*Article* **Determination of Serviceability Limits of a Turboshaft Engine by the Criterion of Blade Natural Frequency and Stall Margin**

#### **Yaroslav Dvirnyk 1,2, Dmytro Pavlenko 2,3 and Radoslaw Przysowa 4,5,***<sup>∗</sup>*


Received: 2 November 2019; Accepted: 4 December 2019; Published: 9 December 2019

**Abstract:** This paper analyzes the health and performance of the 12-stage axial compressor of the TV3-117VM/VMA turboshaft operated in a desert environment. The results of the dimensional control of 4800 worn blades are analyzed to model the wear process. Operational experience and two-phase flow simulations are used to assess the effectiveness of an inlet particle separator. Numerical modal analysis is performed to generate the Campbell diagram of the worn blades and identify resonant blade vibrations which can lead to high cycle fatigue (HCF): mode 7 engine order 30 in the first stage and mode 8 engine order 60 in the fourth. It is also shown that the gradual loss of the stall margin over time determines the serviceability limits of compressor blades. In particular, the chord wear of sixth-stage blades as high as 6.19 mm results in a reduction of the stall margin by 15–17% and a permanent stall at 770–790 flight hours. In addition, recommendations setting out go/no-go criteria are made to maintenance and repair organizations.

**Keywords:** gas-turbine performance; turboshaft; axial compressor; blade; FEM; CFD; erosion; wear; stall margin; compressor surge; brownout

#### **1. Introduction**

Long-term engine performance in adverse operating conditions is one of the main prerequisites for successful helicopter missions. Unlike airplanes, helicopters have to hover for a long time above the ground, often raising a cloud of dust and causing a brownout. Therefore, the ability to ensure long-term operations at elevated dust concentrations is one of their most important features. Studies on the impact of environmental particles on the efficiency of helicopter engines were already being carried out during the development of the first helicopter types [1].

The gas path of the helicopter engine operated in a dusty environment is exposed to contamination [2,3]. Deposition of particles on aerofoils [4,5] and the heat transfer surfaces of the air cooling system [6] can lead to the deterioration of the aerodynamic and thermodynamic properties of these components. Moreover, given that dust particles (no matter how small they are) have abrasive properties, they also cause erosive wear of the engine components and contribute to various types of structural damage [7–9]. Erosive wear of compressor blades leads to reduction in the stall margin of the compressor, an increase in the likelihood of fatigue damage of compressor blades due to changes in their natural frequencies of vibration, and a decrease in the efficiency of the engine due to the wear of the gas path.

Rotating components are sensitive to mechanical damage caused by solid particles [10]. Models describing their impact and related erosive damage have been proposed by many researchers, such as Finney [11], Bitter [12], and Sheldon [13], but predicting the actual aerofoil degradation is still difficult. The main factors affecting the magnitude of erosion consist of the angle of collision, velocity and particle size, blade surface properties, and particle concentration. Van der Walt [14] showed that the wear rate of the aerofoils is directly proportional to dust concentration. The mechanical properties of the material are also an important parameter influencing the mechanisms of erosion [15,16].

Eustyfeev [17] performed experimental and theoretical studies of compressor blades eroded by solids to determine the erosion resistance of the blade material EI-961. The limiting ratio of the particle size to their velocity of contact with the material was determined. However, the experiment was performed on a cut-out sample, not on a real blade, as well as on one type of material, that is EI-961 steel, while most blades of modern compressors are made of titanium alloy.

Batcho [18] and Singh [19] analyzed loss of performance caused both by erosion and deposition. The impact of ingested particles on engine operation depends on the physical and chemical properties of the dust, its composition and concentration. Particles deposited on compressor aerofoils change their geometry and roughness, which leads to a decrease in the efficiency, and consequently, reduces the pressure ratio and performance of the compressor [20].

The time between overhauls (TBO) of an engine operated in a highly dusty environment is much less than that set out by the manufacturer and is reduced by the erosion of the compressor blades. The statistical analysis [21] showed that engines with eroded compressors represent the largest group of engines grounded due to compressor blade damage (30–35%). This value is comparable to the ratio of engine removals due to foreign objects ingested from the runway during take-off (25–30%). The share of aviation incidents, bird ingestion, or human errors during maintenance accounts for 15–20%.

The aim of this study was to assess the serviceability limits of the compressor blades of a helicopter engine operating in a dusty environment. To achieve this goal, the following tasks based on the results of the dimensional control of worn blades were performed:


#### **2. Methods**

The TV3-117 turboshaft (Figure 1) with a maximum power of 1640 kW and overall pressure ratio of 9.4 belongs to one of the world's most popular helicopter engines. It rotates anticlockwise (looking from aft to fore) and has a 12-stage axial compressor with a subsonic inlet and variable inlet guide vanes. Close to the blade tip, the Mach number is M = 0.8–0.92. The tip radius is 163 mm. The number of blades in subsequent stages are as follows: 37, 43, 59, 67, 73, 81, and 89 in further stages.

This work aims to analyze and predict the erosive wear of the TV3-117 engines powering the Mi-8MTV and Mi-24 helicopters in the desert environment of the Republic of Algeria. The region of North Africa and the Middle East is known for extreme ambient temperatures, a high concentration of dust, and sand dispersing up to a height of 6000 m, which is the helicopter operation zone. For example, in North Sudan, dust concentration is 1.3 g/m3, and in Algeria 1.3–1.6 g/m3. The size of particles ingested into the gas path ranges from 0.01 to 2 mm.

The operation of gas-turbine engines in the regions described above, with a high content of dust and sand, inevitably leads to deterioration of the engine [22,23]. Airborne particles may cause substantial erosive wear of compressor blades. As a consequence, tip clearance is widened and the aerofoil profile is modified—in particular, the chord length and the blade thickness appear to decrease, especially above 66% of the span (Figure 2). This leads to a reduction in the compressor pressure ratio and ultimately a reduction in the efficiency of the entire engine.

Firstly, a statistical analysis was carried out to study the nature of compressor wear and to determine its critical components. On the basis of the subsequent regression analysis, patterns of wear of the blades of all compressor stages as a function of engine operation time and dust concentration were established.

The obtained patterns of the chord wear of blades of all compressor stages were used to predict their geometry, depending on the engine operating time. This allowed for the simulation of the flow through the worn compressor and modeling blade vibration, and a determination of the serviceability limit of the blades in terms of structural integrity and stall margin.

For the analyzed engines, the observed impact of erosion on compressor performance was several times greater than deposition due to high dust concentration and fair inlet protection. Moreover, compressor fouling can be reversed by washing. Therefore, particle deposition is not taken into account in this work.

**Figure 1.** TV3-117 turboshaft [24].

**Figure 2.** Erosive wear of compressor blades.

#### *2.1. Blade Inspection*

The inspected engines were grounded and torn down for one of two reasons: Either due to the chord wear of first-stage blades higher than 2 mm at the tip; or due to exceeding the threshold levels of performance parameters such as the gas generator speed *n*<sup>1</sup> or the turbine inlet temperature (TIT).

The dimensional control of blade geometry was performed by measuring the chord and the thickness of the profile at various sections as well as blade height. Ten blades were selected for the evaluation from each stage of 40 engines [25]. In total, 4800 blades were inspected. Statistica 12 was used to analyze the results.

The inspected engines were initially divided into two groups. The first one included those that were operated without a particle separator (IPS), and the second with an IPS.

#### *2.2. Structural Analysis*

Compressor blades of modern turboshafts are characterized by thin profiles and relatively low stiffness, hence forced vibration presents a notable threat to them [26,27]. Compressor blades absorb intensive static and dynamic loads. When rotating under large centrifugal forces, the blades are deformed, which leads among others to a decrease in their twist. As a result of unsteady flow forces, compressor blades vibrate relative to their static deformations. The distribution of dynamic stresses in blades has to be determined to ensure that they remain below the fatigue limit of the material [28,29].

The erosive wear of blades has a significant impact on their strength, which was confirmed by Hamed [1,30]. The properties of the blade surface and its ability to withstand erosion are one of the key factors that determine the reliability of the system, since a variety of surface irregularities can lead to stress concentration that increases the risk of high cycle fatigue (HCF) [31–34].

The main reason for forcing blade vibration is an irregularity of flow in the circumferential and radial direction. The frequencies of the driving forces are multiples of the rotor speed (engine orders) that are equal to the number of obstacles around the circumference, e.g., the number of guide vanes and struts in the gas path. When the speed changes, a number of resonant frequencies can be excited. To predict the frequencies of resonant vibration, a modal analysis of blades and the Campbell diagram are necessary [35,36].

Vibration frequencies and operational deflection shapes can be determined with a certain degree of accuracy by numerical methods, in particular using volumetric finite element models (FEM). At present, this research method is preferred, since the analytical method of calculating frequencies and vibration modes is not suitable for the complex geometry of aerofoils.

Blade geometry was measured using 3D scanning, followed by postprocessing in the CAD system ASCON KOMPAS 16.0. Geometry models of compressor blades were developed in Unigraphics NX8. Compressor blades operated in a dusty environment were described using parameterized solid-state models. By varying the blades height, chord length, and thickness in the corresponding sections of the aerofoil, several variants of blade geometry were generated.

The grid models of blades, created with the ANSYS ICEM CFD grid generator, consisted of 15–18 thousand hexagonal SOLID 185 elements. The natural frequencies of blades were calculated by the ANSYS 14.5 solver.

#### *2.3. CFD Model*

The 3D flow calculation, with an ideal gas as the working fluid, was based on the Navier–Stokes equations and the finite element method (FEM), implemented in the ANSYS CFX solver. The mesh was created in ANSYS Turbo Grid to model the operation of the compressor (Figure 3). A separate mesh flow model (domain) was designed for each stage (Figures 4 and 5). The domains were designed taking into account the possibility of air flow leaking in the radial gap (Figure 6) by adding the interface in the blade tip. The compressor model consisted of 26 domains (Table 1).

The following grid parameters were considered when building the grid:


First, the flow calculation was performed for two rotational speeds: 95% and 98%, where 1% corresponds to 195.37 rpm. An ambient air temperature of 288 K and pressure of 101.325 kPa were assumed. Each rotational speed of the compressor rotor corresponds to certain angles of the variable inlet guide vanes (VIGV) and guided vanes of the further four stages.


**Table 1.** Computational fluid dynamics (CFD) domain and mesh parameters.

**Figure 3.** CFD model of the compressor.

**Figure 4.** Mesh of the flow through the inlet and variable inlet guide vanes.

**Figure 6.** Mesh in the tip gap.

To reduce the required computing power, a single blade with cyclic symmetry along the lateral boundaries of the domain was modeled for each compressor stage. Stage interfaces (Mizing-Plane) between fixed and rotating domains were defined, which allowed for interpolation between interconnected grids, taking into account the laws of mass conservation.

The choice of the turbulence model depends on the nature of the turbulent flow, the required accuracy, the available time and computational resources. The SST K-*ω* turbulence model of Mentera was chosen as more accurate and reliable for the class of flows with a positive pressure gradient. The residual RMS error of <sup>1</sup> × <sup>10</sup>−<sup>6</sup> was assumed as a satisfactory condition of computational fluid dynamics (CFD) convergence and it was achieved after 600–870 iterations.

CFD results were used to estimate the compressor maps, as in another paper about this turboshaft [37]. The mass flow rate was measured at the outlet of the compressor. Each operating point corresponded to a certain mass flow, in the range from 4 to 11 kg/s.

#### *2.4. Modeling Two-Phase Flow through IPS*

The modeling of multiphase flows presents a number of difficulties compared to single-phase flows, since it is necessary to solve the equation of mass, amount of motion, and energy conservation for each phase separately. These equations are much more complicated than for the single-phase case, because they have additional terms that govern the exchange of mass and energy between phases. However, as a result of various concomitant physical phenomena and possible changes in the flow regime, the exact values of many parameters are not always known [6].

To study the operation of the inlet particle separator, we describe multiphase dispersion flows, in which there is a continuous and dispersion phase. The dispersed phases contained many particles distributed in a continuous phase. The Euler model and the ANSYS CFX software were used for modeling. The equations of mass, amount of motion, and energy conservation were solved separately for each phase. In the equations of motion, the interfacial drag force and other forces observed in multiphase dispersed systems were adopted. The calculations determined the local flow rate, temperature, and volume fraction of the dispersed phase. A granulation model was used to account for particle collision, friction, and density of the particles.

To simulate two-phase flow and obtain results on the velocity of motion and distribution of dust particles in the air, the following assumptions were made:


The IPS geometry, mesh, and other details are described in our previous publication [38]. Although particle separation is not the subject of this work, the selected calculation results are presented here to explain the wear process in engines operated with or without the IPS.

#### **3. Results and Discussion**

#### *3.1. Chord Wear*

The results of the dimensional control of the compressor blades show that the largest wear is observed at the blade tip, while a much smaller wear is exhibited near the root (Figure 7). The chord wear of Sections 2–4 can be expressed as a function of the wear of the tip section.

**Figure 7.** Chord wear of the compressor blade in specific cross-sections.

The statistical analysis carried out for individual stages revealed the wear patterns caused by the design features of the compressor [25]. Figure 8 shows the chord wear of the compressor blades for all stages. For engines operated without IPS, the largest wear is observed for the first-stage blades, while for engines operated with IPS, for the sixth-stage blades.

Engines belonging to the first category, operated without IPS, were grounded in accordance with the current standards for the maximum allowable chord wear of first-stage blades i.e., 2 mm at the tip. The effective TBO of the first category engines was only 150–200 FH, provided that it is nominally 1500 FH for this type of engine.

Engines from the second category, operated with IPS, were removed from service due to their operating parameters, such as *n*<sup>1</sup> speed and turbine inlet temperature beyond permissible limits. The effective TBO for the second category was 600–650 FH which indicates a partial effectiveness of the applied IPS [38].

Given the close correlation (R > 0.856) of chord wear among stages 2–12, it is possible to use these regression dependencies for modeling compressor wear. The chord wear of the sixth-stage blades is used as the independent variable. This choice is explained by the fact that, despite the obvious advantages of using the first stage for the dimensional control of blades, which is the most convenient in terms of monitoring and predicting the remaining useful life (RUL) without removing and disassembling the engine, it does not correlate well with the wear of the remaining stages (R < 0.4).

The chord wear of stages 2–12 correlates with the wear of the sixth stage and can be estimated using the linear regression.

$$c\_i = a\_i c\_\theta + b\_i \tag{1}$$

The dependence of the chord wear on the compressor blades on the engine operating time in a dusty environment can be described by a second-order curve (Figure 9).

$$\mathcal{L}\_6 = \left(4.69 \times 10^{-6} \, t + 1.25 \times 10^{-3}\right) \, t \,\delta,\tag{2}$$

where *δ* is dust concentration in g/m3. For the analyzed fleet, the average concentration *δ* =1.6 g/m3 can be assumed:

$$
\sigma\_6 = 7.5 \times 10^{-6} t^2 + 0.02t. \tag{3}
$$

**Figure 8.** Average chord wear of compressor stages for an engine operated without inlet particle separator (IPS) or with IPS.

#### *3.2. Modal Analysis of Blades*

To assess the effect of wear on the natural frequency of the vibration, a modal analysis was performed (Figure 10). For each stage, an individual Campbell diagram was developed describing the nominal and worn profile of the aerofoil. Erosion significantly increases the natural frequencies of the blades. The analysis reveals several resonances excited by engine orders. Some stages were identified with a risk of resonance in the operating speed range (Figure 11).

**Figure 9.** Chord wear of sixth-stage blades in function of flight hours.

**Figure 10.** Modal and kinetostatic stresses of the first-stage blade with a nominal (**left**) and worn aerofoil (**right**) for *n*<sup>1</sup> = 100%.

**Figure 11.** Impact of erosion on the Campbell diagram for critical modes and stages.

Stages and modes for which the risk of resonance is identified:


#### *3.3. Particle Separator*

A particle separator cleans the air entering the engine from dust, sand, dry twigs, leaves, and other foreign objects during taxiing, take-off, and landing from unpaved runways or landing areas [39]. When IPS is turned on, with the engine running, hot air from the compressor enters the dust ejector nozzle. At the same time, as a result of centrifugal forces, part of the airflow entering the engine is pressed to the rear of the central fairing and enters the separator inlet. Most of the cleaned air passes through a separator to the engine inlet. Contaminated air, including foreign particles, enters the dust exhaust pipe, in which a vacuum is created by the ejector. Thus, particles are expelled into the atmosphere.

The results of modeling the speed and trajectory of foreign particles of various sizes flowing in the gas path showed that most of the particles are separated along the internal curved surface of the IPS due to the action of centrifugal forces. However, experience from the operation of helicopters in a dusty environment shows that the use of a particle separator of this type does not solve the problem of erosive wear of compressor aerofoils [38].

IPS is effective at separating large fractions of particles ranging from 50 to 100 μm, almost completely cleaning the air supplied to the engine compressor (Figure 12). If the particle size is relatively small—from 10 to 50 μm, about 20% of the particles enter the engine compressor bypassing the IPS separator under the influence of viscous forces following the flow. (Figure 13).

**Figure 12.** Velocity and concentration of large particles in the inlet of the turboshaft with IPS; particle size 50–100 μm.

#### *3.4. Stall Margin Analysis*

To assess the effect of erosion on the compressor performance, the flow for the nominal (initial) geometry of the blades, as well as the geometry corresponding to the engine operating time of 200, 400, 600, and 800 FH was simulated. Compressor pressure maps (Figure 14) and efficiency maps (Figure 15) were calculated and validated by comparison with available test data from a compressor rig and test cell.

**Figure 13.** Velocity and concentration of small particles in the inlet of the turboshaft with IPS; particle size 10–50 μm.

The compressor maps plotted as a function of flight hours show that with increasing intensity of blade wear, the pressure ratio and its efficiency decreases. Consequently, the stall margin (SM) is also reduced (Figure 14). It was calculated using the following formula:

$$\text{SM} = 100\% \left( \frac{\pi\_{\text{stall}} / \dot{m}\_{\text{stall}}}{\pi\_{\text{ss}} / \dot{m}\_{\text{ss}}} - 1 \right), \tag{4}$$

where *πstall* is the pressure ratio for the stall line, and *πss* is the pressure ratio for the steady state line.

A decrease in the stall margin of the compressor by 15% causes the appearance of a surge during test-cell testing of TV3-117 engines. The analysis of flow through a compressor with blades of varying degrees of wear (Figures 16 and 17) showed that as a result of erosive wear of rotor stages 6–9, the compressor develops a stall in the high speed range, which leads to a surge. The reason for this phenomenon was chord reduction and an increase in the radial clearance. For this reason, the compressor operated in a dusty environment reaches its serviceability limit after 730–750 FH (Figure 18).

**Figure 14.** Dependence of pressure compressor maps on the operating time in a dusty environment.

**Figure 15.** Compressor efficiency versus operating time in a dusty environment.

**Figure 16.** Flow velocity in the new and worn compressor.

**Figure 17.** Flow velocity at 90% blade span.

**Figure 18.** Stall margin of the compressor versus operating time in a dusty environment.

#### **4. Conclusions**

On the basis of the analysis of the geometry of the compressor blades of the TV3-117 engines operated in a dusty environment, it was found that the wear of the blades in all compressor stages occurs uniformly. When the particle separator is used, the largest chord wear is observed for stages 1–6. As a result of a close correlation of chord wear in stages 2–12, their wear can be related to the sixth stage.

An original methodology was developed to assess the influence of erosive wear of blades on compressor performance by modeling three-dimensional flow for various degrees of wear of all rotor stages. It involves measuring the chord wear of blades, calculating the natural frequencies of vibration whilst taking into account the aerofoil wear, numerical calculation of the compressor flow, and analyzing the onset of stall. This approach in combination with established patterns of chord wear over engine operating time allowed for the assessment of the serviceability limits of the blades.

On the basis of a modal analysis of compressor blades with different operating times, the dependency of blade vibration frequency on the chord wear for all stages was established. It was found that when the chord of the first-stage blades has worn more than 4 mm, M7 EO30 resonance can occur. Similarly, chord wear of the forth-stage blades higher than 5.1 mm causes M8 EO60 vibration. The HCF risk for the remaining stages is negligible.

The values of the maximum permissible chord wear of the blades of all stages are determined by the criterion of stall margin of the TV3-117 turboshaft. The chord wear of the sixth-stage blades of 6.19 mm is critical because it is accompanied by a decrease in the stall margin of the compressor by 15–17%, which indicates the appearance of a permanent stall at 770–790 FH. An additional slot for the optical inspection of the sixth stage was designed to allow for field maintenance (Figure 19).

**Figure 19.** New slot designed to inspect the sixth stage.

The standards applicable today determine the maximum chord wear of a blade at the level of 2 mm at the tip, regardless of the use of particle separators. The criteria established in terms of natural vibration frequency and the stall margin allow for an increase in the time between overhauls (TBO) by 200 FH. The defined serviceability limits of the blades enable helicopter users to significantly reduce operating costs by extending the RUL of the engines operated in a desert environment.

The presented methodology can be utilized. However, further studies and more field data are needed to estimate the wear rate of compressor blades operated in other environments. For lower dust concentrations or more effective inlet protection, a more sophisticated model will be necessary to describe particle deposition in the compressor as well as the hot section. Large fleets may require machine learning tools to model significant amounts of diverse and variable data.

**Author Contributions:** Y.D. and D.P. conceived and designed the research; Y.D. processed and analyzed the data; Y.D. and D.P developed FEM and CFD models; D.P. and R.P. verified and evaluated the results. Y.D. and R.P. drew conclusions and produced the paper.

**Funding:** This research received no external funding.

**Acknowledgments:** This publication was prepared within the framework of the AERO-UA project, which has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement No 724034.

**Conflicts of Interest:** The authors declare no conflict of interest. Motor Sich JSC had no role in the design, execution, interpretation, or writing the study. The views, information, or opinions expressed herein are solely those of the authors and do not necessarily represent the position of any organization.

#### **Abbreviations**

The following abbreviations and symbols are used in this manuscript:


#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Predesign Considerations for the DC Link Voltage Level of the CENTRELINE Fuselage Fan Drive Unit**

**Stefan Biser 1,2,\*,†, Guido Wortmann 1,†, Swen Ruppert 3,†, Mykhaylo Filipenko 1,†, Mathias Noe <sup>2</sup> and Martin Boll 1,†**


Received: 2 October 2019; Accepted: 14 November 2019; Published: 20 November 2019

**Abstract:** Electric propulsion (EP) systems offer considerably more degrees of freedom (DOFs) within the design process of aircraft compared to conventional aircraft engines. This requires large, computationally expensive design space explorations (DSE) with coupled models of the single components to incorporate interdependencies during optimization. The purpose of this paper is to exemplarily study these interdependencies of system key performance parameters (KPIs), e.g., system mass and efficiency, for a varying DC link voltage level of the power transmission system considering the example of the propulsion system of the CENTRELINE project, including an electric motor, a DC/AC inverter, and the DC power transmission cables. Each component is described by a physically derived, analytical model linking specific subdomains, e.g., electromagnetics, structural mechanics and thermal analysis, which are used for a coupled system model. This approach strongly enhances model accuracy and simultaneously keeps the computational effort at a low level. The results of the DSE reveal that the system KPIs improve for higher DC link voltage despite slightly inferior performance of motor and inverter as the mass of the DC power transmission cable has a major share for a an aircraft of the size as in the CENTRELINE project. Modeling of further components and implementation of optimization strategies will be part of future work.

**Keywords:** electric propulsion; aircraft; CENTRELINE; DC link voltage level; analytical model; design space exploration

#### **1. Introduction**

While noise levels have been reduced drastically, absolute emissions from aviation, especially carbon dioxide (CO2) and nitrogen oxide (NOx), have been continuously on the rise throughout the last decades, accounting for 2–3% of global greenhouse gas emissions. The large improvements in fuel efficiency are thwarted by the steady increase in passenger numbers [1,2]. The National Aeronautics and Space Administration (NASA) N+3 fundamental technology research for subsonic fixed-wing aircraft [3,4], the Civil Aviation Organization (ICAO) [5,6] as well as the Aviation Transport Action Group (ATAG) [2] and the European Commission's Flightpath 2050 [7] set aggressive target levels for greenhouse gas emissions in the future—up to 75% less fuel burn and 90% NOx emissions by 2050.

To reach those ambitious objectives, a significant improvement in terms of fuel consumption is necessary. These objects can be seen as one of the main drivers for the development of novel propulsion systems [8], resulting in a large increase in interest on partly or fully electrified aircraft [9,10]. The change to electric-based propulsion systems not only affects current designs, it completely changes the setup. Three types of EP for aircraft can be distinguished by their level of hybridization [9]. Turbo-electric propulsion systems have a total hybridization of power, i.e., they generate all power from burning fuel in gas turbines [11,12]. Hybrid-electric concepts involve power generation from burning fuel and electric storage [8,13]. All-electric propulsion systems only have batteries on-board for energy storage [14].

An often-discussed point is the choice of the DC link voltage level for power transmission. The voltage level of state-of-the-art aircraft is currently limited to ±270 V DC to avoid corona at typical flight altitudes [15,16]. Increasing electric power demand of aircraft requires higher voltage levels to reduce weight and keep transmission efficiency at a high level [14,16,17]. Considering DC link voltages from 1000 V to over 5000 V, several studies showed that the mass of the power transmission cable resulting from the choice of the DC link voltage level is an important driver in system mass [15,17]. Up to now, the sensitivity of the electric machine and inverter design for electric aircraft on the DC link voltage has not yet been investigated in a coupled research.

In this paper, we elaborate this problem, having a closer look on the propulsion system of the CENTRELINE project [18] where 8 MW of propulsion power are required. In contrast to conventional aircraft powered by gas turbines, especially turbo-electric and hybrid-electric propulsion systems have significantly increased complexity on the system level and mass. Due to the increased number of components with a large number of parameters influencing the mission performance, a large design space is opened up [9,12]. Brelje and Martins [9] reviewed many approaches which try to incorporate sizing methods for EP on different levels of detail into the conventional aircraft sizing procedure. However, most of them use simple correlations based on stochastic regression or fixed parameters for specific power and efficiency of the components within the electric drivetrain [8,12,19–23]. Accordingly, EP is still poorly understood due to the low fidelity models that are used [9]. Thus, they proposed to develop a fully coupled, multidisciplinary analysis and optimization (MDAO) for electric aircraft.

As such models require enormous calculation power and long calculation time, we follow a different approach in this paper where each component is described by a physically derived, analytical model linking several subdomain models specific to the component together; e.g., for an electric machine, these are electromagnetics, structural mechanics and thermal analysis. All such components models are finally set up as coupled model for the complete system. This approach aims to strongly enhance the prediction accuracy, especially in terms of plausibility, reliability, and traceability of the results, while at the same time keeping the computational effort low to perform large design space explorations.

The paper is structured as follows: Firstly, the CENTRELINE project and the reference propulsion architecture are shortly introduced. Secondly, the specific challenge is presented, and the sizing process of the components is explained in more detail. The results of the large design space exploration are discussed component-wise as well as for the combined system. After a discussion of the results, an outlook of future work will be given.

#### **2. CENTRELINE Project Overview**

CENTRELINE is a project funded by the European Union (EU), serving as a demonstrator for propulsion-aircraft integration with the goal of validating results of former studies on boundary layer ingestion (BLI) such as the FP7 project DisPURSAL [24]. The major change in aircraft design is the implementation of the so-called fuselage fan (FF) drive unit which will be installed in the fuselage aft-end [18]. This device ingests and re-energizes the boundary layer of the fuselage to compensate viscous drag in the fuselage boundary layer flow [18,25].

While DisPURSAL [24] focussed on a conventionally driven fuselage propulsor by a third gas turbine, CENTRELINE will evaluate the possible advantages of a fully turbo-electric drivetrain arrangement to overcome shortcomings of a gas turbine in the aft section such as flow losses associated with the air intake duct, vibration issues and high maintenance costs [26]. The electric powertrain consists of two power generation sets, an electric power transmission system as well as a single fuselage propulsor located in the rear of the aircraft, as shown in Figure 1. Each generation is composed of a wing-podded, advanced geared turbofan (GTF), an electric generator that converts mechanical to electric power, and an AC/DC rectifier. The electric generator is mechanically coupled to an additional free power turbine located in the low-pressure turbine (LPT) section of the GTF even though this is in the hot section, as this has the minimal influence on engine stability. The rectifier will be placed in the nacelle of the GTF to evaluate the potential of air-cooled power electronics [23].

**Figure 1.** H2020 CENTRELINE architecture (turbo-electrically driven fuselage wake-filling) [18].

A DC power transmission system is chosen, consisting of eight parallel lanes to transfer the power as well as protection and switching devices to control it. In the rear, a single propulsion unit including eight parallel DC/AC inverters, one electric motor and the fuselage fan itself is mounted. The inverter converts electric DC to AC power, while the motor converts it to mechanical power to drive the fuselage fan directly without any additional gearbox [27].

Several limitations with regards to installation space of inverter and electric motor (depicted in Figure 2), derived by aerodynamic calculations, must be considered during the sizing process.

**Figure 2.** Fuselage fan drive unit integration with installation space limitations (taken from [27] and modified).

#### **3. Methodolgy**

#### *3.1. Problem Setup*

Brelje and Martins [9] as well as Gesell et al. [12] stated that EP is not applicable as drop-in replacement of existing aircraft propulsion unit but has to be optimized for aircraft and mission requirements. As outlined in Section 1, the proposed approach of sizing components of the electric drivetrain with physically derived, analytical models systematically improves the level of model details in order to raise awareness of importance of a better understanding of EP systems for aircraft. The average run-time per system design point is within the small digits of seconds on an Intel® Core™ i7-6600U CPU @ 2.60 GHz.

The approach was applied to the FF drive unit of the concept aircraft, considering only a reduced system with the electric motor, the DC/AC inverter and the cabling, while other components are neglected. This study aimed to investigate the influence of a change in DC link voltage level of component as well as system KPIs, especially mass and efficiency. Figure 3 shows the block diagram of the setup.

**Figure 3.** Block diagram of FF drive unit including motor, inverter and cable. For sake of simplicity only one inverter-cable configuration instead of all eight parallel lanes which feed the winding systems of the electric motor is shown.

The diagram only shows the parameters which are relevant for the scope of this study, in detail:


General requirements as well as global and local variation parameter define a unique input vector *x*, which results in a unique output vector *KPI*. The necessity of a fully coupled system sizing model is elucidated using the example of the DC link voltage level. Depending on it, the winding layout as well as insulation aspects in the motor vary, which in term influences the selection of semiconductor models during the inverter sizing process. Furthermore, the selection of the DC link voltage level has significant impact on the sizing of the cabling as it is directly proportional to the current. Those effects can only be made visible if the interdependencies between components are considered correctly.

#### *3.2. Requirements, Constraints and Design Target*

The requirements for the electric motor as well as the semiconductor technology of the inverter and the length of the transmission cable stem from the aircraft design process. The electric motor has eight parallel winding systems each fed by a DC/AC inverter and connected to a DC transmission cable, transferring an eighth of the total power, respectively. The range of the DC link voltage of the transmission system is based on a pre-assessment of available semiconductor blocking voltage levels [27]. As the semiconductor module selection is based on the chosen inverter topology and the DC link voltage, five different topologies including two-, three-, and five-level topologies, are considered. Additionally, several internal parameters for each component are varied within reasonable ranges.

Table 1 summarizes the most important global design requirements as well as global and local variation parameters.


**Table 1.** Global requirements and input ranges for global and local variation parameters.

Besides, several constraints, especially concerning the available installation space for motor and inverter, must be considered, as shown in Figure 2. Based on the given input, a large design space exploration is performed. The design target is to identify Pareto-optimal design points with respect to the specified KPIs, i.e., system mass *msys* and efficiency *ηsys*, for the given input sets *x* according to Fang and Qin [28], such that

$$\min\_{\mathbf{x}} \left( \{ m\_{sys}(\mathbf{x}), (1 - \eta\_{sys}(\mathbf{x})) \} \right) \tag{1}$$

where *msys* = ∑*<sup>i</sup> mi* and *ηsys* = ∏*<sup>i</sup> ηi*. Pareto-optimality is chosen as criterion to enable an evaluation of designs with concurring KPIs.

Within the next section, the sizing process of each component is explained in more detail.

#### *3.3. Detailed Description of Sizing Process of Components*

#### 3.3.1. Electric Machine

Several machine types have been studied extensively for electric aircraft [29]. Surface-mounted, permanent magnet synchronous machines (PMSM) have been identified as the most promising technology. PMSM have several DOFs, e.g., the number of poles and slots, the choice of materials for magnets and the laminated core of rotor and stator, the diameter-length-ratio of the rotor *Drot*/*Lrot* or the stator electric current density *Jeff* , which represent the local variation parameters.

The iterative design process of the electric machine includes five disciplines and starts with a defined set of requirements that the final design must fulfil. In the first step, the geometrical layout of the machine is determined. Afterwards, the electromagnetic design includes sizing the magnetic circuit under no-load and load conditions with a non-linear lumped parameter model (e.g., [30,31]), adjusting phase current to reach the required shaft torque *TEM*,*shaf t* as well as the configuration of the winding layout. The electromagnetic calculation is done with Simcenter SPEED 13.06 [31]. The structural mechanics section covers the sizing of the retention sleeve for the magnets for different load cases as well as designing rotor shaft and stator housing. The thickness calculation is based on an analytical press-fit model which evaluates the tangential stresses with respect to the radial overclosure of the fit [32,33]. A further important part of machine design is insulation coordination based on IEC 60664 [34] to ensure secure operating voltage levels. To conclude the design process, a thermal analysis of the motor is performed to ensure that waste heat can be dissipated, which especially dimensions the cooling channels. The allowable conductor current density depends on the heat transfer into the cooling medium, which in turn is a function of the thickness of the insulation hindering heat flux and thus decreases allowable conductor current density. Considering the DC link voltage during the motor design thus facilitates a more realistic design.

Passive masses of the machine including rotor shaft, bearings, housing, and terminal box are hard to calculate as they strongly depend on the actual design as well as installation issues. According to a benchmark of several lightweight motors [35], a factor of two was used to account for the passive masses. This is a fair assumption as the power and rotational speed of the machine is fixed and thus the dimensioning torque for all designs identical.

After those design steps, the result is checked against the requirements. If they are not fulfilled the design process enters a further loop, otherwise the design is valid and the specifications are extracted, e.g., to serve as input for a further component such as the DC/AC inverter.

#### 3.3.2. DC/AC Inverter

DC/AC inverter designs can be realized in different topologies, differing in the height of the single-step output voltage and the total harmonic distortion (THD) [27,36,37]. This study only considers two-level (2L-), three-level (T-type) neutral point clamped (3L(T)NCP-) and fly-cap (3LFC-), and five-level stacked multicell (5LSMC-) DC/AC voltage source inverters. The topologies are explained in detail by the work of Krug [36,37], on which most parts of the design process are based on.

First, the voltage configuration of switches and diodes for the given operation point are determined [36]. A scalable semiconductor characteristic is established from a database with commercially available semiconductor modules and their properties which is then used to perform a coupled calculation of electric and thermal performance. Based on the operation point, the losses of the semiconductor modules are calculated analytically [36,38,39]. The semiconductor modules as well as the heat sink, i.e., a cooling plate, are sized such that the chip temperature increase due to the losses at the calculated operating point is below a certain limit and that the waste heat can be removed via the cooling plate at steady-state conditions. As silicon-carbide MOSFETs (SiC) chips are used, a maximum chip temperature of 150 °C is assumed.

Furthermore, the DC link capacity is calculated based on the expected current ripple calculated analytically using a method provided by Krug [36]. Again, a database of commercially available capacitors is used. The sizing process also includes a mass estimation of the gate driver units and a simple housing but misses a calculation of electromagnetic interference (EMI) filters on AC and DC side. To account for the additional mass of the filters, a factor of two is applied.

#### 3.3.3. DC Power Transmission Cable

The power transmission cables are routed from the generators through pylon and wing to the fuselage center to the fuselage aft, where the motor is located. The total length of the cables adds up to 86.0 m. Due to the long distance, aluminum cables are chosen instead of copper cables because of their higher ratio of electrical conductivity to specific weight. A transmission system with eight parallel DC lanes is used to mitigate the risk of a complete power loss in the event of a cable failure [27].

The geometrical design of the DC transmission cable only consists of a circular conductor material, i.e., copper or aluminum, and a concentric layer of insulation material, as electromagnetic field effects and thus shielding can be neglected. The cable sizing process is based on a coupled model of electrical and thermal analysis as well as insulation coordination. The electrical domain covers the calculation of electrical parameters such as voltage levels, currents and resistances as well as the ohmic losses in the conductor considering the conductor material. Insulation coordination determines the necessary thickness of insulation to avoid arcing according to IEC 60664 [34]. A thermal analysis determines the size of the conductor by iteratively solving a one-dimensional, cylindrically symmetrical heat equation [40]. The size of the conductor is chosen such that the losses can be dissipated via natural convection and radiation at given ambient temperature *Tamb*, and the maximum admissible conductor temperature *Tcond* at the insulation interface is not exceeded.

#### **4. Results and Discussion of Design Space Exploration**

The results of the design space exploration are discussed for the single components as well as for the combined system.

#### *4.1. Results for Components*

#### 4.1.1. Electric Machine

First, the results of the design space exploration for the electric motor are evaluated. Figure 4 shows the gross mass *mEM* (active and passive masses) of the electric machine against its electric efficiency *ηEM* at full load for three different DC link voltages *VDC*, ranging from 1500 V to 3000 V. Each marker represents a valid motor design according to the design process described in Section 3.3.1; however, the graph only shows a fraction of the calculated design for better readability.

The Pareto fronts for each voltage level with respect to minimum mass and maximum efficiency (i.e., minimum losses) are depicted as line plots. As can be seen, the higher is the voltage level, the heavier are the machines for identical efficiency levels. This is a result of the increasing thickness of slot, phase, and turn insulation due to the increased voltage level, which in turn leads to a larger necessary slot area. This effect is intensified by the larger cooling channels required due to inferior heat transfer through the thicker insulation and not compensated by lower current levels.

The influence of increased DC link voltage level *VDC* on the machine efficiency is minor, as especially the copper losses depend on the electric current density of the wire which was unchanged in this study. The slight decrease of efficiency is due to the increased core losses as the absolute loss is proportional to the iron mass at roughly constant specific iron losses. The best machines reach power densities up to 11.0 kW kg<sup>−</sup>1, torque densities up to 50.0 N m kg−<sup>1</sup> and efficiencies up to 98.0%. The peripheral speed of the rotor is moderately high (*vrot* ≈ 90 m s<sup>−</sup>1) due to the large size (*Drot* ≈ 0.4 m), wherefore the high power density can be attributed mainly to a high electromagnetic and thermal utilization of materials. This also reflects in a high current density in the stator (*Jeff* = 17.5 A mm<sup>−</sup>2), a value which is challenging but not impossible with state-of-the-art cooling techniques [41].

**Figure 4.** Gross mass against efficiency of electric machine for different DC link voltage levels *VDC* (*PEM*,*shaf t* = 8.0 MW, *nEM*,*shaf t* = 2100 min<sup>−</sup>1, *Jeff* = 17.5 A mm−<sup>2</sup> and *fEM*,*el* = 420–1750 Hz). Line plots indicate Pareto-fronts for different voltage levels. Not all results shown for better readability (≈5%).

#### 4.1.2. DC/AC Inverter

As the design of the inverter is based on the use of databases of commercially available semiconductor modules with discrete blocking voltages and capacitors and utilizes different topologies, a different visualization method is chosen. Masses and efficiencies for different topologies are plotted against varying DC link voltage.

Figure 5 shows mass respectively efficiency of several inverter topologies against varying DC link voltage. In Figure 5a, the discrete steps at certain voltages can be seen clearly, which are determined by the allowable utilization accounting for derating at high flight altitudes [27] and the blocking voltage of the semiconductors as well as the number of inverter levels. The maximum efficiency within the different sections which can be reached increases with higher DC link voltages *VDC* for the different topologies except for the 2L inverter due to its inferior loss characteristic at high voltages.

Within the sections, the mass of the inverters decreases slightly due to the lower currents as well as the decreasing current ripple on DC link capacitors and the fly caps. The volatility in mass results from the discrete choice of capacitors out of a database. The absolute higher mass for further increased DC link voltage can be explained by the specifically higher mass of the semiconductor modules with higher blocking voltages (see Figure 5b).

The most lightweight inverters reach power densities up to *PInv*,*AC*/*mInv* ≈ 55 kV A kg−<sup>1</sup> (gravimetric) and *PInv*,*AC*/*VInv* ≈ 50 kV A dm−<sup>3</sup> (volumetric) at efficiencies of up to *<sup>η</sup>Inv* = 99.0% for maximum chip temperature *Tchip* = 150 °C at a fluid inlet temperature *Tamb* = 75 °C for a specific resistance *Rth*,*ca* = 0.012 KW−<sup>1</sup> of the heat sink surface to ambient (in this case the liquid). Those values are in good accordance with a literature review [42].

**Figure 5.** Efficiency *ηInv* and gross mass *mInv* for varying DC link voltage *VDC* and different inverter topologies (*PEM*,*el* = 1.0 MV A, cos *ϕ* = 0.85, *Tchip* = 150 °C, *Tamb* = 75 °C and *fEM*,*el* = 1000 Hz).

#### 4.1.3. DC Transmission Cable

Figure 6 shows the DC link voltage vs. the cable mass for the eight DC lane configuration specified in Section 3.2 for aluminum as well as copper conductors. While aluminum has only 60% of the electrical conductivity compared to copper, its density is roughly a third, resulting in 50% of the cable mass per ampere. The mass of the cable reduces significantly with increasing voltage as the current decreases and with that the Ohmic losses, which are the sizing parameter for the conductor cross-section.

**Figure 6.** Cable mass *mCab* vs. varying DC link voltage *VDC* of the DC power transmission cable (*PInv*,*DC* = 1.0 MW, *TCab* = 120 °C and *LCab* = 86.0 m).

With increasing DC link voltage level *VDC*, the cable mass decreases roughly indirectly. For unscreened, stranded aluminum cables with silicone rubber insulation comparable to those in [43], the absolute mass of the insulation only increases slightly for higher voltage levels as the current and thus the conductor cross-section decreases significantly. The relative mass share of the insulation compared to the total cable mass rises from roughly 15% for 1500 V to 35% for 3000 V. A high DC link voltage is thus favorable, while a point of minimum cable mass might exist for even higher voltages when insulation starts to dominate the total cable mass.

#### *4.2. System Results*

Figure 7 shows the results of the CENTRELINE DSE study. Figure 7a outlines the results with respect to the total system mass *mSys* and system efficiency *ηSys* of the FF drive unit including the electric motor as well as eight DC/AC inverters and DC power transmission lanes.

**Figure 7.** CENTRELINE DSE study results: (**a**) Total system mass *mSys* against system efficiency *ηSys* of FF drive unit for different DC link voltage levels *VDC*. Line plots indicate Pareto-front for different voltage levels. Not all results shown for better readability (≈5%). (**b**) Breakdown of component masses *mi* including electric machine, inverters, and cables for Pareto-optimal point for different DC link voltages *VDC* at similar system efficiency *ηSys* ≈ 94.0%.

Two effects superpose each other. First, the general trend implies that, with higher DC link voltage level *VDC*, lower system masses and higher system efficiencies are possible. As both motor and inverter tend to be heavier the mass of the cable must have a major share and its decrease in mass overcompensates the growth of the others. The second effect is more subtle and is explained for the two voltage levels 2100 V and 2200 V. The first effect would imply that a voltage level of 2200 V is more suitable on system level; however, this is clearly not the case. While the difference in motor and cable mass for the two voltages is negligible, it is not in terms of inverter efficiency and mass (see Figure 5) as *VDC* = 2100 V represents one of the above-mentioned section boundaries where the selected semiconductor blocking voltage and thus the module changes. This results in a quantifiable effect on system level.

To clarify this, Figure 7b shows the mass breakdown of the single components for the Pareto optimal points at a similar system efficiency *ηSys* = 94.0% for the four different voltage levels. Motor and inverter masses only feature a minor rise in absolute terms, while cable masses reduce significantly but still have a significant to dominating share (roughly 30–65% of the total mass depending on the voltage level).

It can be concluded that a high voltage level is beneficial in terms of system mass, however not for the electric machine and some inverter topologies, without considering further implications such as available semiconductor modules for even higher voltages or increasing complexity. Possible solutions might be the insertion of DC/DC converters after rectifier and before inverter to reduce the voltage level of motor and inverter and increase the one of the DC transmission system or the use of superconducting technologies [11]. This must be evaluated qualitatively.

#### **5. Conclusions**

The possibilities of coupled, analytical models for sizing EP systems have been demonstrated for the CENTRELINE project. A sensitivity study on the impact of the DC link voltage level revealed opposing effects on the component properties while on the system level the trade was clearly dominated by the DC power transmission cable. The underlying effect of discrete steps for the blocking voltages of semiconductor modules on the system performance has been made visible. Two potential technical solutions to decrease the influence of the DC power transmission system were given.

Further work will focus on the modeling of other EP components, especially heat exchangers to account for additional masses due to the thermal management needed for an electric aircraft, as well as refining existing models, e.g., passive masses of electric machines and EMI filters of inverters. Optimization strategies will be implemented.

A complete integration of the presented approach by mirroring back the results of the electric propulsion sizing process into the preliminary aircraft design loop would unlock high potential to improve quality of results as well as to optimize them.

**Author Contributions:** Conceptualization, S.B., M.B. and M.F.; methodology, S.B., M.B. and M.F.; software, S.B., M.B., S.R. and G.W.; formal analysis, S.B.; investigation, S.B.; data curation, S.B.; writing—original draft preparation, S.B.; writing—review and editing, S.B., M.B. and M.F.; visualization, S.B.; supervision, M.B., M.F. and M.N.; project administration, M.B.; and funding acquisition, G.W. and M.B.

**Funding:** Part of this work was conducted within the CENTRELINE project, which has received funding from the European Union's Horizon 2020 research and innovation programme under Grant Agreement No. 723242.

**Acknowledgments:** They authors want to thank their colleagues Andreas Mayr, Christoph Marxguth, Mabroor Ahmed and Torsten Schröder for fruitful discussions.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Abbreviations**

The following abbreviations are used in this manuscript:


FC Fly Cap

SMC Stacked Multi Cell

EMI Electromagnetic Interference

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Aerospace* Editorial Office E-mail: aerospace@mdpi.com www.mdpi.com/journal/aerospace

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com

ISBN 978-3-0365-4224-9