Original Research Article

International Journal of Spray and

Autoignition flame transfer matrix: Combustion Dynamics

2022, Vol. 14(1-2) 72–81
© The Author(s) 2022
Analytical model versus large eddy Article reuse guidelines:
sagepub.com/journals-permissions
simulations DOI: 10.1177/17568277221086261
journals.sagepub.com/home/scd

Francesco Gant1 , Alexis Cuquel1 and Mirko R. Bothien2

Abstract
Modern gas turbines need to fulfil increasingly stringent emission targets on the one hand and exhibit outstanding oper-
ational and fuel flexibility on the other. Ansaldo Energia GT26 and GT36 gas turbine models address these requirements
by employing a combustion system in which two lean premixed combustors are arranged in series. Due to the high inlet
temperatures from the first stage, the second combustor stage predominantly relies on autoignition for flame stabiliza-
tion. In this paper, the response of autoignition flames to temperature, pressure and velocity excitations is investigated.
The gas turbine combustor geometry is represented by a backward-facing step. Based on the conservation equations an
analytical model is derived by solving the linearized Rankine-Hugoniot conditions. This is a commonly used analytical
approach to describe the relation of thermodynamic quantities up- and downstream of a propagation stabilized flame. In
particular, the linearized Rankine-Hugoniot jump conditions are derived taking into account the presence of a moving dis-
continuity as well as upstream entropy inhomogeneities. The unsteady heat release rate of the flame is modelled as a linear
superposition of flame transfer functions, accounting for velocity, pressure, and entropy disturbances, respectively. This
results in a 3×3 flame transfer matrix relating both primitive acoustic variables and the temperature fluctuations across
the flame. The obtained analytical expression is compared to large eddy simulations with excellent agreement. A discussion
about the contribution of the single terms to the modelling effort is provided, with a focus on autoignition flames.

Keywords
Autoignition, flame transfer matrix, sequential combustion, gas turbines, Rankine-Hugoniot conditions
Date received: 5 November 2021; accepted: 29 January 2021

Introduction autoignition flames, respectively.2,1 In order to assess the

thermoacoustic characteristics of a combustion system, a
The energy landscape is in the midst of a profound trans- common approach is to make use of the network modelling
formation. In order to fulfill the goals of the Paris approach. The complex system is divided into multiple sub-
Agreement, power generation needs to be decarbonized. systems for which (thermo-)acoustic transfer functions can
This will result in a fast increasing share of variable renew- be derived.3 Usually, one of these subsystems contains the
able power, which has to be balanced by technologies that flame response to acoustic fluctuations. This so-called flame
allow for dispatchable power generation able to flexibly transfer matrix (FTM) relates fluctuating quantities up- and
respond to a varying load demand and renewables produc- downstream of the flame. For propagation stabilized lean
tion. Especially, in so-called Power-to-X-to-Power natural gas flames, the FTM can be conveniently measured
schemes, gas turbines are predestined to take over this role.
In order to guarantee best-in-class operational and fuel
flexibility with ultra-low emissions, Ansaldo Energia 1
Ansaldo Energia Swit., Haselstrasse 18, Baden 5400, Switzerland
GT26 and GT36 feature a sequential combustor architec- 2
Zürich University of Applied Sciences (ZHAW), Institute of Energy
ture.1 Sequential combustion systems consist of two com- Systems and Fluid-Engineering, Technikumstrasse 9, Winterthur 8401,
bustor stages. The first lean-premix stage is mainly Switzerland
aerodynamically stabilized (vortex breakdown, flame
propagation), whereas flame stabilization in the second Corresponding author:
Mirko R. Bothien, Zürich University of Applied Sciences (ZHAW),
stage relies predominantly on autoignition. For both Institute of Energy Systems and Fluid-Engineering, Technikumstrasse 9,
flames, fuel and oxidant are premixed and for ease of nota- Winterthur 8401, Switzerland.
tion they are in the following referred to as propagation and Email: mirko-bothien@zhaw.ch

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License (https://
creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission
provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).
Gant et al. 73

under atmospheric conditions due the fact that the turbulent The following integral conservation equations relate the
flame speed is only weakly pressure dependent.4,5For auto- up- and downstream quantities across a flame front19,21:
ignition stabilized flames, this straightforward approach is
Δ(ρ u) = u fl Δρ (1a)
not possible. This is due to the fact that flame stabilization,
autoignition delay time, and reaction rates strongly depend Δ(ρ u2 + p) = u fl Δ(ρ u) (1b)
on the mean pressure level and the flame dynamics are sen-
sitive to oscillations of temperature and pressure. An alter- 1 1
native to measuring FTM under full engine pressure is to Δ(ρu(cp T + u2 )) = u fl Δ(ρ(cv T + u2 )) + Q̇ (1c)
2 2
perform unsteady large eddy simulations (LES) coupled
where p is the pressure, u the velocity, ρ the density, T the
with system identification (SI) techniques. This method-
ology has been introduced for propagation stabilized temperature, Q̇ the flame heat release rate per unit area, and
flames6,7 and has been recently used to characterize reheat u fl the flame location velocity. The coefficients cp and cv
flame dynamics.8,9 Lately, research efforts on autoignition denote the mixture specific heat capacity, at constant pres-
flame thermoacoustics for gas turbine applications have sure and at constant volume respectively. The subscripts
increased. Research at Ansaldo Energia of Bothien and denote states up- (1) and downstream (2) of the flame and
co-workers focuses on flame transfer functions and matri- the symbol Δ stands for the difference between quantities
ces.8–11 Schulz and Noiray12,13 as well as Aditya et al.14 in 2 and in 1, e.g. Δp = p2 − p1 . Closure of the set of
and Gruber et al.15 investigate the occurrence of propaga- Eqs. (1) is done via the ideal gas law, pi = ρi RTi with
tion and autoignition stabilized regimes of reheat flames. i = 1, 2. Together with the system given by Eqs. (1) this
In contrast to propagation stabilized flames, only a few is a nonlinear algebraic system of four equations for the
studies on analytical modelling approaches for autoignition four unknown downstream quantities u2 , p2 , ρ2 , T2 .
flame acoustics exist.16–18 Although the characterization of
autoignition flame dynamics by combined LES/SI
approaches shows very promising results, it comes with Linearization of the conservation equations
the drawback that retrieving the FTM for the whole
engine operating regime and different burner variants in Acoustic disturbances are small fluctuations with respect to
new development programmes is far too time consuming. the mean value. Each physical quantity can be written as a
In this paper, we therefore focus on deriving an analytic sum of a mean value and an acoustic fluctuation as:
φ(x, t) = φ  denotes the mean quantity
 + φ′ (x, t), where (·)
expression for the dynamics of autoignition stabilized ′
flames starting from the general conservation equations, and (·) the fluctuating term. φ is a general flow variable,
which is then benchmarked with LES/SI results. For this e.g., pressure p, velocity u, density ρ.
purpose, the Rankine-Hugoniot jump conditions relating If the acoustic amplitudes are sufficiently small, φ′ ≪ φ
,
fluctuating quantities up- and downstream of the flame the acoustic equations may be derived by a first-order
are derived. Chu19 was the first to derive these jump condi- approximation of the conservation equations neglecting
tions across a flame front in 1953. Additionally, analytical nonlinear second- or higher-order effects (φ′2 < <φ′ ).
models for autoignition flame transfer functions (FTFs) to Linearization of Eqs. (1) then yields the following two
acoustic and entropic disturbances17,11 are incorporated in sets of equations, one for the mean parts
the FTM modelling, resulting in a closed-form expression ρ2 u2 = ρ1 u1 (2a)
directly comparable to LES/SI results.
This work is an extension of two previous research ρ2 u22 + p2 = ρ1 u21 + p1 (2b)
papers,8,20 revisited and deepened with the inclusion of new
material from recent research on autoignition flames.11,18 ρ2 u2 (cp T2 + u22 /2) = ρ1 u1 (cp T1 + u21 /2) + Q̇ (2c)

p2 /(ρ2 T2 ) = p1 /(ρ1 T1 ) = R (2d)

The Rankine-Hugoniot jump conditions and one for the fluctuating ones
The Rankine-Hugoniot conditions are one-dimensional inte-
Δ(ρ′ u + ρu′ ) = u′fl Δρ (3a)
gral conditions describing mass, momentum and energy con-
servation across a discontinuity – in the present case a flame. Δ(p′ + 2ρuu′ + u2 ρ′ ) = 0 (3b)
This discontinuity can be modelled to be at rest or in motion.
As was for example shown in a recent publication,10 in reheat 1
Δ(ρu(cp T ′ + uu′ ) + (ρ′ u + ρu′ )(cp T + u2 )) =
flames the flame front position is highly responsive to the 2
modification of the autoignition delay time of the unburnt ′ 1
mixture. This in turn is influenced by the excitation applied = Q̇ + u′fl Δ(ρ(cv T + u2 )) FQ1 C
2
at the system boundary. Including this phenomenon in the
model is crucial for autoignition flames. p′2 /p2 = ρ′2 /ρ2 + T2′ /T2 , (3c)
74 International Journal of Spray and Combustion Dynamics 14(1-2)

where for ease of notation here and in the following the where Q̃ = p1Q̇u1 and use has been made of the speed of
overbar is dropped. It is also assumed that the mean flame sound c2 = γp/ρ and of the Mach number M ≜ u/c. The
velocity is zero, i.e., u fl = 0, due to the fact that the excita- independent variables p′ , u′ and T ′ have been normalized
tions of the inlet parameters exhibit zero mean value. so that they have the dimension of a velocity. Equation
(6) provides the downstream fluctuating quantities as func-
tion of the upstream ones. We focus now on the last two
Solution for the mean quantities ′
summands and try to express Q̇ and u′fl as functions of
First, we focus exclusively on the mean quantities, given in the upstream quantities p′1 /(c1 ρ1 ), RT1′ /c1 and u′1 .
Eqs. (2). These are four nonlinear coupled equations in the
unknowns u2 , p2 , ρ2 , T2 . The system can be solved analyt-
ically and the solution expanded in the Mach number19,21: Representation of the unsteady heat release rate
u2 ρ1 T2 γ − 1 Q̇ It is assumed that the individual contributions of pressure,
= = =1+ + O(M12 ) (4a) velocity and temperature fluctuations on the unsteady heat
u1 ρ2 T1 γ p1 u1
release rate can be linearly superposed and that they are
p2 Q̇ related to the relative heat release rate fluctuations by the
= 1 − (γ − 1) M 2 + O(M13 ) (4b)
p1 p1 u1 1 frequency dependent flame transfer functions.
Bothien et al. 8 showed that for the linear case relative
Interestingly, these relations would also hold for the general
heat release rate fluctuations can be expressed as
variables u(x, t), p(x, t), ρ(x, t), T(x, t) and not only for
′
their mean values if a problem without moving discontinu- Q̇ p′ u′ T′
ity would have been considered. = Fp 1 + Fu 1 + FT 1 = · · ·
Q̇ p1 u1 T1
Linearization of Eqs. (4) is common22,6 and yields: ⎡ ′ ⎤
p1
′ (7)
T2 Q̇ p′  ⎢ ρ 1 c1
⎥
u′2 = u′1 + ( − 1)u1 ( − 1 ) (5a) γ
c1 F p
1
u1 F u
γ ⎢ ′ ⎥
c1 F T ⎣ u 1 ⎦ ,
T1 Q̇ p1 ′
RT1
′ c1
T2 Q̇ u′1
p′2 = p′1 − ( − 1)ρ1 u21 ( + ) (5b) where Fp , Fu , and FT are the complex-valued, frequency
T1 Q̇ u1
dependent flame transfer functions of the single contribu-
′
T2 Q̇ T ′ p′ u′ tions. This approach based on the superposition of contribu-
T2′ = T1′ + ( − 1)T1 ( + 1 − 1 − 1 ) (5c) tions is particularly useful to model the problem and
T1 Q̇ T1 p1 u1
understand the physical mechanisms involved. The key
Since in the current framework the flame is changing its axial point to justify this methodology is the linearity assump-
location due to the ignition delay time dependence on tempera- tion, which has been carefully validated in a previous
ture, Eqs. (5) are not valid and need to be derived with u′fl ≠ 0. work,8 showing that the superposition principle is applic-
In the Discussion section we will elaborate on the reason why able as long as we consider small temperature fluctuations.
Eq. (5) are not suitable to study autoignition flames.

Analytic flame transfer function models

Solution for the fluctuating quantities
The analytical formulations for the autoignition flame trans-
Equations (3) for the fluctuating quantities are linear and fer functions are obtained by merging the model of
can be recast in a matrix form: Zellhuber et al.17 for acoustic fluctuations with the model
⎡ ⎤⎡ p′2 ⎤ of Gant et al.11 for entropic disturbances. The unsteady
γM1 cc12 u1
u2 −γM1 cc12
ρ 2 c2
⎢ c1 p2 + γM1 M2 2M1 −γM1 M2 ⎥⎢ ⎥ heat release rate can then be expressed in the frequency
⎣ c2 p 1 ⎦⎢
⎣ u′2 ⎥ ⎦ = ··· domain as18:
γ2 RT2′
0 γM12 uu21 γ−1 M1 cc21 ′
c2 Q̇ u′1 1 p′1 s′1 −iωτ −ω2 σ2 /2 p′1 −iωτ −ω2 σ2 /2
⎡ ⎤⎡ p′1 ⎤  = + − e e − φ e e
γM1 1 − γM1 Q̇ u1 γ p1 cp p
p1
ρ 1 c1
⎢ 1 + γM 2 ⎥⎢ ⎥
⎣ 1 2M1 −γM12 ⎦⎢ ⎣ u′1 ⎥ ⎦ + ··· p′ (x fl ) −ω2 σ 2 /2 s′
− τφT ωT1 ie−iωτ e−ω σ /2 1
2 2
˜ ˜ γ2 ˜ RT1′ +φp e
−Q̇γM γM − Q̇
1
2
1 M + Q̇γM γ−1 1 1 p1 cp
c1
⎡ ⎤ ⎡ ⎤ (8)
0 ′
u1
u2 −1
⎢ 0 ⎥ Q̇ ⎢ ⎥ ′ where φp and φT are autoignition pressure and temperature
⎣ ⎦u1 + ⎣ 0 ⎦u fl , (6)
Q̇ ˜ sensitivity factors,17,11 τ is the mean ignition delay time, σ
Q̇
u1 p1 1− p2
p1 + Q̇ is a measure of the flame thickness, p′ (x fl ) is the acoustic
Gant et al. 75

pressure fluctuation at the flame location, and s′1 is an where ΔhF is the fuel lower heating value. The hypothesis
entropic fluctuation: of perfectly premixed mixture allows us to neglect equiva-
lence ratio fluctuations. Linearization of Eq. (11) around the
s′1 T1′ γ − 1 p′1
= − (9) mean condition (with u fl = 0 as explained above) gives:
c p T1 γ p1
′
Q̇ ρ′ u′ u′fl
Four different physical mechanisms are responsible for the = 1+ 1− (12)
unsteady heat release rate: Q̇ ρ1 u1 u1
and combination with Eq. (7) yields the expression for the
u′ p′ s′ fluctuating flame velocity:
• ( u11 + 1γ p11 − c1p )e−iωτ e−ω σ /2 are mass flow fluctuations
2 2

⎡ ′ ⎤
at the location 1 at the time t − τ. They are convected p1
to the flame and cause a heat release rate fluctuation at  ⎢ ρ 1 c1
⎥
u −1
u′fl = −γM1 Fp − 1 FγM FT + 1 ⎢ ′ ⎥
⎣ u1′ ⎦ (13)
time t. 1

s′ RT1
• −(τφT ωT1 i)e−iωτ e−ω σ /2 c1p are entropic fluctuations at
2 2
c1

the location 1 at the time t − τ. They affect the autoigni-

tion time of the mixture as it travels downstream towards
the flame front. Final expression
p′
−φp p11 e−iωτ e−ω σ /2
2 2
• are pressure waves affecting the Equation (6) can be expanded inserting Eq. (7) for the
reaction progress of the mixture as it travels downstream unsteady heat release rate, Eq. (13) for the moving flame,
towards the flame front. and Eq. (10) for the FTFs. For the mean quantities in the
p′ (x ) matrices in Eq. (6) , a first order approximation of Eqs.
• φp p1 fl e−ω σ /2 are changes of acoustic pressure at the
2 2

(4) yields:
flame location. They have an important effect on the
flame response at high frequency. u2 ρ1 T2 γ − 1 Q̇
= = =λ≜1+ (14a)
u1 ρ2 T1 γ p1 u1
More details on Eq. (8) and its physical interpretation can p2
be found in the literature.17,11,18 If the acoustic field is =1 (14b)
p1
assumed to be compact, τω < <1 (which is valid for the
plane wave transfer matrix approach considered here) and Substituting Eqs. (14) in Eq. (6), inverting the matrix on the
for low Mach numbers, M < <1, it can be shown that pres- left hand side and computing the products we obtain:
sure fluctuations at the flame x fl can be related to those at ⎡ ′ ⎤ ⎡ √ √
p2
the location 1 by p′ (x fl , t) ≈ p′ (x1 , t). Additionally, using λ + λγ(λ − 1)M12
⎢ ρ2 c′ 2 ⎥ ⎢
Eq. (9), the flame transfer functions Fp , Fu , and FT are ⎢ u2 ⎥ = ⎣ −γ(λ − 1)M1 + γλ(1 − λ)M 3
⎣ ′⎦ √ 1
obtained by comparison of Eq. (8) to Eq. (7): RT2
(γ − 1) λ (λ − 1)M 2
c2 1
1 γ−1 ⎤⎡ p′1 ⎤
Fp = ( − φp (1 − eiωτ ))e−iωτ e−ω σ /2 −
2 2
FT (10a) 0 0
γ γ ⎢ ρ1 c′ 1 ⎥
1 + (1 − γ)(λ − 1)M12 0 ⎥ ⎦⎢ ⎥
⎣ u1′ ⎦+
Fu = e−iωτ e−ω
2 σ 2 /2
γ−1 λ−1
(10b) − γ √λ M1 √1
λ
RT1
c1
−iωτ −ω2 σ 2 /2 ⎡ √ √ √ ⎤⎡ ′ ⎤
FT = −(1 + τφT ωT1 i)e e (10c) A λF p A γMλ1 Fu A λF T
p1
ρ 1 c1
⎢ ⎥⎢ ′ ⎥
⎢
+⎣ BγλM 1 Fp BλFu BγλM1 FT ⎥ ⎢ ⎥
√ √ √ ⎦⎣ u1′ ⎦
C λF p C γMλ1 Fu C λF T RT1
c1

Solution for the moving flame A ≜ −2γ(λ − 1)M12

The dependence of u′fl on the upstream fluctuations can be 1

B ≜ 1 − + (λ − 1)(γ + 1)M12
expressed conveniently if the flame is assumed to be a thin λ
discontinuity between the unburnt and the burnt mixture. C ≜ (λ − 1)(1 − γ)M12
The heat release rate per unit area depends on the amount (15)
of fuel crossing the flame front per unit time, i.e., the rela-
tive velocity between the flame and the mean flow as well as where λ is defined in Eq. (14a). The first matrix in Eq. (15)
the density of the fuel and can be written as represent the static contribution of the flame, whereas the
second matrix in Eq. (15) represent the dynamic, frequency-
Q̇ = ρ1 (u1 − u fl )ΔhF , (11) dependent contribution. Each column of the FTM depends
76 International Journal of Spray and Combustion Dynamics 14(1-2)

on a single element of the flame transfer function, therefore means of a composition transported filtered density function
there are no cross dependencies between FTF elements in method based on a Eulerian formulation.
the FTM. This is to be expected since a linear framework Three compressible simulations are run, exciting the
is considered. Equation (15) can be simplified to Eq. (27) outlet pressure, the inlet temperature and the inlet velocity,
given by Chen et al.21. respectively. The excitation is provided by forcing at dis-
crete frequencies and in form of a low-pass filtered discrete
random binary signal (DRBS), which allows a frequency
Numerical setup broadband excitation of the flame. In order to avoid thermo-
Figure 1 shows the simplified reheat combustor used in this acoustic instabilities, the respective opposite side of the
study to benchmark the analytic model described above. domain that is not excited is set to non-reflecting.
This backward-facing step (BFS) geometry has been used The flame dynamics investigated in this work are
by Bothien et al. 8 to derive the 3×3 flame transfer matrix assumed to be linear and time-invariant. Hence, the
for an autoignition flame. In the following, the geometry, system response is obtained from the sum of the convolu-
the numerical set up, the combustion model and the tions of its finite impulse responses with their respective
system identification methodology are briefly explained. input signals. For this, a suitable system identification (SI)
More detailed information is provided in Ref.8. method is applied to reconstruct the multi-parameter
Perfectly premixed air and methane are injected through system from the large eddy simulations. This method is
the inlet of the domain. On the bottom surface a no-slip well suited to identify the flame response to broadband
boundary condition assures the presence of a boundary input signals while minimizing noise.9In Ref. 8, the single
layer, whereas periodic boundary conditions are applied steps of this procedure are explained in detail, and can be
on the two side walls. The top surface is a symmetry summarized as follows:
plane so as to have a closer representation to the realistic
geometry with a two-sided BFS. The geometry is meshed 1. Velocity, pressure and temperature signals are extracted at
with 2.11 million hexahedral cells and the simulations are a number of different axial location as time-dependent
run on the commercial software ANSYS Fluent v17.0. section-averaged quantities. Each signal is shifted to a ref-
Turbulence scales larger than the mesh size are computed, erence location applying a characteristic-based filter
whereas the Smagorinsky model with Lilly dynamic pro- (CBF).24 This methodology is able to filter out turbulent
cedure is used to model the subgrid-scale. The turbulent noise and retain only the meaningful signal information.
Schmidt number is set to 0.7 for scalar flux and for 2. The flame is assumed to behave as a linear time invariant
species. The combustion model relies on tabulated chemis- system. The inputs to this system are the upstream vel-
try and a progress variable and is particularly suited to study ocity, pressure and temperature signals, the output is the
autoignition flames dynamics.23 In particular, the reaction volume-integrated heat release rate. The finite impulse
kinetics are tabulated and obtained from a 0D homogeneous responses (FIR) of the system are calculated from its
reactor calculation with detailed chemistry. A composite input and output information using the Wiener-Hopf equa-
progress variable is defined and utilized to parametrize tion.6 The FIR transformed in the frequency domain
the reaction evolution, and all thermo-chemical quantities provide the flame transfer functions Fu , Fp and FT , for
of interest are tabulated as function of this progress variable the velocity, pressure and temperature respectively.
and of the mixture fraction. The combustion model solves 3. To identify the FTM the same system identification pro-
the transport equations of the progress variable and the cedure is applied, considering in this case as output
mixture fraction and reads the other variables values from signals the downstream velocity, pressure and tempera-
the tables, including intermediate species and products.23 ture fluctuations. Nine frequency dependent functions
The chemistry-turbulence interaction is modelled by Fij are obtained, relating up- to downstream fluctuations:
⎡ ′ ⎤ ⎡ ⎤
p2 ⎡ ⎤ p′1
ρ2 c2
⎢ ′ ⎥ F 11 F12 F13 ⎢ ρ 1 c1
⎥
⎢ u2 ⎥ = ⎣ F21 F22 F23 ⎦⎢ u′1 ⎥. (16)
⎣ ′⎦ ⎣ ′⎦
RT2 F31 F32 F33 RT1
c2 c1

Discussion
Flame transfer functions
In Figure 2, the individual flame transfer functions are
plotted over the Strouhal number, which is defined as Sr ≜
Figure 1. Backward-facing step used for LES reproduced from fu/L with L being the step height and u the mean flow
Bothien et al.8. velocity.
Gant et al. 77

Figure 2. Individual flame transfer functions of the BFS autoignition flame. LES/SI result: solid blue, LES discrete forcing: yellow
diamonds, analytical model of Eqs. (10): red line with crosses.

Comparison of the gains (top row) reveals the import- of magnitude calculations8 show that for a typical reheat
ance of pressure and temperature fluctuations on the gener- flame p′1 /p1 ≈ 0.1u′1 /u1 and therefore that velocity fluctua-
′
ation of heat release rate fluctuations for autoignition tions can significantly impact Q̇ /Q̇ likewise.
flames. For example at Sr = 0.33, |Fp | = 10 and |FT | =
35 whereas |Fu | is two orders of magnitude smaller
(|Fu | = 0.3). The fact that temperature fluctuations can Flame transfer matrix
play such a large role is due to the exponential dependence Figure 3 shows the full 3×3 transfer matrix. The LES/SI
of the mixture ignition delay time on temperature.11 results (solid blue) are reproduced from Bothien et al.8. The
Consequently, the flame position is largely affected and analytic FTM model given in Eq. (15) is presented in combin-
so is the heat release rate fluctuation.10 It has to be noted ation with the LES/SI FTFs (red line with crosses). As
that this effect is very distinct for the BFS configuration expected, the model show excellent agreement for all ele-
studied here since the resulting flame is almost only auto- ments. Significant differences are present only in the phase
ignition stabilized. This can be deduced from the almost of the elements F12 , F31 and F32 . They are due to the fact
vertical flame front in Figure 1 and is due to the very low that the corresponding gains are small and hence the phase
turbulence levels at the inlet. For a real engine reheat combus- estimation is error-prone. All elements with gains larger than
tor, the contribution of autoignition to the flame stabilization 0.5 are well reproduced both in gain and in phase confirming
is still larger than the contribution of propagation, as can be the correctness of the assumptions that are made in deriving
deduced for example from the flame shapes shown by Eq. (15). Additionally, this excellent match also verifies the
Yang et al.9, but will be significantly less pronounced. consistency of the system identification methodology
The analytic model (red line with crosses) Eq. (10) is in applied in Ref. 8 by means of a completely independent ana-
excellent agreement to the discrete forcing results of the lytical approach. All the elements in the third column of
LES (yellow diamonds). Compared to the broadband Figure 3 present non-negligible values, showing how incom-
results (blue, solid), the analytic model is not following ing temperature fluctuations have a significant impact on the
the wavy pattern. acoustic and entropic fields downstream the flame. This fre-
Before discussing the flame transfer matrix, an important quency dependent effect is largely function of the autoignition
point needs to be highlighted: the large difference in magni- delay time of the reactive mixture and, consequently, of the
tude among the flame transfer function gains |Fp |, |Fu | and mean inlet temperature, pressure and of the fuel compos-
|FT | in Figure 2 does not necessarily imply a similar large dif- ition.17,18,11 In general, variations of these parameters increas-
ference in the contribution of pressure, velocity and tempera- ing the autoignition delay time, i.e. shifting the flame in a more
ture fluctuations to the heat release rate fluctuations. Otherwise downstream position, will result in larger gains.
said, even if |Fu | is very small in Figure 2, its contribution to Figure 3 verifies the correctness of the Rankine-Hugoniot
′
the heat release rate fluctuations Q̇ /Q̇ cannot be neglected. relations in Eq. (15) alone. In Figure 4 we aim at verifying
For example, the fact that |Fp | is approximately 30 times the correctness of the Rankine-Hugoniot relations Eq. (15)
larger than |Fu | does not mean that the heat release rate fluc- together with the analytical FTFs from Eq. (10) (purple line
′
tuations Q̇ /Q̇ caused by pressure fluctuations p′1 /p1 will be with crosses). The same considerations presented for
30 times larger than those caused by velocity fluctuations Figure 3 hold. Additionally, Figure 4 presents a green line
u′1 /u1 . This is because, as shown in the first line of Eq. 7, with diamonds, which corresponds to the classical linearization
′
the heat release rate fluctuation Q̇ /Q̇ is affected also by of the Rankine-Hugoniot relations, Eqs. (5). In the last row of
p′1 /p1 and u′1 /u1 , and not only by Fp and Fu . Simple order Figure 4 this line is significantly differing from the other
78 International Journal of Spray and Combustion Dynamics 14(1-2)

Figure 3. Flame transfer matrix from LES/SI approach8 (blue, thick line), analytic formulation of Eq. (15) with numeric FTFs (red line
with crosses).

Figure 4. Flame transfer matrix from LES/SI approach8 (blue, thick line), analytic formulation of Eq. (15) with analytic FTFs Eq. (10)
(purple line with crosses) and classical model Eqs. (5) (green line with diamants).
Gant et al. 79

Figure 5. Flame transfer matrix from LES/SI approach8 (blue, thick line); analytic formulation of Eq. (15) with analytic FTFs Eq. (10),
simplified retaining only O(1) terms in the Mach number M1 (purple line with circles); analytic formulation of Eq. (15) with analytic FTFs
Eq. (10), simplified retaining only O(1) terms, Eq. (17) (red line with crosses).

curves. The reason for the observed mismatch is that autoigni- In Figure 5 a simplified version of Eq. (15) is shown,
tion stabilized flames are characterized by non-negligible oscil- allowing analytical insight into which terms are indeed
lations with respect to their mean position even for relatively necessary to properly model the numerical FTM, and
small fluctuations of the inlet parameters.10This is accounted which ones are negligible. We consider in this regard two
for by the flame velocity fluctuation term u′fl . When dealing simplification approaches. The first simplification scheme
with thermoacoustics of propagation flames this term can be consists in neglecting all terms O(M1 ) or of higher order in
sometimes neglected without affecting the results signifi- the Mach number M1 , e.g. terms multiplied by M1 and M12 .
cantly,22 but necessarily needs to be included to correctly This is justified since for the present study M1 ∼ O(10−1 ),
represent the generation of entropy waves.25 Similarly, as and therefore those terms are at least 10 times smaller
can be seen from the comparison between analytic model than O(1) terms. We present the result of this simplifica-
and LES/SI in the last row of Figure 4, this is the case for auto- tion with the purple line with circles in Figure 5. The
ignition flames. The classic Eqs. (5) (green line with diamonds) second simplification scheme consists in neglecting all
that are often successfully used in the literature for capturing terms of order O(10−1 ) or lower. This is different from
the flame acoustic response22 are not able to correctly repro- the first scheme since terms like M12 FT and M1 Fp are
duce the LES/SI results, because they do not account for the not neglected, due to the fact that FT is O(102 ) and Fp
flame velocity fluctuation term u′fl . is O(101 ). The result of this latter simplification is:

⎡ ⎤ ⎡√ ⎤⎡ p′1 ⎤ ⎡ √ ⎤⎡ p′1 ⎤
p′2
λ 0 0 0 0A λFT
⎢ ρ2 c′ 2 ⎥ ⎢ ⎢ ρ1 c′ 1 ⎥ ⎢ ⎥⎢
ρ 1 c1
⎥
⎢ u2
⎣ ′
⎥=⎣ 0
⎦ 1 0⎥ ⎢ ⎥ +
⎦⎣ 1 ⎦ ⎣ BγλM1 Fp
u BλFu BγλM1 FT ⎦⎣ u′1 ⎥
⎢
⎦
RT1′ √ ′
RT2
0 0 √1λ 0 0C λFT RT1 (17)
c2 c1 c1

1
A ≜ −2γ(λ − 1)M12 B≜1− C ≜ (λ − 1)(1 − γ)M12
λ
80 International Journal of Spray and Combustion Dynamics 14(1-2)

and it is presented with a red line with crosses in Figure 5. exclusively in terms of Mach number dependence, but
The superiority of this second approach is clear when com- requires consideration on the order of magnitude of all
paring the two schemes with the numerical FTM (blue thick terms involved in the expression, and especially FTFs.
line). The large differences in FTFs order of magnitudes
(Figure 2) does not allow for a direct, Mach-dependent sim- Acknowledgements
plification of Eq. (15), but instead requires a slightly more This project has received funding from the European Union’s
complex approach, where also the magnitude of the FTF Horizon 2020 research and innovation programme under the
terms is taken into account. Marie Sklodowska-Curie grant agreement No. 765998,
Before concluding, it is of interest to briefly discuss the ANNULIGhT.
perfectly premixed assumption. Considering the same
geometry and the same operating conditions, inhomogene- ORCID iDs
ities in the fuel distribution on the inlet surface would Francesco Gant https://orcid.org/0000-0001-5787-361X
impact the presented results. In the case of spatial inhomo- Mirko R. Bothien https://orcid.org/0000-0002-1699-8931
geneities the flame front will probably increase its thickness
and, depending on the level of unmixedness, move References
upstream. This would be the result of locally rich spots
1. Pennell DA, Bothien MR, Ciani A et al. An Introduction to
that would autoignite early, generating reactive kernels the Ansaldo GT36 Constant Pressure Sequential Combustor.
that would anticipate the reaction also in the leanest In Volume 4B: Combustion, Fuels and Emissions, Paper No.
regions. The impact on the flame transfer functions is not GT2017-64790. Charlotte, North Carolina, USA: ASME.
easily predictable, but based on the above consideration ISBN 978-0-7918-5085-5. DOI: 10.1115/GT2017-64790.
an effect could be expected through the parameter σ, http://proceedings.asmedigitalcollection.asme.org/proceeding.
which measures the flame thickness. The other situation 2. Güthe F, Hellat J and Flohr P. The Reheat Concept: The
of interest, namely a temporal dependence of the amount Proven Pathway to Ultralow Emissions and High Efficiency
of fuel injected in the domain, is also practically very rele- and Flexibility. J Eng Gas Turbine Power 2009; 131(2):
vant. A closed form relation for the flame transfer function 021503. DOI: 10.1115/1.2836613. http://GasTurbinesPower.
asmedigitalcollection.asme.org/articl.
has been obtained by Zellhuber et al.26 and a numerical
3. Schuermans B, Guethe F, Pennell D et al. Thermoacoustic
study has been performed by Scarpato et al.27, which
Modeling of a Gas Turbine Using Transfer Functions
shows that the impact of equivalence ratio fluctuations on Measured Under Full Engine Pressure. J Eng Gas Turbine
the flame dynamics is similar to that of velocity Power 2010; 132(11): 111503. DOI: 10.1115/1.4000854.
perturbations. http://GasTurbinesPower.asmedigitalcollection.asme.org/articl.
4. Schuermans B, Bellucci V, Guethe F et al. A Detailed
Analysis of Thermoacoustic Interaction Mechanisms in a
Turbulent Premixed Flame. In Volume 1: Turbo Expo 2004,
Conclusions Paper No. GT2004-53831. Vienna, Austria: ASME. ISBN
In this paper, the Rankine-Hugoniot jump conditions are 978-0-7918-4166-2, pp. 539–551. DOI: 10.1115/GT2004-
applied to an autoignition stabilized flame. In conjunction 53831. http://proceedings.asmedigitalcollection.asme.org/
with analytic expressions for the flame transfer functions proceeding.
5. Kobayashi H, Tamura T, Maruta K et al. Burning Velocity of
relating upstream acoustic pressure, velocity as well as fluc-
Turbulent Premixed Flames in a High-pressure Environment.
tuating temperature to the heat release rate fluctuations, a
Symposium (International) on Combustion 1996; 26(1):
3×3 transfer matrix is derived. The analytic transfer func- 389–396. DOI: 10.1016/S0082-0784(96)80240-2. https://
tion is compared to the results of previously conducted linkinghub.elsevier.com/retrieve/pii/S008207849680240.
large eddy simulations coupled with system identification 6. Polifke W, Poncet A, Paschereit C et al. Reconstruction of
routines of the same combustor geometry. It can be con- Acoustic Transfer Matrices by Instationary Computational
cluded that the analytic model excellently agrees to the Fluid Dynamics. J Sound Vib 2001; 245(3): 483–510.
simulations. DOI: 10.1006/jsvi.2001.3594. http://linkinghub.elsevier.
Additionally, it is shown how it is necessary to retain com/retrieve/pii/S0022460X01935941.
flame speed fluctuations in the derivation of linearized 7. Huber A and Polifke W. Dynamics of Practical Premixed
Rankine-Hugoniot conditions, if one is interested in model- Flames, Part I: Model Structure and Identification.
International Journal of Spray and Combustion Dynamics
ling entropy wave generation of autoignition stabilized
2009; 1(2): 199–228. DOI: 10.1260/175682709788707431.
flames. This is due to the fact that autoignition flames
http://journals.sagepub.com/doi/10.1260/175682709788707431.
strongly react to temperature fluctuations due to the expo- 8. Bothien M, Lauper D, Yang Y et al. Reconstruction
nential temperature dependence of the ignition delay time. and Analysis of the Acoustic Transfer Matrix of a Reheat
Finally, the linearized Rankine-Hugoniot conditions are Flame From Large-Eddy Simulations. J Eng Gas Turbine
simplified neglecting terms of order O(10−1 ) or smaller, and Power 2018; 141(2): 021018. DOI: 10.1115/1.4041151.
it is shown that this procedure cannot be performed http://gasturbinespower.asmedigitalcollection.asme.org/articl.
Gant et al. 81

9. Yang Y, Noiray N, Scarpato A et al. Numerical Analysis of the Response of Reheat Flames to Entropy Waves. Combust
Dynamic Flame Response in Alstom Reheat Combustion Flame 2020; 222: 305–316. DOI: 10.1016/j.combustflame.
Systems. In Volume 4A: Combustion, Fuels and Emissions, 2020.09.005. https://linkinghub.elsevier.com/retrieve/pii/
Paper No. GT2015-42622. Montreal, Quebec, Canada: ASME. S001021802030389.
ISBN 978-0-7918-5668-0. DOI: 10.1115/GT2015-42622. http:// 19. Chu BT. On the generation of pressure waves at a plane flame
proceedings.asmedigitalcollection.asme.org/proceeding. front. In International Symposium on Combustion 1952.
10. Gant F, Scarpato A and Bothien MR. Occurrence of Multiple Cambridge, Massachusetts, USA.
Flame Fronts in Reheat Combustors. Combust Flame 2019; 20. Gant F and Bothien MR. Autoignition flames transfer
205: 220–230. DOI: 10.1016/j.combustflame.2019.04.013. matrix modeling. In 26th International Congress on Sound
https://linkinghub.elsevier.com/retrieve/pii/S001021801930158. and Vibration (ICSV26), volume 2. Montreal, Quebec,
11. Gant F, Bunkute B and Bothien MR. Reheat Flames Response Canada.
to Entropy Waves. Proc Combus Inst 2021; 38(4): 6271–6278. 21. Strobio Chen L, Bomberg S and Polifke W. On the jump con-
DOI: 10.1016/j.proci.2020.05.007. https://linkinghub.elsevier. ditions for flow perturbations across a moving heat source. In
com/retrieve/pii/S154074892030007. 21st International Congress on Sound and Vibration
12. Schulz O and Noiray N. Combustion Regimes in Sequential (ICSV21). Beijing, China. https://www.researchgate.net/
Combustors: Flame Propagation and Autoignition At Elevated publication/260278261_On_the_Jum.
Temperature and Pressure. Combust Flame 2019; 205: 22. Bellucci V, Schuermans B, Nowak D et al. Thermoacoustic
253–268. DOI: 10.1016/j.combustflame.2019.03.014. https:// Modeling of a Gas Turbine Combustor Equipped With
linkinghub.elsevier.com/retrieve/pii/S001021801930108. Acoustic Dampers. J Turbomach 2005; 127(2): 372–379.
13. Schulz O, Jaravel T, Poinsot T et al. A Criterion to Distinguish DOI: 10.1115/1.1791284. http://Turbomachinery.
Autoignition and Propagation Applied to a Lifted asmedigitalcollection.asme.org/article.
Methane–air Jet Flame. Proc Combus Inst 2017; 36(2): 23. Kulkarni R, Bunkute B, Biagioli F et al. Large Eddy
1637–1644. DOI: 10.1016/j.proci.2016.08.022. https:// Simulation of ALSTOM’s Reheat Combustor Using
linkinghub.elsevier.com/retrieve/pii/S154074891630411. Tabulated Chemistry and Stochastic Fields-Combustion
14. Aditya K, Gruber A, Xu C et al. Direct Numerical Simulation Model. In Volume 4B: Combustion, Fuels and Emissions,
of Flame Stabilization Assisted by Autoignition in a Reheat paper no. GT2014-26053. Düsseldorf, Germany: ASME.
Gas Turbine Combustor. Proc Combus Inst 2018; 37(2): ISBN 978-0-7918-4569-1. DOI: 10.1115/GT2014-26053.
2635–2642. DOI: 10.1016/j.proci.2018.06.084. https:// http://proceedings.asmedigitalcollection.asme.org/proceeding.
linkinghub.elsevier.com/retrieve/pii/S154074891830267. 24. Kopitz J, Bröcker E and Polifke W. Characteristics-based
15. Gruber A, Bothien MR, Ciani A et al. Direct Numerical filter for identification of acoustic waves in numerical
Simulation of Hydrogen Combustion At Auto-ignitive simulation of turbolent compressible flow. In 12th
Conditions: Ignition, Stability and Turbulent Reaction-front International Congress on Sound and Vibration (ICSV12).
Velocity. Combust Flame 2021; 229: 111385. DOI: 10. Lisbon, Portugal. https://www.researchgate.net/publication/
1016/j.combustflame.2021.02.031. https://linkinghub. 255738482_Characteri.
elsevier.com/retrieve/pii/S001021802100100. 25. Strobio Chen L, Bomberg S and Polifke W. Propagation and
16. Ni A, Polifke W and Joos F. Ignition Delay Time Modulation as a Generation of Acoustic and Entropy Waves Across a Moving
Contribution to Thermo-Acoustic Instability in Sequential Flame Front. Combust Flame 2016; 166: 170–180. DOI: 10.
Combustion. In Volume 2: Coal, Biomass and Alternative 1016/j.combustflame.2016.01.015. https://linkinghub.
Fuels; Combustion and Fuels; Oil and Gas Applications; elsevier.com/retrieve/pii/S001021801600028.
Cycle Innovations, Paper No. 2000-GT-0103. Munich, 26. Zellhuber M, Tay Wo Chong L and Polifke W. Non-Linear
Germany: ASME. ISBN 978-0-7918-7855-2. DOI: 10.1115/ Flame Response at Small Perturbation Amplitudes -
2000-GT-0103. http://proceedings.asmedigitalcollection.asme. Consequences for Analysis of Thermoacoustic Instabilities.
org/proceeding. In Proceedings of the European Combustion Meeting.
17. Zellhuber M, Schuermans B and Polifke W. Impact of Cardiff, UK: FESCI.
Acoustic Pressure on Autoignition and Heat Release. 27. Scarpato A, Zander L, Kulkarni R et al. Identification of
Combustion Theory and Modelling 2014; 18(1): 1–31. DOI: Multi-Parameter Flame Transfer Function for a Reheat
10.1080/13647830.2013.817609. http://www.tandfonline. Combustor. In Volume 4B: Combustion, Fuels and Emissions,
com/doi/abs/10.1080/13647830.2013.8176. Paper No. GT2016-57699. Seoul, South Korea: ASME.
18. Gant F, Gruber A and Bothien MR. Development and ISBN 978-0-7918-4976-7. DOI: 10.1115/GT2016-57699.
Validation Study of a 1D Analytical Model for the http://proceedings.asmedigitalcollection.asme.org/proceeding.