betaFeedback

WorksPhysics and Astronomy

Coupled thermoacoustic resolvent analysis of a model two-stream coaxial combustor

Abstract

We derived a resolvent operator to analyze the coupled flame-acoustic effect in a model two-stream coaxial combustor. The theoretical analysis accommodates both the hydrodynamic effect of the active flame and chamber acoustic effect, as well as their coupling effect. Utilizing the coupled thermoacoustic resolvent, we are able to identify the optimal forcing of different mechanisms and their corresponding optimal responses. When this new modeling tool is applied to a two-stream coaxial model combustor, it reveals distinctly different forcing-response characteristics in the coupled system compared to those in a purely hydrodynamic linear system. Acoustic energy peaks, indicative of resonance, are observed. Additionally, it has been found that the acoustic forcing does not maintain a rank-one property. Furthermore, the modes most receptive to the flame-acoustic coupling effect can be identified using a novel sensitivity analysis approach. Interestingly, by employing the coupled thermoacoustic resolvent, we demonstrate that forcing can be designed to negate or counteract perturbations caused by the forcing of another attribute, potentially aiding in the development of effective thermoacoustic control strategies.

*Corresponding author

Email: [email protected] (Y. Lv)

Information for Colloquium Chairs and Cochairs, Editors, and Reviewers

Novelty and Significance Statement

This study proposes a coupled thermoacoustic (CTA) resolvent analysis framework, aiming to gain new insights into the forcing-response relations of closed-loop thermoacoustic systems. Diverging from prior research, this approach shifts the focus to the combustor rather than the flame. It allows for the exploration of various forcing mechanisms and their corresponding responses. Intriguingly, the CTA resolvent can be utilized to design input forcings that cancel or counteract the responses led by an existing forcing, This novel theoretical approach has the potential as an efficient tool for thermoacoustic control.

Author Contributions

  • LC: designed research; performed research; analyzed data; wrote the paper.

  • ZQ: designed research; performed research; analyzed data; wrote the paper.

  • YL: designed research; performed research; analyzed data; wrote the paper.

Authors’ Preference and Justification for Mode of Presentation at the Symposium

The authors prefer OPP presentation at the Symposium, for the following reasons:

  • This paper proposes a novel thermoacoustic analysis method that is different from other related works.

  • This work improves current tools’ predictive capabilities for analyzing thermoacoustic instability.

  • The presentation is focusing on the prediction results, incorporating only the essential background information.

Introduction

Classical thermoacoustic studies on propulsion systems often focus on the stability analysis to identify the unstable eigenmodes. It has been recognized that the thermoacoustic system also exhibits the transient-growth or pseudo-resonance due to the non-normality nature of the linear system. The related short-term behaviors may equally cause detrimental consequences to combustion systems. To effectively analyze such short-term behaviors, the analysis based on the resolvent operator has gained increasing attention. The resolvent analysis directly reveals the amplification property of a non-normal linear system to external perturbations. In the fields of thermoacoustics, the resolvent analysis was used to construct the relation between velocity perturbation (input) and heat-release fluctuation (output) for a laminar M-shaped premixed flame [1] and for a laminar premixed slot flame [2]. More recently, the resolvent operator was leveraged to study the effects of swirl on the flame frequency response [3]. The efforts of exploiting the capability of the resolvent operator were also extended to the studies of turbulent flames. The flame responses to external forcings were investigated in terms of both passive [4] and active [5] jet flames. It has also been taken to predict the flow response in turbulent swirling flames to acoustic perturbations [6].

In the previous resolvent analyses, the focus is often placed on the hydrodynamics in the flame region at the low Mach number limit, with the goal of finding the forcing under which heat-release rate fluctuations are maximized[2, 5, 7]. However, as such the flame is considered as a decoupled open-loop system while the chamber effect that delineates acoustic reflection and resonance is ignored. Although this technique could be useful for the construction of flame transfer functions (FTF), it does not provide a complete picture that describes the coupling between flame and acoustics [8]. More importantly, the optimal forcing structure identified in those resolvent analyses does not necessarily play the same role as that in the coupled flame-acoustic system. These research issues inspire us to derive a coupled resolvent analysis framework which can be used to construct the forcing-response relations of thermoacoustic system. Toward characterizing the thermoacoustic modality in the coupled systems, the studies [9, 10] invoked the compact flame assumption and deemed acoustic velocity at the reference point as a uniform perturbation to the flame. A multi-scale expansion approach was also proposed to model the force between the hydrodynamic and the acoustic field [11]. Some of these coupling strategies will be leveraged in our theoretical framework.

The objective of this study is to construct a coupled thermoacoustic (CTA) resolvent analysis framework, with the aim of gaining new insights into the forcing-response relations of closed-loop thermoacoustic systems. The key difference from the previous studies is that the focus is on the combustor instead of the flame. The forcing mechanism of different attributes can be considered and the corresponding responses can be identified. Intriguingly, by exploiting the (CTA) resolvent, we are able to design input forcings that cancel or counteract the responses led by an existing forcing. The potential of the new theoretical method as a tool for effective thermoacoustic control will be demonstrated.

The rest of the paper is structured as follows. In § 6, the governing equations and framework of coupled thermoacoustic resolvent analysis are derived. In § 7, the configuration of the model combustor and computation setup are described. Some Numerical techniques are introduced in § 8. The results of the coupled thermoacoustic resolvent analysis are presented and discussed in § 9, as well as the comparison with baseline results. Finally, in section 6, the main conclusions are summarised.

Figure 1. Illustration of coupled thermoacoustic resolvent analysis.

Mathematical formulations

The low-Mach approximation is employed to simplify the representation of combustion physics since this condition is commonly fulfilled for gas-turbine combustors. The hydrodynamic effects are preserved in the flame model and examined from the linearized reactive flow equations. The acoustic propagation and reflection in the whole domain are described using the Helmholtz equation. A non-compact source term is added to the Helmholtz equation to model the induced heat release response of the flame. The acoustics are coupled to the flame/flow momentum equation by matching the velocity disturbance at a given location. A similar coupling strategy was employed previously in the studies [9, 10], where however the hydrodynamic modality was not explicitly addressed.

Governing equations of flame

The combustion is governed by the reactive flow equations. Here we consider a Burke-Schumann flame case by assuming an infinitely fast chemistry model. The governing equations are written as:

∂tρ+∇⋅(ρu⃗)=0(1)\partial_t \rho+ \nabla\cdot(\rho\vec{u}) = 0 \tag*{(1)}
∂t(ρu⃗)+∇⋅(ρu⃗u⃗T)=−∇p+∇⋅τ(2)\partial_t (\rho\vec{u}) + \nabla\cdot(\rho\vec{u}\vec{u}^{T}) = -\nabla p + \nabla\cdot\boldsymbol{\tau} \tag*{(2)}
∂t(ρZ)+∇⋅(ρu⃗Z)=∇⋅(ρD∇Z)(3)\partial_t (\rho Z) + \nabla\cdot(\rho\vec{u}Z) = \nabla\cdot(\rho D \nabla Z) \tag*{(3)}
τ=μ((∇u⃗)+(∇u⃗)T)−23μ(∇⋅u⃗)I⃗(4)\boldsymbol{\tau} = \mu\left((\nabla\vec{u}) + (\nabla\vec{u})^{T}\right) - \frac{2}{3}\mu(\nabla\cdot\vec{u})\vec{I} \tag*{(4)}

in which ρ\rho is the density, u⃗\vec{u} is the velocity, τ\boldsymbol{\tau} is the viscous stress tensor, pp is the hydrodynamic pressure and ZZ denotes the mixture fraction. Given the fast chemistry assumption, the species concentration YY, temperature TT, and transport properties (including mass diffusivity DD and dynamic viscosity μ\mu) are functions of ZZ and can be pre-tabulated with respect to ZZ.

Linear model of combustion

To examine the effects of convective perturbation on the flame, we therefore treat the flame as an amplifier that takes certain external forcing as input and gives back heat-release disturbance as the output. The input-output relations are constructed by considering the resolvent of the linearized reactive flow equations. The combustion linear system is derived from Eq. (1) and can be cast to the form:

Bh0∂qh′∂t=Ah0qh′+fh′(5)B_h^0 \frac{\partial q_h'}{\partial t} = A_h^0 q_h' + f_h' \tag*{(5)}

where Ah0A_h^0 and Bh0B_h^0 are matrices evaluated around the steady-state flame solution, q0q^0. The qh′q_h' denotes a set of independent variables of the linearized system, for instance, qh′=(u⃗′,p′,Z′)hq_h' = (\vec{u}',p',Z')_h, which is solved in the flame region; and herein the forcing fh′f_h' denotes certain hydrodynamic or compositional perturbations. Once the modal transformation (q′,f′)=(q^,f^)exp⁡(iωt)(q',f') = (\hat{q},\hat{f})\exp(i\omega t) is introduced, the linear system is transformed to

q^h=Rhf^h,Rh(ω)=(iωBh0−Ah0)−1(6)\hat{q}_h = R_h \hat{f}_h,\qquad R_h(\omega) = (i\omega B_h^0-A_h^0)^{-1} \tag*{(6)}

where RhR_h is the resolvent operator providing an input-output mapping of the linear system. Through the singular value decomposition of RhR_h, we are able to seek the optimal forcing with which the energy gain of ∥q^h∥2/∥f^h∥2\|\hat{q}_h\|^2/\|\hat{f}_h\|^2 is maximized. Note that for combustion problem one often concerns the energy norm of heat release fluctuation, ∥Q^∥2\|\hat{Q}\|^2, in which Q^\hat{Q} can be expressed as

Q^=ρ0cp0(iωT^h+u⃗0⋅∇T^h+u⃗^h⋅∇T0)+(ρ^hcp0+ρ0c^ph)u⃗0⋅∇T0(7)\hat{Q}=\rho^0 c_p^0(i\omega\hat{T}_h+\vec{u}^0\cdot\nabla\hat{T}_h+\hat{\vec{u}}_h\cdot\nabla T^0)+(\hat{\rho}_h c_p^0+\rho^0\hat{c}_{p h})\vec{u}^0\cdot\nabla T^0 \tag*{(7)}

where cpc_p is the heat capacity at constant pressure. For Burke-Schumann flames we may write Q^=ChQ(ω)q^h\hat{Q}=C_h^Q(\omega)\hat{q}_h since T^\hat{T}, ρ^\hat{\rho} and c^p\hat{c}_p are all functions of Z^\hat{Z}.

∣f^∣2|\hat{f}|^2∣q^∣2|\hat{q}|^2
f^h,ux\hat{f}_{h,u_x}∫Ωh∣f^h,ux∣2r dr dx\int_{\Omega_h}|\hat{f}_{h,u_x}|^2r\,dr\,dx∫Ωh∣Q^∣2r dr dx\int_{\Omega_h}|\hat{Q}|^2r\,dr\,dx
f^h,Z\hat{f}_{h,Z}∫Ωh∣f^h,Z∣2r dr dx\int_{\Omega_h}|\hat{f}_{h,Z}|^2r\,dr\,dx∫Ωh∣Q^∣2r dr dx\int_{\Omega_h}|\hat{Q}|^2r\,dr\,dx

Table 1. Energy norm to RhR_h

∣f^∣2|\hat{f}|^2∣q^∣2|\hat{q}|^2
f^h,ux\hat{f}_{h,u_x}∫Ωh∣f^h,ux∣2r dr dx\int_{\Omega_h}|\hat{f}_{h,u_x}|^2r\,dr\,dx∫Ωp^a2γp02r dr dx\int_{\Omega}\frac{\hat{p}_a^2}{\gamma p_0^2}r\,dr\,dx
f^h,Z\hat{f}_{h,Z}∫Ωh∣f^h,Z∣2r dr dx\int_{\Omega_h}|\hat{f}_{h,Z}|^2r\,dr\,dx∫Ωp^a2γp02r dr dx\int_{\Omega}\frac{\hat{p}_a^2}{\gamma p_0^2}r\,dr\,dx
f^a\hat{f}_a∫Ω∣f^a∣2r dr dx\int_{\Omega}|\hat{f}_a|^2r\,dr\,dx∫Ωp^a2γp02r dr dx\int_{\Omega}\frac{\hat{p}_a^2}{\gamma p_0^2}r\,dr\,dx

Table 2. Energy norm to RctaR_{cta}

Note that in this study, Q^\hat{Q} is defined as a local quantity with spatial dependency, not an integral quantity as employed for compact flame analyses [9, 12]. Since the resolvent operator directly relates the heat-release perturbation with the input velocity disturbance, hence it has been used to effectively estimate FTFs for thermoacoustic analysis [1, 2].

Acoustic model

The acoustics in the combustor are described using the Helmholtz equation with a source term that represents the heat release of the flame. This model preserves the dominant source of sound generation from the combustion region and is sensible for the low-Mach regime. It reads:

1γ0p0∂2pa′∂t2−∇⋅(1ρ0∇pa′)=γ0−1γ0p0∂Q′∂t(8)\frac{1}{\gamma_0 p_0}\frac{\partial^2 p_a'}{\partial t^2}-\nabla\cdot\left(\frac{1}{\rho_0}\nabla p_a'\right)=\frac{\gamma_0-1}{\gamma_0 p_0}\frac{\partial Q'}{\partial t} \tag*{(8)}

where p0p_0 and ρ0\rho_0 are the pressure and density of the steady-state solution and γ0\gamma_0 is the adiabatic index. We introduce the solution variable q^a=(p^a,r^a)\hat{q}_a=(\hat{p}_a,\hat{r}_a) with r^a=iωp^a\hat{r}_a=i\omega\hat{p}_a and perform the modal transformation as in § 6.2. Equation (5) can be recast to the linear form

Aa0q^a=iωBa0q^a−iωCaQQ^−f^a(9)A_a^0\hat{q}_a=i\omega B_a^0\hat{q}_a-i\omega C_a^Q\hat{Q}-\hat{f}_a \tag*{(9)}

with the matrices Aa0A_a^0, Ba0B_a^0 and CaQC_a^Q to be evaluated from the steady-state flame solution. The additional acoustic forcing f^a\hat{f}_a is introduced to the right-hand side of Eq. (9) to examine the transient growth of the system. In a typical engine configuration, the forcings of such kind may originate from turbomachinery parts or arise from structural vibration. Note that the acoustic velocity perturbation u⃗a′\vec{u}_a', although not directly solved, can be derived from the pressure field.

Coupled thermoacoustic resolvent analysis

From the physical view of compressible flow, the flame and acoustic effects are always coupled in actual combustors, which potentially indicates thermoacoustic instability. To investigate the coupling effect, we follow the commonly used approach [13, 14, 15]. Specifically, the acoustic velocity perturbation at a reference location is selected as the input to interrogate the flame model. The reference location is typically close to and upstream of the flame. This treatment may be translated as a mathematical expression:

u⃗h′(x)=u⃗a′(x),for x∈Γref,(10)\vec{u}_h'(x)=\vec{u}_a'(x),\quad\text{for }x\in\Gamma_{\mathrm{ref}}, \tag*{(10)}

Here Γref\Gamma_{\mathrm{ref}} denotes the location where this matching condition is enforced. For notational convenience, we may augment Eq. (10) in terms of solution variables:

Ph(x−xref)q^h=Pa(x−xref)q^a,(11)P_h(x-x_{\mathrm{ref}})\hat{q}_h=P_a(x-x_{\mathrm{ref}})\hat{q}_a, \tag*{(11)}

where PP plays a similar role as a delta function. Combining Eqs. (6), (9) and (11), we obtain the following coupled model relation:

\begin{bmatrix} \hat{q}_h\\ \hat{q}_a\\ \eta \end{bmatrix} = \underbrace{\begin{bmatrix} \mathrm{i}\omega B_h^0-A_h^0 & 0 & P_h^{T}\\ \mathrm{i}\omega C_a^0C_h^0 & \mathrm{i}\omega B_a^0-A_a^0 & 0\\ P_h & -P_a & 0 \end{bmatrix}^{-1}}_{R_{cta}} \begin{bmatrix} \hat{f}_h\\ \hat{f}_a\\ 0 \end{bmatrix}, \tag*{(12)}\end{bmatrix}

where η\eta is simply an auxiliary variable. RctaR_{cta} denotes the coupled thermoacoustic resolvent analysis operator. For thermoacoustic problem, the output response of interest is the energy of acoustic pressure, defined as

Ea=12∫Ωp^a2γp0r drdx(13)E_a=\frac{1}{2}\int_{\Omega}\frac{\hat{p}_a^2}{\gamma p_0}r\,\mathrm{d}r\mathrm{d}x \tag*{(13)}

of which the integration is performed for the whole domain of the combustor, Ω\Omega. The optimal forcing that leads to the maximum of EaE_a is directly relevant to thermoacoustic prediction and control. Two types of input forcing, which correspond to distinct physical mechanisms, are preserved to examine the non-normal behavior of this two-way coupled system.

Combustor configuration and computational setup

Figure 2 shows the configuration of the axisymmetric model combustor considered in this study. The combustor facilitates a two-stream coaxial diffusion flame. The inner and outer nozzles issue the fuel and oxidizer, respectively. The inner nozzle diameter is dd while the width of the outer one is 0.4d0.4d. A small gap of 0.1d0.1d separates these two nozzles. The combustor domain extends to x∈[−50d,240d]x\in[-50d,240d] along the axial direction while keeping a height of 20d20d. As discussed in the previous section, the flame linear analysis is performed on a subdomain, Ωh\Omega_h, where combustion occurs. Ωh\Omega_h is encompassed by the boundaries (interfaces) Γinlet\Gamma_{\mathrm{inlet}} and Γflame\Gamma_{\mathrm{flame}}. Γinlet\Gamma_{\mathrm{inlet}} is located at xflame,l=−5dx_{\mathrm{flame},l}=-5d, while Γflame\Gamma_{\mathrm{flame}} is defined with xflame,r=160dx_{\mathrm{flame},r}=160d and Rflame=8dR_{\mathrm{flame}}=8d. The dimension of Ωh\Omega_h is purposely chosen so that the quantitative results of linear analysis are adequately captured within Ωh\Omega_h.

Figure 2. Illustration of the combustor configuration and the computational domain.

The nonlinear simulation of reactive flow is conducted on the whole domain to generate the base-flow solution, q0q^0. In this study, we are specifically interested in the oxy-rich carbon-free combustion technology. Therefore, the fuel and oxidizer are selected to hydrogen and oxygen, respectively, and issued through the nozzles with a fixed velocity ratio of U1/U2=0.2U_1/U_2=0.2. The ambient temperature T1T_1 and pressure are 300 K and 1 atm. Under this operating condition, the Reynolds numbers corresponding to the fuel and oxidizer streams are 200 and 625, respectively. The characteristic Mach number is about 0.03.

Special attention is paid to the boundary conditions for resolvent analysis. The settings are similar to those in a recent study [16]. On Γaxis\Gamma_{\mathrm{axis}} (m=0m=0): Neumann condition for ur′u_r', p′p' and Z′Z' while Dirichlet condition for ur′u_r' and uθ′u_\theta'. On Γoutlet\Gamma_{\mathrm{outlet}} or Γflame\Gamma_{\mathrm{flame}}: stress-free condition for u⃗′\vec{u}' while Neumann condition for the other variables. On Γwall\Gamma_{\mathrm{wall}}: no-slip conditions for velocity and Neumann condition for the other variables. On Γinlet\Gamma_{\mathrm{inlet}}, the u⃗′\vec{u}' is zeroed out for standard resolvent while the matching condition Eq. (11) is enforced for coupled resolvent analysis. At the combustor inlet, q′q' are zeroed out.

Numerical techniques

The governing equations for both baseflow calculation and resolvent analysis are discretized with a finite element method in the open-source software FreeFem++ [17]. The P2 elements are used for the velocity components, and the P1 elements are used for the pressure and mixture fraction. The mesh for the whole domain comprises about 160,000 triangles, resulting in 1.6 million degrees of freedom for Eq. (12).

The baseflow is obtained by solving a steady-state problem with the Newton iteration method. The resolvent analysis is conducted as a singular value problem to optimize the energy of the response field with the choice of quadratic norms:

Mq,cta1/2RctaLctaMf,cta−1/2=XctaΣctaYcta∗(14)M_{q,cta}^{1/2}R_{cta}L_{cta}M_{f,cta}^{-1/2}=X_{cta}\Sigma_{cta}Y_{cta}^{*} \tag*{(14)}

Here Mf,ctaM_{f,cta} and Mq,ctaM_{q,cta} are the weighting matrices to measure the energy of forcing and response mode.

LctaL_{cta} has only 0 and 1 as elements to restrict the dimension of the linear system. The optimal and suboptimal modes are contained in the matrices XctaX_{cta} and YctaY_{cta}, which are ordered by their corresponding energy gains σcta,i2\sigma_{cta,i}^{2} in Σcta\Sigma_{cta}. This singular value decomposition is recast as an eigenvalue problem and solved by the implicitly restarted Arnoldi method. The matrix inversions in the numerical procedure are carried out with MUMPS and SuperLU provided by PETSc library [18].

Results and Discussion

Base flow

The steady axisymmetric base flow is shown in Figure 3 for the axial velocity on the top and temperature on the bottom.

Steady base flow for the axial velocity and temperature

Figure 3. Steady base flow for the axial velocity (top) and temperature (bottom).

Baseline acoustic modes

The acoustic eigenvalue problem with passive flame is conducted first as the baseline results for the coupled thermoacoustic resolvent analysis. The first six acoustic eigenmodes are shown in Figure 4. Here, the acoustic eigenfrequencies λc\lambda_c are the natural frequencies of the model combustor, and the St=ωd/U1St = \omega d/U_1 is the Strouhal number. The eigenmodes 3 and 5 are primarily concentrated in the inner nozzle, while the remaining eigenmodes are distributed throughout the combustor and may strongly interact with the flame.

Acoustic eigenspectrum and eigenmodes

Figure 4. Acoustic eigenspectrum and eigenmodes.

Resolvent analysis

The leading five resolvent gains versus different Strouhal numbers are shown in Figure 5 for both RhR_h and RctaR_{cta} to different components in f^h\hat{\boldsymbol{f}}_h and f^a\hat{\boldsymbol{f}}_a. For the RctaR_{cta}, several peak values exist in the optimal gain curves while only one peak value for the RhR_h.

Resolvent gains of five leading resolvent modes for $R_{cta}$ and $R_h$

Figure 5. Resolvent gains of five leading resolvent modes for RctaR_{cta} and RhR_h.

No acoustic feedback is introduced in the standard resolvent analysis for RhR_h, and the flame is regarded as an energy amplifier for solely hydrodynamic perturbations. It exhibits rank-1 property with optimal gain σ12\sigma_1^2 greatly larger than other suboptimal gains. There is only one peak value in the optimal gain curves located at St=0.80St = 0.80 for f^h,ux\hat{f}_{h,u_x} in the streamwise momentum equation, and at St=0.65St = 0.65 for f^h,Z\hat{f}_{h,Z} in the mixture fraction equation.

The optimal forcing and response mode for f^h,ux\hat{f}_{h,u_x} are shown in the Figure 6a. The spatially developing structure with its maximum amplitude located in the mixing layer is tilted by the mean shear, and this can be attributed to the Kelvin-Helmholtz and Orr mechanisms in the previous research [4]. For the f^h,Z\hat{f}_{h,Z} shown in the Figure 6b, the optimal forcing structure representing component perturbations is mainly confined in the inner nozzle upstream of the flame and is convected by the mean flow to the flame front. Compared to the responding field of momentum perturbations shown in Figure 6a, here the diffusion flame is directly disturbed by the component perturbations, and the Q^\hat{Q} is strong both near the flame front due to the chemical reactions and downstream of the flame due to the mean flow convection. In the coupled resolvent analysis for RctaR_{cta}, several peak values in the gain curves are close to the acoustic eigenfrequencies λc\lambda_c. Here, the eigenspectrum λcta\lambda_{cta} for the coupled system is evaluated and shown in Figure 7. The flame-acoustic coupling introduced in RctaR_{cta} not only combines the acoustic-related and hydrodynamic-related eigenmodes, but also leads to the shift of λc\lambda_c. The resolvent norm could be expanded with its eigenvalues λcta\lambda_{cta} and eigenvectors VctaV_{cta} as:

Optimal forcing and response modes of $R_h$

Figure 6. Optimal modes of RhR_h. (a) Optimal forcing (left) and response (right) mode of RhR_h with respect to f^h,ux\hat{f}_{h,u_x} at St=0.80St = 0.80. (b) Optimal forcing (left) and response (right) mode of RhR_h with respect to f^h,Z\hat{f}_{h,Z} at St=0.65St = 0.65. All fields are normalized by their maximum value.

Coupled thermoacoustic eigenspectrum

Figure 7. Coupled thermoacoustic eigenspectrum.

∥Rcta(ω)∥≤∥Vcta−1∥∥Vcta∥1dist⁡(iω,λcta)(15)\left\|R_{cta}(\omega)\right\| \leq\left\|V_{cta}^{-1}\right\|\left\|V_{cta}\right\| \frac{1}{\operatorname{dist}(i\omega,\lambda_{cta})} \tag*{(15)}

Here the resonance term is related to the strong peaks in the gain curves of RctaR_{cta}, especially in the case of f^a\hat{f}_a. The pseudo-resonance term originates from the non-normal nature of both the thermoacoustic system and its hydrodynamic part. The amplification induced by pseudo-resonance can be observed from the small peaks in the gain curves of f^h,ux\hat{f}_{h,u_x} around StSt between 0.2 and 1.5.

Besides the optimal gains, the corresponding forcing and response modes of RctaR_{cta} and RhR_h with respect to f^h\hat{f}_h are compared. Here the frequency St=0.15St = 0.15 near the first strong peak value in the gain curves of RctaR_{cta} are taken as an example. The results are shown in the Figure 8a for f^h,ux\hat{f}_{h,u_x} and Figure 8b for f^h,Z\hat{f}_{h,Z}. The response mode of hydrodynamic velocity u^h\hat{u}_h in RctaR_{cta} is caused by the linear superposition of the boundary condition due to the acoustic velocity u^a\hat{u}_a at the reference point and external forcing f^h\hat{f}_h in the flame domain Ωh\Omega_h. This result indicates that the acoustic-flame coupling terms in RctaR_{cta} could result in different flame responses that is ignored in RhR_h. One possible reason for the different behavior of RctaR_{cta} and RhR_h is the weighting matrix MqM_q used for output response measurement. The energy of acoustic pressure pap_a is directly measured in RctaR_{cta}, while only heat release rate fluctuation Q^\hat{Q} in RhR_h. According to the Rayleigh criterion, the thermoacoustic system could become unstable when pap_a and Q^\hat{Q} are in the same phase, which could not be investigated with RhR_h.

Comparison between $R_{cta}$ and $R_h$

Figure 8. Comparison between RctaR_{cta} and RhR_h. (a) Optimal forcing and response mode of RctaR_{cta} (left) and RhR_h (right) with respect to f^h,ux\hat{f}_{h,u_x} at St=0.15St = 0.15. (b) Optimal forcing and response mode of RctaR_{cta} (left) and RhR_h (right) with respect to f^h,Z\hat{f}_{h,Z} at St=0.15St = 0.15. All fields are normalized by their maximum value.

In addition to the different behaviour of RctaR_{cta} and RhR_h with respect to hydrodynamic forcing f^h\hat{f}_h, the roles of different forcing type f^h\hat{f}_h and f^a\hat{f}_a in RctaR_{cta} are investigated. Here f^h\hat{f}_h could be regarded as the forcing in the level of flame subdomain Ωh\Omega_h and plays roles mainly near the flame front, while f^h\hat{f}_h in the level of whole combustor domain Ω\Omega. From Figure 5, the gain curve has many strong peak values. Here the frequencies St=0.15St = 0.15 and St=1.65St = 1.65 near the first two resonance eigenfrequencies are chosen for comparison. The optimal forcing and response modes of f^h,ux\hat{f}_{h,u_x} and f^h,Z\hat{f}_{h,Z} are shown in the Figure 9a and Figure 9b.

Comparison between $\hat{f}_{h,u_x}$ and $\hat{f}_{h,Z}$ to $R_{cta}$

Figure 9. Comparison between f^h,ux\hat{f}_{h,u_x} and f^h,Z\hat{f}_{h,Z} to RctaR_{cta}. (a) Optimal forcing and response mode of f^h,ux\hat{f}_{h,u_x} (left) and f^h,Z\hat{f}_{h,Z} (right) to RctaR_{cta} at St=0.15St = 0.15. (b) Optimal forcing and response mode of f^h,ux\hat{f}_{h,u_x} (left) and f^h,Z\hat{f}_{h,Z} (right) to RctaR_{cta} at St=1.65St = 1.65. All fields are normalized by their maximum value.

It is found that different hydrodynamic or compositional forcing in Ωh\Omega_h can result in acoustic response with similar patterns but different amplitude amplification in Ω\Omega. Compared to the baseline acoustic eigenmodes at the two frequencies, the response modes here could be regarded as response modes origin from perturbed acoustic eigenmodes. The corresponding results of f^a\hat{f}_a are shown in the Figure 10. The forcing f^a\hat{f}_a is similar to the adjoint acoustic eigenmode in the previous research[19]. The corresponding response modes resemble the acoustic fields induced by f^h\hat{f}_h but with different amplitude amplification.

Optimal forcing and response mode of $R_{cta}$ to $\hat{f}_a$

Figure 10. Optimal forcing and response mode of RctaR_{cta} to f^a\hat{f}_a at St=0.15St = 0.15 (left) and St=1.65St = 1.65 (right). All fields are normalized by their maximum value.

Conclusions

A coupled thermoacoustic resolvent analysis framework is derived and applied to investigate the flame-acoustic interaction of a model two-stream coaxial combustor. Both the hydrodynamic and chamber acoustic effects are incorporated into the framework and coupled by matching the velocity disturbance at a reference location. The forcing-response relations of the flame-acoustic closed-loop system are constructed with the proposed resolvent method. It is found that the optimal gain corresponding to the coupled resolvent exhibits the peaks induced by pseudo-resonance amplification and acoustic resonance. In addition, the coupled resolvent allows us to evaluate the optimal forcing and response modes corresponding to different forcing mechanisms, including the hydrodynamic or compositional perturbations around the flame front and the acoustic forcing present in the whole chamber. It is recognized that the optimal response mode to the hydrodynamic perturbation preserves the rank-1 property, while the optimal response mode to acoustic forcing loses the rank-1 property at certain frequencies. Finally, it is proven that the adoption of optimized external input forcing can effectively cancel the acoustic pressure induced by other perturbations in the combustor. This showcases the potential of leveraging the derived resolvent framework to design active control strategies for thermoacoustic systems.

Declaration of competing interest

The authors have no conflicts to disclose.

Acknowledgments

This work is supported by the NSFC Basic Science Center Program for “Multiscale Problems in Nonlinear Mechanics” (No. 11988102).

References

References

  1. [1]M. Blanchard, T. Schuller, D. Sipp, P. Schmid, Response analysis of a laminar premixed M-flame to flow perturbations using a linearized compressible Navier-Stokes solver, Phys. Fluids 27 (4) (2015).
  2. [2]C. Wang, T. L. Kaiser, M. Meindl, K. Oberleithner, W. Polifke, L. Leshafft, Linear instability of a premixed slot flame: flame transfer function and resolvent analysis, Combust. Flame 240 (2022) 112016.DOI
  3. [3]C. S. Skene, P. J. Schmid, Adjoint-based parametric sensitivity analysis for swirling M-flames, J. Fluid Mech. 859 (2019) 516–542.DOI
  4. [4]M. Casel, K. Oberleithner, F. Zhang, T. Zirwes, H. Bockhorn, D. Trimis, T. L. Kaiser, Resolvent-based modeling of coherent structures in a turbulent jet flame using a passive flame approach, Combust. Flame 236 (2022) 111695.DOI
  5. [5]T. L. Kaiser, G. Varillon, W. Polifke, F. Zhang, T. Zirwes, H. Bockhorn, K. Oberleithner, Modelling the response of a turbulent jet flame to acoustic forcing in a linearized framework using an active flame approach, Combust. Flame 253 (2023) 112778.arxiv.org/abs/2211.13888
  6. [6]T. L. Kaiser, L. Lesshafft, K. Oberleithner, Prediction of the flow response of a turbulent flame to acoustic perturbations based on mean flow resolvent analysis, J. Eng. Gas Turbines Power 141 (11) (2019) 111021.
  7. [7]T. Sayadi, P. J. Schmid, Frequency response analysis of a (non-) reactive jet in crossflow, J. Fluid Mech. 922 (2021) A15.
  8. [8]M. Ihme, Combustion and engine-core noise, Annu. Rev. Fluid Mech. 49 (2017) 277–310.DOI
  9. [9]L. Magri, M. P. Juniper, Global modes, receptivity, and sensitivity analysis of diffusion flames coupled with duct acoustics, J. Fluid Mech. 752 (2014) 237–265.
  10. [10]A. Orchini, M. P. Juniper, Linear stability and adjoint sensitivity analysis of thermoacoustic networks with premixed flames, Combust. Flame 165 (2016) 97–108.DOI
  11. [11]Y.-C. See, M. Juniper, Multiple-scale thermo-acoustic stability analysis of a coaxial jet combustor, Proc. Combust. Inst. 36 (3) (2017) 3863–3871.DOI
  12. [12]S. J. Illingworth, I. C. Waugh, M. P. Juniper, Finding thermoacoustic limit cycles for a ducted Burke-Schumann flame, Proc. Combust. Inst. 34 (1) (2013) 911–920.DOI
  13. [13]C. F. Silva, F. Nicoud, T. Schuller, D. Durox, S. Candel, Combining a Helmholtz solver with the flame describing function to assess combustion instability in a premixed swirled combustor, Combust. Flame 160 (9) (2013) 1743–1754.DOI
  14. [14]P. Salas, Physical and numerical aspects of thermoacoustic instabilities in annular combustion chambers, Ph.D. thesis, Université Sciences et Technologies-Bordeaux I (2013).
  15. [15]E. Mothéau, F. Nicoud, T. Poinssot, Mixed acoustic–entropy combustion instabilities in gas turbines, J. Fluid Mech. 749 (2014) 542–576.DOI
  16. [16]C. M. Douglas, W. Polifke, L. Lesshafft, Flash-back, blow-off, and symmetry breaking of premixed conical flames, Combust. Flame 258 (2023) 113060.DOI
  17. [17]F. Hecht, New development in FreeFem++, J. Numer. Math. 20 (3-4) (2012) 251–266.DOI
  18. [18]S. Balay, W. Gropp, L. C. McInnes, B. F. Smith, PETSc, the portable, extensible toolkit for scientific computation, Argonne National Laboratory 2 (17) (1998).
  19. [19]L. Magri, M. P. Juniper, Sensitivity analysis of a time-delayed thermo-acoustic system via an adjoint-based approach, J. Fluid Mech. 719 (2013) 183–202.

Paper details

vlnc:W_xb3w8p5t
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 8

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents