modeling of large avionic structures in electrical

MODELING OF LARGE AVIONIC STRUCTURES IN ELECTRICAL NETWORK SIMULATIONS Alexandre Piche (1), Richard Perraud (1), Christophe Lochot (2) (1)

EADS Innovation Works France (2) AIRBUS France

[email protected], [email protected], [email protected]

ABSTRACT The extensive introduction of carbon fiber reinforced plastics (CFRP) in conjunction with an increase of electrical systems in aircraft has led to new electromagnetic issues. This situation has reinforced the need for numerical simulation early in the design phase. In this context, we have proposed [1] a numerical methodology to deal with 3D CFRP avionic structures in time domain simulations at system level. This paper presents the last results on this subject and particularly the modeling of A350 fuselage in SABER computation containing the aircraft power distribution. 1. INTRODUCTION In aeronautic industry, the replacement of metallic elements by CFRP ones involves new electromagnetic issues such as the sizing and optimization of functional current path. The functional current corresponds to the current that feeds the equipments and returns through the structure. Previously, the metallic fuselage handles the functions of signal and fault current return and voltage reference for all electrical equipments. In composite aircrafts, two main points need to be considered: • •

A poor conductive structure could induce critical voltage drops for equipment. Depending on the frequency, a certain amount of current could flow through CFRP assemblies and involves local temperature increases with possible mechanical damages.

Therefore, an internal metallic structure called Electrical Structure Network or ESN is introduced on A350 to replace metallic fuselage inherent functions (electrical and environmental) and especially ensure the functional current return.

Figure 1. Electrical Structure Network of A350 In this context, numerical modeling appears to be a powerful way for current distribution prediction in ESN design phase. In a first time, we have presented in 2008 a wired approach [2] for studying current return in a composite fuselage in a frequency range comprised between the DC to few tens of kHz. In a second time, we have included this approach in a global methodology [1] to treat large avionic structures in time domain simulations at system level. Network Quality investigations are effectively addressed by simulation using time domain circuit solvers like SABER. This tool deals with non linear loads with several levels of complexity (from a behavioural model to a component one, by using component library). The metallic fuselage of civil aircrafts was considered until now as a perfect ground plane and was never represented in SABER. For A350, the ESN could be seen as a discrete and non perfect ground plane which must be introduced in electrical network simulations. The objective of this paper is thus to take into account of the ESN impact in SABER simulation containing a realistic model of aircraft power distribution. In a first step, we recall the key points of the numerical methodology and in a second step we present its application on the ESN A350.

2. METHODOLOGY We recall in this part [1] the methodology to model composite structures in time domain circuit solvers. The global system could be divided in two parts: • •

The structure: CFRP skin, internal CFRP elements, internal metallic elements, junctions between elements, cables The equipments : devices connected at the end of cables

fractions. Once a rational macromodel is available, a state-space representation is possible, allowing the implementation in a time domain circuit solver like SABER. The global methodology is described in the figure below. 3D ASERIS BE model of avionic structure + routing of power cables

N ports Y matrix The structure is, for EMC aspects and low level signals, considered as linear from an electromagnetic point of view. The equipment, on the contrary, is mostly non linear (ex: rectifier bridges). As soon as we have to deal with non linear components in the whole simulation chain, the global system have to be solved in time domain. Fortunately, the structure characterization could be done in the frequency domain due to its linearity.

Rational approximation of frequency responses

Compression : ASERIS BE / Multiport computation No approximation

Vector Fitting

State space representation for time domain circuit solvers

2.1. Frequency characterization of the structure The structure characterization is done in the frequency domain with a Boundary Element Method (BEM) adapted to very low frequencies near DC regime (EADS software ASERIS BE). We have developed an approach [2] which consists in the representation of any internal elements (metallic or CFRP) by wires with linear resistances. The CFRP skin is modeled by a surface triangulation with a surface impedance mainly controlled by the presence of metallic mesh over it. Assemblies between structural elements are modeled via equivalent junctions (wires with local resistances). This wired approach is made possible by the very low frequency range of interest for which elements could be represented with a good accuracy by equivalent resistances and inductances. 2.2. Reduction technique A reduction technique, called Multiport or N-port computation, is then used to reduce the problem at the ports of interest. The ports of the system are defined in place of variable local parameters (sources, impedances, junctions…). The reduction consists in finding without approximation (compared to classical BEM formulation) a N-ports matrix relation (Y or Z) between currents and voltages on ports. This multi-scale technique is the exchange interface between 3D linear models and circuit simulations with non linear devices.

Figure 2. Methodology to model large avionic structures in time domain circuit solvers 3. MODELING OF ESN A350 IN TIME DOMAIN CIRCUIT SOLVER The objective of this part is to apply the methodology to the ESN A350 in order to take into account of its impact in time domain simulations containing the aircraft power distribution. 3.1. 3D BEM model of A350 fuselage As mentioned previously, an internal metallic structure (ESN) has been introduced on A350 to replace metallic fuselage inherent functions and especially to ensure the functional current return. The first step is thus to build a 3D BEM model of the A350 fuselage [2] (skin, internal CFRP elements and ESN). The figure 3 shows the internal structure of the model. Side 2

Crown area

Pax floor

2.3. Vector Fitting technique The last step of the methodology is the approximation of the frequency domain tabulated data. To do this, we use a « vector fitting » technique which decomposes each term of the N-port Y matrix in a sum of rational

Cargo floor

Side 1

Front

Figure 3. ESN A530 3D BEM model, internal structure

3.2. Definition of ports, N-port computation and Vector fitting technique In this problem, the ports of the system are defined in place of sources (power distribution) or loads (equipments). We build a three phases route in the fuselage, connecting the equipments (loads) to the power center (figure 4).

The ports 12 to 35 are dedicated to equipments, they are distributed on the 3 phases along the fuselage. Once the definition of ports made, we perform the N-port computation and the « vector fitting » technique. 3.3. Validation of ESN macromodel Before using the ESN macromodel in a network quality investigation, we validate its behaviour with a direct comparison with ASERIS BE. Obviously, this comparison has to be done for a linear configuration of sources and loads (i.e. ASERIS BE is a linear BEM solver working in the frequency domain). In SABER, we simulate the ESN macromodel with realistic models of VFG and ATU (ports 1 to 11) and the following scenario for equipments (port 12 to 35).

Loads in crown area

port12

Position in the model, n° of phase Frame 22, ph 1

27.60 Ω

port13

Frame 32, ph 1

29.05 Ω

port14

Frame 38, ph 1

11.27 Ω

port15

Frame 50, ph 1

9.20 Ω

port16

Frame 52, ph 1

31.26 Ω

port17

Frame 66, ph 1

128.51 Ω

port18

Frame 70, ph 1

35.90 Ω

port19

Frame 86, ph 1

9.34 Ω

port20

Frame 22, ph 2

44.61 Ω

port21

Frame 32, ph 2

245.33 Ω

port22

Frame 38, ph 2

5000 Ω

port23

Frame 50, ph 2

18.40 Ω

port24

Frame 52, ph 2

23.00 Ω

port25

Frame 66, ph 2

137.83 Ω

port26

Frame 70, ph 2

5000 Ω

port27

Frame 86, ph 2

5000 Ω

port28

Frame 22, ph 3

5000 Ω

port29

Frame 32, ph 3

6.48 Ω

port30

Frame 38, ph 3

171.48 Ω

port31

Frame 50, ph 3

18.40 Ω

port32

Frame 52, ph 3

19.71 Ω

port33

Frame 66, ph 3

42.81 Ω

port34

Frame 70, ph 3

34.84 Ω

port35

Frame 86, ph 3

39.77 Ω

Power distribution

Figure 4. Routing of power cables Concerning the power distribution, two entities have to be defined ; the VFG (Variable Frequency Generator) and the ATU (Auto-Transformer Unit). The VFG is a conventional wound rotor synchronous alternator and provides the 230V AC to the primary of the ATU. The ATU converts the 230V AC into 115V AC (secondary connected to the loads). The SABER models of VFG and ATU require respectively 4 and 7 ports. • • • • •

Port 1, 2, 3 : VFG, phase 1 2 3 Port 4 : VFG, neutral (connected to ESN) Port 5, 6, 7 : Primary of ATU, phase 1 2 3 Port 8, 9, 10 : Secondary of ATU, phase 1 2 3 Port 11 : ATU, neutral (connected to ESN)

The figure 5 focuses on VFG and ATU in the aircraft.

ATU neutral

To the loads in crown area

Front VFG neutral

VFG, phase 1 2 3 Primary of ATU phase 1 2 3

Pax floor

Value

Table 1. Scenario of equipments / loads

Secondary of ATU phase 1 2 3 Cargo floor

Figure 5. Zoom on the power distribution

An equipment is modeled by a resistive load calibrated to inject the right amount of current in the ESN.

We could not represent in ASERIS BE the non linear behaviour of VFG and ATU. Consequently, we note the voltages on the three phases at the secondary of ATU obtained in the SABER simulation. • • •

Secondary of ATU, ph1 : V1 = 115.96V / 0° Secondary of ATU, ph2 : V2 = 117.97V / -120° Secondary of ATU, ph3 : V3 = 116.89V / -240°

Then, we inject V1, V2 and V3 as voltage sources in ASERIS BE ESN model and we perform the classical BEM computation. Finally, we compare with SABER the voltages available at the equipments at 800Hz (VEQUIPMENT). The complete validation process is summarized in the figure 6. 3D BEM model of A350 + routing of power cable

Definition of ports

N-port computation Classical BEM formulation Vector fitting

MAST model

ASERIS BE simulation • Scenario of loads (table 1) • Voltage sources

SABER simulation • ESN macromodel • Scenario of loads (table 1) • Realistic SABER model of VFG and ATU

Secondary of ATU

Comparison on VEQUIPMENT Figure 6. Validation process The following table presents the comparison between SABER and ASERIS BE on VEQUIPMENT.

port1

SABER (Volts RMS) 232.89

ASERIS BE (Volts RMS) XXX

port2

236.24

XXX

port3

234.51

XXX

port4

0

XXX

port5

232.89

XXX

port6

236.24

XXX

port7

234.51

XXX

port8

115.96

115.96

port9

117.97

117.97

port10

116.89

116.89

port11

0

0

port12

117.08

117.14

port13

117.29

117.36

port14

117.14

117.21

port15

117.14

117.20

port16

117.26

117.33

port17

117.27

117.34

port18

117.37

117.43

port19

117.34

117.41

port20

119.03

119.00

port21

118.88

118.85

port22

119.39

119.35

port23

119.34

119.31

port24

119.23

119.19

port25

119.23

119.20

port26

119.45

119.41

port27

119.56

119.53

port28

114.82

114.74

port29

114.76

114.68

port30

114.27

114.19

port31

114.32

114.24

port32

114.29

114.20

port33

114.27

114.18

port34

113.92

113.83

port35

113.84

113.75

VFG VFG neutral Primary of ATU Secondary of ATU ATU neutral

Table 2. Comparison between SABER and ASERIS BE on VEQUIPMENT We observe a very good correlation on VEQUIPMENT between SABER and ASERIS BE ; the ESN macromodel is thus validated. 3.4. Network quality investigation The macromodel could now be used in transient systemlevel simulations. To illustrate, we perform in SABER a global simulation (VFG, ATU, ESN model) and we

observe in time domain the voltage fluctuation into ESN (voltage between a given grounding of equipment and a reference located in nose fuselage). The figures 7 and 8 present respectively the SABER test bench and the voltage fluctuation for a given load.

-0.2

Voltage fluctuation (V)

-0.4 -0.6 -0.8

Behavioral model of VFG

-1.0 -1.2 -1.4

Behavioral model of ATU

0

Time (ms) 10 20 30 40 50 Figure 8. Voltage fluctuation for a given load

4. CONCLUSION The increased use of carbon fiber reinforced plastics (CFRP) in civil aircraft involves new challenges for electrical architecture. The strong need for simulation in design phase has lead to the development of specific methods to deal with large avionic structures. Time domain circuit solvers like SABER are typically used to model equipments with non linear components and to check the electrical network quality. Consequently, we have developed a numerical methodology to take into account of 3D CFRP structures in time domain simulations at system level. This paper recalls this methodology and presents the modeling of A350 in a SABER computation containing the aircraft power distribution. The next step is to introduce realistic models of transformer rectifiers to complete the global power distribution. They generate in the aircraft the 28V DC from the 230V AC. The objective in this case is to predict the ESN impact on DC loads. 5. REFERENCES 1. R. Perraud et al. “Electromagnetic modelling of large scale structures with non linear devices”, EMC Europe 2010, September 13-17, 2010, Wroclaw, Poland Figure 7. SABER test bench

2. I. Revel et al. “Modeling strategy for functional current return in large CFRP structures for aircraft applications”, EMC Europe 2008, September 8-12, 2008, Hamburg, Germany