gap magnetic flux distribution, and the material parameters: A s -the stator electric loading, and B m -the air gap flux magnitude. The values of the design feature factors depend on the type of the machine, the type of the power supply (shape of the current waveform), the air gap flux distribution and the field and armature winding distributions [22]. The main generator stage designed process was conducted based on the sizing Equation (1) and the parameters presented in Table 1. The parameters for the design process were derived from the reverse engineering of the original generator. The design process assumed that the new exciter should allow for the higher rotational speed of the BSG. In addition, the excitation winding should generate a sinusoidal distribution of magnetomotive force (MMF) to generate a sinusoidal distribution of the air gap flux density. Part of the solution for the proposed assumption was to use a high number of rotor slots. The outer diameter of the original and new field structure of the main generator is D r = 173.8 mm, the airgap length is δ = 0.6 mm and the core length is l = 73 mm. The most important difference is that in a salient pole solution the air gap length varies, and, in a non-salient solution, it is constant. An additional advantage of the uniform air gap length is the decrease in non-uniform pole shoe saturation of the field during the under load operation of the BSG. The prototype field dimensions are partly based on the original generator construction and partly on the selected manufacturing technology, in particular, the technology of field winding construction and manufacturing. In the proposed manufacturing process, a square profile wire was selected for the field winding to maximize the fill factor. The minimum tooth thickness was calculated and then verified using FEM models for the machine to sustain the centrifugal forces and also due to the maximum value of the flux density in the field core. Figure 4 shows the dimensions of the main generator stator and both the commercial and the prototype generator rotors. The original salient pole excitation winding has 42 turns per pole and the design cylindrical prototype has 40 turns per pole. The cross section of the original and prototype excitation winding wire is the same and is equal to 4.5 mm 2 . The proposed design prototype apart from allowing higher rotational velocity of the machine should also decrease the high-order harmonic content of the induced EMF. Both the air gap flux density spatial distribution and the field and armature winding spatial distribution influence the harmonic content of the induced EMF. To analyze this influence, the air gap length distribution functions and field and armature winding distribution functions have to be defined. The air gap length distribution function is defined as: where α δ defines the average length of the air gap, α 0 and α 2k−1 are the relative components of the air gap length distribution function (α 0 = 1), φ s is the angular position along the stator and θ r is the displacement angle of the rotor axis in reference to the stator axis. In case of the prototype machine with cylindrical stator and rotor, this function is defined as: The armature winding distribution is defined based on the winding MMF distribution as: where x refers to phase a, b or c in three phase machine, i xs is the x'th phase armature current and the MMF xs (i xs , φ s ) is the x'th phase magnetomotive force distribution. This force is defined as: where N s represents the number of turns of equivalent sinusoidally distributed armature winding, A s,2k−1 are the relative amplitudes of armature winding MMF distribution (A s,1 = 1), θ xs is the angular displacement of x'th phase in reference to phase a (θ as = 0) and k MMF s is the number of odd harmonics used to approximate the armature MMF distribution. The same winding distribution and magnetomotive force distribution functions can be defined for field winding: where i f d is the field current, φ r is the angular position along the rotor, N f d represents the number of turns of the equivalent sinusoidally distributed field winding, A f d,2k−1 are the relative amplitudes of field winding MMF distribution (A f d,1 = 1) and k MMF f d is the number of odd harmonics used to approximate the field MMF distribution. Based on the winding distribution the MMF distribution and the air gap length distribution, one can calculate the mutual inductance between field winding and phase winding: If Equations (3), (4) and (7) are applied to relation (8), the function for mutual inductance is defined as: where: where µ 0 is the magnetic permeability of vacuum, l is the machine length and r is the distance from the axis of machine to the middle of the air gap. As can be noticed, the mutual inductance harmonic components depend on the air gap length and the armature and field winding distribution high-order harmonics. In the designed prototype, the armature winding was not changed. However, both the air gap length distribution and the field winding distribution were changed. The proposed solution for more sinusoidally distributed excitation winding MMF was to use a high number of rotor slots.
Because the prototype has a cylindrical construction and more sinusoidally distributed winding, then the commercial generator the resulting EMF is much more sinusoidal. Figure 5 shows the MMF distribution for both the commercial and the prototype main generator field windings. Simplified mechanical calculations have been performed to approximate forces and pressures acting on the prototype main generator field core. The force acting on the weakest part of the rotor was calculated as a sum of two forces: • The centrifugal force acting on rotor tooth-F tooth , • The centrifugal force acting on the winding in the rotor slot F Cu .
The centrifugal force acting on the rotor tooth was calculated as: where ω is the mechanical rotational velocity of the rotor, ρ Fe is the density of rotor core, l is the length of the main generator part of the machine, a is tooth thickness at the bottom of the tooth, b is the tooth thickness at the top of the tooth, c is the height of the tooth and r 1 is the bottom radius of the rotor tooth. The force acting on the copper in the slot was calculated as: where d is the width of the copper profile wire in the rotor slot. The resulting force acting on the rotor tooth from both the tooth itself and the copper in the slot is about 23 kN. The resulting pressure in the thinnest part of the rotor tooth is about 126 MPa. Because the resulting machine dimensions and the resistance of excitation winding did not change for the prototype and original commercial generator, the thermal analysis of the prototype was not conducted. However, if this solution was to be implemented in a commercial power system, such analysis would have to be conducted.
Three-dimensional (3D) geometric models of the machine are then developed as the parametric model using CAD software. This approach enables easy modification of the geometric models and studying of the influence of the geometrical parameters and material properties on the magnetic field distribution, power losses, integral parameters (inductances, torque), mechanical stresses, etc. (Figure 3).

FEM Simulation
The FEM simulations (using Cedrat/FLUX2D software (version 9.3, Cedrat, Grenoble, France)) have been carried out to verify the designed BSG. Due to a lack of information about the materials used in the commercial generator, the FEM simulations were verified using measurement data. As a result of reverse engineering processes for the given dimensions and measured behavior of the commercial machine, the material parameters of the stator and rotor magnetic core were established. Those parameters were calculated using the iteration process by minimizing the error between measurements of no-load state and FEM simulation results. The initial relative permeability µ r = 8000 and the magnetic polarization at saturation of 1.6 T parameters were set for the commercial machine stator and rotor core. For the prototype field core of the main generator, M530-50A steel sheets have been used. Unfortunately, the manufacturing process and relatively poor stacking factor caused the necessity for additional varnish coating insulation between steel sheets. During the laser cutting process of the prototype core steel sheets, a slag caused by laser cutter needed to be removed. This process also damaged the original varnish insulation coating on the surface of the steel. Because of this decrease in the stacking factor and the damage done to the cut edge of the electrical steel, the material magnetic characteristic has changed. The field core parameters for the simulation were set to µ r = 500 (initial relative permeability) and the magnetic polarization at saturation of 1.55 T.
The model was implemented in the Cedrat/FLUX2D software using a dedicated Python script. For the discretization of the FEM model, the air gap was divided into three sections: the stator static mesh section, the rotor moving mesh region and the air gap re-meshing region for the mesh generation during the rotor movement. The resulting mesh and geometry of both the commercial and prototype generator are shown in Figure 6. The resulting mesh generation process led to a total number of about 36 thousand nodes for the original salient pole main generator of BSG and about 53 thousand for the prototype. The simulation process was conducted with set maximum variation between the iterations of integral calculation of less than 0.1%. Table 2 shows the maximum values of magnetic field densities in the crucial parts of the commercial and designed machines. Table 2. Maximum values of flux density of the main generator part of BSG. FEM simulation results at no-load conditions for the commercial salient-pole and prototype cylindrical generators are shown on Figure 7. The magnetic equipotential lines are shown for both the commercial and the prototype BSG main generator part. In this simulation, the main generator field winding has been supplied with nominal no-load current and FEM computations in magnetostatic conditions have been conducted. The amplitude of the normal component of the air gap flux density of both types of the main generator field has the same maximum value of 0.8 T (Figure 8). The prototype cylindrical generator has got more sinusoidal distribution of air gap flux density (Figure 8). In addition to electromagnetic calculations, the mechanical calculations concerning the centrifugal forces have been conducted. The simulation of this force acting on one of the rotor teeth was analyzed. The stress analysis module of Autodesk Inventor software (version 2014, Autodesk, San Rafael, CA, USA) was used to perform the mechanical FEM simulation. The rotation speed of 16 krpm has been applied to the rotor. The materials used for stator (steel sheets), windings (copper), and slot wedges (bras) were defined. The stress analysis results are shown in Figure 9. In the simulation, only the copper inside of the slots is considered as the copper in the end windings will not be held by the rotor core itself. The electromagnetic and mechanical FEM simulations verified the assumptions and results of the analytical calculations. The proposed solution delivers more sinusoidal distribution of the excitation winding MMF and the maximum tensile stress due to centrifugal force does not exceed maximum tensile stress allowed for M530-50A steel.

Measurements
Measurements have been carried out for the no-load (open circuit) and load conditions of the prototype and commercial BSG. Measurements of the no-load back EMF in the function of field current for constant speed are shown in Figure 10. This shows that the prototype's main generator field requires more than two times the no-load nominal excitation current. The main factor in this is the smaterial and manufacturing technology used for the core of the main generator field. The difference in number of turns between the prototype and the original machine is only 5% and the air gap length is the same. This also means that the armature reaction of the prototype generator is diminished. The no-load back EMF waveform of the prototype has less high-order harmonic components than the commercial one ( Figure 11). In both machines, the stator winding is the same which means that the back EMF waveform is dependent on the air gap flux density spatial distribution which in case of a prototype generator is more sinusoidal.
For the load conditions measurements, a load impedance consisting of 0.8 Ω active part and 0.15 Ω reactive inductive part (@400 Hz) has been used. The measurement results of the voltage and current harmonic components in load conditions are shown in Figure 12. As can be observed, the designed prototype with non-salient pole field of the main generator produces less high-order harmonics in the back EMF waveform. In addition to no-load and under load steady state test, a transient state of the symmetric short circuit was performed ( Figure 13). This test was performed by shorting the armature terminals when the machine was operating at no-load condition with the nominal voltage at the terminals. As can be observed, the excitation current is higher in the prototype generator than in the commercial one and equal in steady state to the nominal no-load excitation current.

Conclusions
The comparative study of two types of BSG main generator exciter has been conducted. The salient pole and cylindrical exciter construction have been compered. The exciter and subexciter of BSG were not changed. The design process was composed of analytical calculation, electromagnetic FEM calculations, and mechanical FEM calculations. The mechanical calculations were conducted to analyze the centrifugal force impact on the designed field of the main generator at higher than nominal (16 krpm while nominal is 8 krpm) rotational velocity, whereas the electromagnetic FEM simulations have been conducted in order to verify the analytically calculated dimensions and parameters of the main generator exciter.
The BSG with the prototype non-salient field has a higher volume and mass of the magnetic core (cylindrical core-9.3 kg, salient pole core-8.1 kg). The volume and mass of the copper used for the field coils are slightly smaller in design and build a prototype (cylindrical field windings copper-2.7 kg, salient pole field windings copper-2.9 kg).
The measurements were conducted at a nominal rotational speed of 8 krpm due to the same exciter and subexciter in both generators. The results of the measurements show significant advantages of the designed solution. Because of more sinusoidal MMF distribution of the field winding, the no-load phase voltage contains less high-order harmonics than the voltage of the silent pole commercial BSG (GT40PCz8). The possibility of operation at higher rotational speeds and better quality of produced electrical energy are a good indicator for the development of the BSG for the MEA variable frequency power system.
The designed generator also has its flaws. It is slightly heavier than the original, and the cooling of the field winding will be different (lack of axial vents in the designed rotor). The selected technique of the manufacturing leaves much to be desired and will have to be modified to archive better magnetic permeability of the field core.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript: