Chinese  |  English
Home Table of Contents

15 April 2023, Volume 63 Issue 4
    

  • Select all
    |
    Review
  • Applications and prospects of machine learning in turbulent combustion and engines
    AN Jian, CHEN Yuxuan, SU Xingyu, ZHOU Hua, REN Zhuyin
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 462-472. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.001
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Significance] With the development of combustion science, large amounts of data containing various kinds of effective physical information are generated by numerical simulation and experimental measurement. Traditional research methods mainly apply model-based physical rules to illustrate such information. However, as the amount of data increases, data-driven methods have gradually gained research attention. Due to the remarkable success of machine learning (ML) techniques in data analysis and processing, they also offer a new way of processing large amounts of data in the field of combustion. [Progress] This study reviews the applications of ML in turbulent combustion, including chemical reactions, combustion modeling, engine performance prediction and optimization, and combustion instability prediction and control. The challenges and future prospects are also discussed. In the area of chemical reactions, the use of ML has been successfully demonstrated for the simplification and optimization of chemical mechanisms as well as for the efficient representation of chemical systems. Similarly, ML applications have produced encouraging results for modeling subgrid-scale processes and for parameterizing PDFs, often outperforming physics-based closure models in a priori studies. However, caution should be exercised in extrapolating these findings to a posteriori applications. Moreover, further studies are necessary to examine the performance of these data-driven models that are typically generated for specific operating conditions in practical applications. To address the limitations of regression models, physics-informed neural networks provide avenues for incorporating physical principles and other fundamental consistencies that are necessary for enabling robust combustion simulations. As for applications in engines, robust intelligent control via ML has only become feasible for combustion experiments in recent years, mainly due to the developments of deep learning. As such, these methods are still not feasible for commercial applications. This is largely caused by the lack of confidence in ML models under unseen conditions, especially in safety-critical applications, and by the large amounts of online training required for the convergence of current ML methods. [Conclusions and Prospects] Given such a background, robustness study is still a top priority. Although many successful studies on the combination between ML and combustion research have been accomplished, the conceptualization of combustion problems in ML frameworks remains a laborious task. Formulating the problem into an ML framework is a prerequisite for the issue to be successfully solved using ML. Clarifying the combustion problem and carefully selecting and preprocessing the obtained data are important. In addition, the careful selection of the ML model, the loss function, and the training and tuning of the model are necessary components for building a predictive model. Moreover, the ML models exhibit various degrees of predictive uncertainties, which are exacerbated by the lack of interpretability in complex models. Therefore, there is an urgent need to establish ML methods with physical insights. More attempts, such as sample construction method, modeling method, and uncertainty quantification or sensitivity analysis, should be conducted to effectively verify the performance of the model. This ensures that the model abides by the laws of physics and that it can accurately represent the simulated system. The holistic combination of data-driven methods with physical insights could have profound impacts on all areas of combustion science and technology, such as data-assisted modeling and simulation techniques, in situ control and optimization strategies, and data-driven screening of alternative fuels.
  • Review of subgrid models of the equation of state in the large eddy simulation of transcritical and supercritical flows and combustion
    ZHOU Mingshuo, DING Siyu, WANG Xingjian
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 473-486. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.013
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Significance] With the requirement of high efficiency and heavy thrust, fluid mixing and combustion under transcritical and supercritical conditions have been increasingly used in advanced propulsion systems. Transcritical and supercritical fluids exhibit numerous mixing and combustion properties different from those of subcritical (low-pressure) fluids because of thermodynamic nonidealities and transport anomalies. Therefore, elucidating the fundamental characteristics of transcritical and supercritical fluid flows and combustion is vital. Extreme operating conditions pose severe challenges to experimental measurements and optical diagnostics, while large eddy simulation (LES) can be used to detail flow structures and combustion dynamics in a relatively affordable manner. To capture abnormal variations in thermophysical properties, introducing a real-fluid equation of state (EOS) is vital, as the EOS is required for the closure of governing equations. [Progress] In high-fidelity simulations of transcritical and supercritical flows, cubic EOSs, including the Soave-Redlich-Kwong EOS (SRK EOS), the Peng-Robinson EOS (PR EOS), and volume-translation methods, have been frequently used owing to their relatively simple forms and high computation efficiency. In LES-based system equations, EOS filtering introduces an unclosed subgrid term, which is generally neglected in ideal-gas problems. However, this subgrid term plays a substantial role in transcritical and supercritical flows owing to the highly nonlinear changes in thermophysical properties. For density-based solvers, subgrid pressure is required, while subgrid density is required for pressure-based solvers. For demonstration, the relative magnitude of the subgrid density with respect to the filtered density is evaluated using the results of direct numerical simulation (DNS) of the transcritical liquid oxygen/methane mixing layer. The contribution of the subgrid density increases with the filtering size, with up to 60% of the DNS-filtered density in the transcritical mixing regions. To account for the subgrid density effect, various types of subgrid density models are considered, including the Reynolds-filtered model (RFM), gradient model (GM), scale similarity model (SSM), and filtered density function (FDF) model. In RFM, the EOS is evaluated using Reynolds-filtered primitive variables rather than Favre-filtered variables, which leads to an underestimation of the filtered density and limited improvement in accuracy. GM is analogous to the Smagorinsky model with both static and dynamic forms, while SSM is formulated upon the assumption of scale similarity between the grid-filter and test-filter levels. Both GM and SSM improve the accuracy of the filtered density to some extent. FDF shows the overall best performance, with the lowest modeling errors. The model assumes a beta distribution form of the Favre FDF to model the subgrid-scale fluctuations. In summary, GM, SSM, and FDF show evident improvements in the filtered density, while RFM needs further improvement to consistently achieve the required accuracy. The development of the subgrid pressure is briefly introduced, and RFM best approximates the subgrid pressure. [Conclusions and Prospects] Future research on the subgrid modeling of the real-fluid EOSs for transcritical and supercritical flows is needed in the following aspects. First, a more accurate and computationally efficient EOS for high-fidelity simulations needs to be developed. Second, more universal and effective subgrid models of EOSs and other thermophysical properties should be explored. The prevailing multidisciplinary methods, such as deep neural networks and random forests, show great potential to improve both the accuracy and efficiency of LES. Enhanced models would strengthen the predictive capability of current modeling and simulation tools and support the development of next-generation propulsion systems.
  • Mechanism and design optimization of acoustic dampers for attenuating combustion instabilities
    YU Zhijian, YANG Qianwen, WANG Yichen, YANG Dong, ZHU Min
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 487-504. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.024
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Significance] Combustion instability is a crucial issue in developing low-emission gas turbine combustors. Meanwhile, future combustors will possess higher temperatures, wider operation parameter ranges, and more complicated geometric structures (e.g., axially staged and annular). Thus, combustion instabilities have the features of high amplitude levels, multiple and time-varying frequencies, and the coexistence of several modes. Passive control employing acoustic dampers for attenuating oscillations has many advantages, such as simple structures, high reliabilities, and low costs. However, the present damper designs encounter great challenges for future combustors. Therefore, multimode wide-absorption acoustic dampers and systematic design optimization methods for multiple dampers must be investigated. [Progress] This paper briefly reviewed the research progress of mechanisms and optimization methods for acoustic dampers and the recent corresponding work conducted by the authors. First, the mechanisms for conventional and multi-bandwidth acoustic dampers were analyzed. Previous acoustic models for holes assumed that the thickness is ignorable and two open ends are inserted into semi-infinite space. A novel semianalytic acoustic model for short holes was proposed to consider sound-vortex interactions in detail. Sound generation and absorption can be well predicted by this model. Additionally, the performance of holes was sensitive to the shape of the hole edges. To broaden the absorption bandwidth of the Helmholtz resonators, parallel perforated materials were installed at the neck of the resonators. A theoretical model was derived to calculate the sound absorption coefficient of this type of resonator and effectively captured the nonlinear effect at the neck. Traditional resonators only possess a single frequency band for suppressing instabilities. Two multi-bandwidth resonators based on elastic membranes and multiple cavities were proposed by the authors. The results showed that elastic membranes and multiple cavities may introduce new frequency bands. Meanwhile, a low-order network model coupled with nonlinear flame dynamics and the acoustic models of resonators was developed to successfully predict thermoacoustic instabilities in cylindrical and annular combustors. The stability of annular combustors could be affected by the asymmetrical flame responses after introducing acoustic resonators. Subsequently, we examined the effects of complex thermoacoustic parameters on the acoustic characteristics of resonators and optimization strategies for designing multiple resonators. A theoretical model combined with the energy equation was established to explore the effect of a temperature difference on resonator performances. The results showed that the entropy disturbance caused by large temperature differences could affect the thermoacoustic stabilities of combustors. The cross-section of resonators was another critical factor in influencing the resonator properties. The acoustic characteristics of perforated liners with variable cross sections were theoretically and experimentally explored. Decreasing the cross-section increased the range of absorption frequency bands. The introduction of resonators for suppressing thermoacoustic instabilities changes the acoustic modes of combustors and the intrinsic modes of flames. An available strategy considering these influences was determined for reasonably designing the resonators. There are many adjustable parameters when multiple resonators are employed simultaneously. An efficient multiparameter adjoint-based optimization strategy for multiple resonators was developed. This algorithm is based on treating the low-order network model by performing the adjoint method. [Conclusions and Prospects] The next generation of multimode wide-absorption acoustic resonators urgently needs to be explored. Moreover, the effects and mechanisms of nonlinearity, mean flow, temperature difference, and other complex physical parameters on the properties of acoustic resonators need to be further explored. Meanwhile, the effectiveness and robustness of current optimization strategies for designing multiple resonators must be improved.
  • State of the art and challenges of flamelet method in gas turbine combustor simulation
    ZHANG Guihua, WU Yuxin, WU Jiahao, ZHANG Yang, ZHANG Hai
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 505-520. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.026
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Significance] Given the consistent increase in the number of evaluation parameters, enrichment of fuel types in gas turbine combustors, and in-depth investigation of the combustion technology, the numerical simulation of combustion processes in gas turbines has become crucial. In various turbulent combustion models, the flamelet method couples numerous chemical components with a small number of scalars by preconstructing a table, which can reduce the number of transport equations to be solved while considering the detailed chemical reactions. The flamelet method, owing to its accuracy and computational efficiency, provides a primary alternative to numerical simulation for gas turbine combustors. [Progress] The present study reviews the advancement of flamelet methods. Subsequently, in view of the future development trend of gas turbine combustors with multiple working conditions, numerous parameters, and low pollution, we reviewed the relevant models and application scope of flamelet methods and analyzed their application and challenges in gas turbine combustors considering the following four aspects: the application of the flamelet method in an adaptive turbulent combustion model, optimized selection of progress variables, coupling with the turbulent model, and its application in pollution analysis. The application of the flamelet method in an adaptive turbulent combustion model includes its coupling with other turbulent combustion models and the development of a multi-region flamelet method. Toward this end, exploring appropriate identification techniques of different combustion modes is crucial, and the machine learning method is a robust tool to address this challenge. The optimized selection of progress variables involves multi-phase flow combustion and multi-fuel and multi-jet problems. However, a universal progress variable that can act as a representative of all problems is lacking. The flamelet method requires different expressions of progress variables in different problems. The development of universal optimization methods form the primary research aim. The coupling with the turbulent model mainly includes the presumed and transport probability density function methods; however, the balance of accuracy and computation cost remains to be elucidated. The application of the flamelet method in pollutant analysis requires solving additional transport equations of pollutants and modifying the expressions of source terms. [Conclusions and Prospects] Based on the review of previous literature, we recommend developing specific validation methods for each submodel in the flamelet method, performing further studies on the coupling effect of different submodels, and obtaining more data on real gas turbine combustion chambers to guide the development of a flamelet method suitable for real gas turbine combustion chambers.
  • Recent progress and challenges in combustion kinetic model of jet fuel
    WU You, YANG Ming, YANG Bin
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 521-545. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.025
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Significance] The numerical simulation of an aero-engine conduces to accelerate the engine research and development processes, which requires an accurate and reliable combustion kinetic model of jet fuel. This paper reviews the research progress of jet fuel (Jet A and RP-3) and their surrogate fuels in recent years. It includes the main composition of jet fuel, physical and chemical property measurement and combustion experiments of jet fuel and surrogate fuels, construction methods for surrogate fuel, surrogate fuel's composition and combustion kinetic models, and verification of the combustion experiment and the kinetic model of surrogate fuel by the combustion characteristics of jet fuel. The key to the numerical simulation of jet fuel combustion is to develop the surrogate fuel, as well as to reasonably construct and simplify its combustion kinetic mechanism, so that it can not only describe the combustion kinetic characteristics with high fidelity, but also be coupled with computational fluid dynamics (CFD) simulation. According to different research purposes, developing an appropriate reduced kinetic model of jet fuel and applying it to the numerical simulation of an aero-engine help to understand the combustion processes in engines. Furthermore, it also provides technical support for engine design optimization and performance improvement. [Progress] The development of the combustion kinetic model of jet fuel is mainly divided into three parts: the experimental study of jet fuel, the construction of surrogate fuels, and the combustion kinetic model of the surrogate fuel. This paper focuses on the progress and challenges of the above three aspects. In experimental study, researchers mainly focus on the combustion characteristics of ignition, flame speed, and oxidation products. Jet A and its surrogate fuels have well-established physical and chemical properties, as well as combustion experimental data that have been extensively validated. Although study of RP-3 starts relatively late, many high-temperature ignition and laminar flame experiments have been conducted, with relatively fewer low-temperature ignition and oxidation experiments. There is an urgent need to conduct accurate fundamental combustion experiments covering a wider range of temperatures, pressures, and equivalence ratios that encompass the operating conditions of aero-engines and establish the fundamental combustion databases of jet fuel and its related component fuels. In terms of surrogate fuel, its construction methods can be roughly divided into four categories, including matching hydrocarbon composition, key physical and chemical properties, characteristic functional groups, and combustion experimental results. Currently, most surrogate fuels match a few characteristics of jet fuels, and the choosing of matching parameters rely heavily on the researcher's experience and research goals. Therefore, it is still necessary to discuss how to develop a systematic construction method for surrogate fuels: One that does not depend on the researcher's experience and only requires selecting the matching parameters of the surrogate fuel based on the research objective. The systematic construction of surrogate fuels requires a fundamental combustion experimental database for jet fuel, while detailed combustion kinetics mechanisms of single- and multiple-components fuels in jet fuel are also necessary. In the study of combustion kinetics models, the surrogate fuels, kinetic mechanisms and combustion experimental data for RP-3, influenced by the experimental research, are mostly one-to-one correspondence. This leads to a lack of universality in surrogate fuels and their detailed mechanisms. Due to the lack of low-temperature ignition and oxidation data, research on the combustion kinetics models of jet fuel primarily focuses on high-temperature ignition and laminar flame speed, with relatively less attention given to the low-temperature mechanisms. Furthermore, in practical applications, detailed kinetics models need to be simplified for use in CFD, so appropriate reduced mechanisms for jet fuel are necessary. [Conclusions and Prospects] The development of the jet fuel combustion kinetic model should move towards database-based and modular construction. Therefore, it is urgent to establish the fundamental combustion database of jet fuel and its related component fuels, as well as continuously develop systematic methods for constructing surrogate fuels, simplifying and optimizing kinetic models. Thus, suitable systematic construction of jet fuel combustion kinetics models can be achieved according to different research objectives.
  • Research progress on control technologies for flame synthesis of condensed-phase nanomaterials
    SHEN Chang, SHAO Sen, GUO Qifeng, ZHOU Yuxin, YOU Xiaoqing
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 546-559. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.016
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Significance] Flame synthesis is a method for the preparation of condensed-phase nanomaterials. It is energy efficient, cost effective and has the potential for large scale industrial applications. [Progress] This review examined the control technologies for the morphology, composition and microstructure of flame synthesized condensed-phase nanomaterials from four aspects, i.e., flame stability, flame temperature and species profiles, product particle size and morphology, and product physicochemical properties. The flame stability control part mainly introduced the swirl stabilization method, including both strong swirl stabilization (Sg, tan＞5 or S＞0.6) and weak swirl stabilization (S≤0.6). The calculation and design criteria of different swirl numbers for different types of swirlers were evaluated. Methods such as the addition of pilot flame, the increase of active component concentrations at the flame nozzle, the use of high-enthalpy precursor solution, and the addition of sheath gas were also summarized. For the control of flame temperature and species profiles, adjustments might be made to the fuel-air equivalence ratio, oxygen flow rate for atomization, or precursor liquid flow rate and concentration. The installation of the cooling meshes and quenching rings, water addition to precursor liquid and high-enthalpy solvent replacement were also good strategies. The particle size and morphology of the product might be controlled indirectly by adjusting the flame temperature and species profiles which could affect the particle dynamics process. From another perspective, it might also be regulated by substrate materials, droplet microexplosion, precursor ultrafine atomization, inclusion of high boiling point active agent, and plasma discharge, etc. These were realized by controlling the droplet size before particle formation and regulating particle attributes for precipitation, collision, and condensation, thereby changing its size and shape. Four primary approaches for controlling physicochemical properties were summarized, including crystal structure control, element doping, core-shell structure design and post-heat treatment. The crystal structure control mainly included temperature-induced phase transitions and doping-induced phase transitions. The element doping was mainly achieved by regulating precursor liquid with different components and different proportions. The core-shell structure design was mainly realized by using the different precipitation characteristics of different solutes or by the installation of auxiliary devices to stagger the time and space distributions of the two different kinds of materials. The post-heat treatment process primarily consisted of the annealing procedures at different atmospheres, temperatures and time durations to remove unwanted combustion residues and defects, and to induce phase transformation, etc. [Conclusions and Prospects] The first two aspects of this review mainly focus on the control of the external parameters such as flame temperatures and species profiles, while the latter two on the control of the internal properties of the materials such as particle size and physicochemical properties. In actual experiments or production, the control technologies will need to be comprehensively used according to specific situations. Considering the complexity of the composition of the precursor liquid, the stability, volatility, and precipitation characteristics had better be comprehensively analyzed before flame synthesis, so that the mechanism identification and active regulation of the material nucleation and other processes will be enabled from the initial stage of particle formation.
    • Research Article
  • Effect of gas flow mixing uniformity on the flashback characteristics of turbulent jet flames
    LI Dan, Lü Hailu, ZHANG Yang, ZHANG Hai, ZHOU Tuo, Lü Junfu
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 560-571. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.006
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Modern gas turbine combustors often adopt premixed combustion technology for lower NO[WTBX]x emissions. In premixed combustion, the flame flashback is an important issue, particularly for hydrogen-rich fuels burned in micromix burners. The combustible gas in the micromix burner is mixed near the burner nozzle outlet, resulting in a nonuniform species mole fraction in micromix turbulent jets. This mixing-induced nonuniformity of the species deserves careful investigation for its impact on the flame flashback. [Methods] By designing four mixing modes (MM 1—4) of H2, CO, and air, this work experimentally studied the effect of gas flow mixing conditions on the flashback characteristics of turbulent premixed jet flames. MM 1: H2, CO, and air directly enter the nozzle through a straight tube. MM 2: H2 and CO flow into a section of PVC (Polyvinyl chloride) pipe through a tee joint and then mix with air through a straight tube. MM 3: H2, CO, and air directly enter the nozzle through a Venturi tube. MM 4: H2, CO, and air enter a chamber equipped with flow conditioning components (sintered metal plate, glass balls, etc.) and then enter the nozzle through a Venturi tube. From MM 1 to MM 4, the degree of the corresponding mixing uniformity of H2, CO, and air increased. The flashback phenomenon is captured using a high-speed camera integrat with the schlieren method. The flow field near the burner exit is measured using particle image velocimetry. [Results] The results showed that the onset of the flame flashback at different fuel/air mixing modes always occures near the burner wall, similar to the classical “boundary layer flashback” phenomenon. High-speed camera images indicated that the near-stoichiometric premixed flame was blue, surrounded by a diffusion flame layer before flashback. For the worst mixing mode, the region of this diffusion flame layer where the flashback starts appeared orange, indicating higher hydrogen concentration in this region. Upon changing to better mixing conditions, the orange diffusion flame disappeared, and the starting point of the flashback was not specific on one side. Under this fuel-lean condition, no surrounding diffusion flame layer appeared, and thus the mixing uniformity could not be directly evaluated through the flame color. As the mixing uniformity was improved, the flashback velocity decreased. The velocity distributions near the burner exit at different mixing modes were top hat-shaped, suggesting that the influence of the mixing conditions on the velocity distribution was unclear. [Conclusions] The mixing-induced nonuniformity of the species has little effect on the flame flashback mechanism. Schlieren images clearly distinguish the onset and position of the flashback. The flashback process observed from schlieren images agrees well with the high-speed camera result. The flashback flow velocity differs among mixing modes, and the flashback is more likely to occur when the mixing condition is worse. The effect of mixing conditions on the flashback is mainly due to the change in the flame propagation speed near the burner rim. The results of this study shed light on the antiflashback burner design.
  • Study of combustion characteristics and flame stabilization mechanism of hydrogen-containing micromix jet flames
    HU Xuechao, BI Xiaotian, LIU Ce, SHAO Weiwei
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 572-584. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.027
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Micromix combustion is an excellent low-pollution combustion technology. However, the instability of micromix combustion based on multiple small flames, especially high-frequency oscillation under high hydrogen content, is still unclear. [Methods] Herein, the emission performance and oscillation characteristics of micromix combustion under different hydrogen enrichments were studied. Furthermore, an experimental study on the combustion instability of hydrogen-rich fuel was conducted using a novel micromix burner under atmospheric pressure and preheated air at 673 K, which provided a reference for practical engineering applications. Power spectral density was used for spectral analysis. Phase-space reconstruction was applied to analyze the developmental changes in the dynamical system and determine the limit cycle oscillations. Proper orthogonal decomposition (POD) was used to analyze flame dynamics under oscillating conditions, and the time coefficients and spatial distribution characteristics of the modes were extracted. Dynamic pressure sensors were arranged in the air inlet and exhaust outlet contraction sections to measure pressure fluctuations. A high-speed camera system was used to realize the fast acquisition of chemiluminescence signals. The NOx emission, dynamic pressure, flame structure, and other combustion characteristics were studied under different hydrogen contents, from pure methane to pure hydrogen. [Results] The results showed that: 1) The micromix burner had an excellent low-emission performance for pure hydrogen with< 5 μmol/mol NOx at 15% O2 and could adapt to a wide hydrogen content to achieve stable combustion. These characteristics indicated that this micromix burner could be directly applied to designing hydrogen turbine combustion chambers. 2) The oscillatory combustion phenomenon occurred when the hydrogen content was between 10% and 20%. Under those conditions, the phase-space reconstruction trajectory manifested as limit cycle oscillation, and the root mean square values of pressure fluctuation were >1%, representing strong correlation structures. High-order harmonics were also found. Heat release was shown as a periodic overall increase and decrease, and the periodic formation and axial propagation of flame vortices could be observed. The flames with high hydrogen contents fluctuated at a high frequency of >900 Hz, but the amplitude of these flames was low. 3) Time-average images were used to characterize the flame structure under different conditions. The decreasing flame height with increasing hydrogen content contributed to the changes in the heat release concentration position. On the one hand, it affected the coupling relationship between the heat release fluctuation and pressure fluctuation, on the other hand, it shortened the period of pressure fluctuation, corresponding to the increase in main frequency. 4) Between 10% and 20% hydrogen content, the first-order mode was a volume oscillation, which was identical to the main frequency of the whole oscillation, and the second-order mode was an axial oscillation, which was twice the main frequency of the oscillation. With the increase of hydrogen content, the main POD modes switched from the axial mode to flame interaction. [Conclusions] The oscillation conditions and the instability characteristics of the hydrogen-containing fuel were obtained via data analysis. The experimental results could be used to master the mechanism of combustion instability and provide a reference for developing control technology for combustion instability.
  • Extinction characteristics of premixed flames of typical hydrogen-rich fuel gas
    ZOU Jun, LI Zhaoxing, ZHANG Hai, Lü Junfu, ZHANG Yang
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 585-593. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.009
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Adaptability of fuel is an essential design requirement of advanced gas turbines. Hydrogen-rich fuel gas can be obtained from a variety of sources, and its use will be an important part of the future development of gas turbines. When using gas turbine combustion technology to burn hydrogen-rich fuel, if the flame goes out, it will lead to unsafe equipment. As a result, the extinction of turbulent hydrogen-rich fuel flame is the key problem when we design gas turbine combustors. [Methods] In this study, the optimized experimental approach and numerical simulation method of counterflow flame are used to compare the extinction strain rate of two typical hydrogen-rich fuel gases under laminar and turbulent combustion conditions, and the main reasons for the difference were examined. Two typical hydrogen-rich fuel gases used in this study are called FA and FB in this paper. As far as FA is concerned, the CO component ratio of fuel is higher, the dilution ratio is lower, and the calorific value is significantly higher. FA can be considered as the typical hydrogen-rich synthetic gas obtained from entrained flow coal gasification, and FB can be regarded as the typical hydrogen-rich synthetic gas obtained from fluidized bed coal gasification, both of which have certain representative significance. The upper nozzle of the counterflow flame produces nitrogen, while the lower nozzle produces premixed fuel with varying equivalence ratios. The equivalence ratio covers a range of 0.4-1.0. The gas temperature at the nozzle outlet is 300 K. The particle image velocimetry (PIV) system is used to obtain the velocity information of the flow field at the nozzle outlet. The turbulent transport model is added to the OPPDIF code for numerical simulation. [Results] The results demonstrated that, within the range of working conditions studied, the numerical simulation method used in this paper could well predict the extinction strain rate of laminar and turbulent flames. The difference between experimental and simulation results was less than ±10% for a laminar counterflow flame and ±40% for a turbulent counterflow flame. Due to the instability of the turbulent flow field, the measurement of the extinction strain rate fluctuated greatly during the turbulent combustion experiment, and the error bar was greater than that of the laminar combustion experiment. [Conclusions] Under laminar combustion conditions, hydrogen-rich fuel gas with a higher mole fraction of active radicals such as H, O, and OH in the flame front has a higher reaction rate and heat release rate of key chemical reactions, so it can resist higher flame stretching deformation. With the increase of the equivalence ratio, the extinction strain rate indicates an upward trend. Turbulence not only improves the mixing of active groups and reactants, thereby improving the reaction, which increases the rate of key chemical reactions and heat release in the reaction area, but it also improves heat transfer of the flame from inside to outside, resulting in a lower internal temperature of the flame.
  • Experiment on the autoignition characteristics of hydrogen and acetylene jets in a turbulent hot coflow
    LIU Guijun, LIU Jiayue, ZHANG Yang, WU Yuxin
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 594-602. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.018
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Hydrogen fuel gas turbine is the key equipment for large-scale hydrogen fuel power generation toward realizing the goal of carbon neutrality. Lean premixed combustion is an important technology for reducing the NOx emissions of modern gas turbines. However, compared with those of traditional hydrocarbon fuels, the high flame-propagation speed, wide flammability limit, and extremely low ignition energy of hydrogen increase the risk of autoignition and flashback in the premixed duct. Besides, the high mass-diffusion rate and flame-propagation speed of hydrogen in the sequential combustor with the second (reheat) stage of autoignition-stabilized flame render the flame stabilization mechanism is different from that of traditional hydrocarbon fuels. This study aims to understand the difference between hydrogen and acetylene as a hydrocarbon fuel in turbulent hot coflow in terms of autoignition type, flame structure, and stabilization mechanism. [Methods] A jet-in-coflow burner is used to conduct autoignition experiments. Based on our previous studies, acetylene, as an important small-molecule hydrocarbon, is selected as a hydrocarbon fuel for comparison with hydrogen. The fuel jet is injected into the hot coflow air through the fuel injection tube installed at the center axis of the burner. After being heated via the electric preheater, the compressed air flowed into the quartz tube to form a turbulent hot coflow exceeding the fuel ignition temperature. The photographs and OH chemiluminescence images of autoignition are obtained using a digital camera and an intensified charge-coupled device camera. The liftoff height, defined as the distance between the fuel injector exit and flame base (autoignition location), is a crucial parameter determining autoignition characteristics that strongly correlates with the flame stabilization mechanism. A ruler is mounted parallel to the quartz tube as a reference to measure the average liftoff height by comparing the autoignition location with the ruler mark. [Results] The results showed that with decreasing fuel mole fraction or increasing fuel jet velocity, the autoignited flame type of hydrogen and acetylene changed from the attached flame to random spots. Compared with that of acetylene, the flame zone of hydrogen random spots with light red was more compact. For the OH chemiluminescence-based flame structure, which was affected by the high mass-diffusion rate and flame-propagation speed of hydrogen, the independent autoignition points of hydrogen were undetectable. The flame zone became a continuous region. The liftoff height of hydrogen random spots was less sensitive to the fuel jet velocity than that of acetylene. The volatility of the hydrogen random point location was weak. Moreover, based on the mixing-strain model in the previous study, we found that the acetylene flame was mainly stabilized via autoignition kinetics. However, autoignition kinetics and flame propagation jointly determined the flame stabilization of hydrogen. The correlation of the liftoff height of hydrogen further indicated that flame propagation played a key role in flame stabilization. [Conclusions] In conclusion, this study, through experiments, reveals the difference in autoignition characteristics between hydrogen and acetylene in turbulent hot coflow, demonstrates the importance of flame propagation in stabilizing the hydrogen-autoignited flame, and modifies the liftoff height prediction model.
  • Single droplet ignition and combustion characteristics of multi-component heavy oil
    CHEN Jian, ZHANG Yang, ZHANG Hai
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 603-611. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.037
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Heavy oil is one of the heavy end products of the petroleum refining process and has high energy density, low price, and poor ignition and combustion performance. There is an increasing demand for heavy oil and substantial variation in the composition of heavy oil provided by different suppliers. Understanding the combustion characteristics of heavy oil, especially those related to ignition and burnout, is critical to safely use and effectively tune and control the combustion of heavy oil. The development of heavy oil burners can yield significant insights into this endeavor. The combustion of heavy oil in the furnace is essentially the combustion of a large number of liquid droplets. The study of the ignition and b[KG-*4]urnout characteristics of a single liquid droplet can lead to a better understanding of the combustion of heavy oil. [Methods] The composition of multi-component heavy oil is analyzed, and the single drop experiment is carried out. Considering the need to measure the temperature at the droplet center and take pictures of the static droplet during the experiment, the hanging drop method is chosen for studying single droplets of three multi-component heavy oils at different temperatures. Also, experiments are performed in tubular furnaces with different initial diameters to understand the correlation between the ignition and combustion characteristics of a heavy oil droplet and its composition, ambient temperature, and initial diameter. The droplet ignition delay is defined as the duration between the entrance of the droplet into the tubular furnace and its ignition. All experiments are conducted under visible light to enable observation of the variation in the ignition delay of liquid droplets and the time taken for the burn-out of their volatile components. Then the experiment is repeated under a strong backlight to better observe the change in the droplet radius during combustion. The obtained images are binarized by the MATLAB program, and the 256 brightness levels of the original image are converted into pure black or pure white, respectively, through threshold adjustment. The pure black part denotes liquid droplets and thermocouples. The pixel count of the black regions is determined, the region of interest (ROI) is set manually to exclude the thermocouple wire, and the actual area of the droplet is calculated. The characteristic radius is defined as that corresponding to the equivalent circle area equal to that of the droplet image.[Results] The ignition delay and the burnout time of a heavy oil droplet were found to decrease with increasing temperature and increase with increasing initial diameter. However, given that these variations were not directly related to the proportion of the heavy components of the droplet, it was unreasonable to try to correlate the proportion of the heavy component with the combustion performance of the droplet. Due to the complex mix of components with different boiling points, the expansion behavior of the heavy oil droplets as a function of time during combustion strongly influences their combustion characteristics. The thermogravimetric-mass spectrometry (TG-MS) characterization could explain the frequency and amplitude of expansion. The expansion times of the three samples were consistent with the number of pyrolysis peaks in TG-MS results, and the expansion amplitude was positively correlated with the height of the pyrolysis peaks. [Conclusions] The multi-component heavy oil droplets are found to undergo micro-explosions during combustion due to the complexity of their composition and large differences in the volatilities of the individual components. The intensity of micro-explosion during droplet combustion is defined. It is found to be positively correlated with the proportion of heavy components, the initial diameter of the droplets, and the ambient temperature.
  • Automatic generation method of a chemical reactor network for predicting NOx emissions
    GAO Qiaodong, LEI Fulin, ZHANG Zhedian
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 612-622. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.011
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] In the design and development of combustion chambers, fast and reliable emission prediction tools are needed. One of the most extensively utilized technologies today is computational fluid dynamics (CFD) combined with a chemical reactor network (CRN). The manual division of the chemical reactor network is complex, and the calculation of chemical reactor parameters is rough. It can quickly and efficiently divide, build, and solve the CRN to estimate NOx emissions based on the CFD calculation results by developing an automatic generation method for the chemical reactor network. [Methods] The program for the automatic generation of CRN is developed based on Python language, and Cantera is integrated to solve the CRN. The CFD results are obtained by numerically simulating six working conditions of a lean premixed burner and five working conditions of a micromix burner. The CRN automatic generation program divides the combustion chamber domain into different reaction regions based on the division criteria of temperature field, velocity field, and geometric parameters. Meanwhile, when the cells are clustered, the chemical reactor parameters and mass flow rates between the chemical reactors are calculated. The different regions are replaced by reactors, such as the perfectly stirred reactor, and linked by flow controllers. The NOx emissions are obtained by solving the CRN through Cantera and compared with the experimental values. [Results] The parameters of the CRN could be accurately calculated by post-processing the CFD results with the CRN automatic generation program. Under different conditions in the same combustion chamber, the cells could be classified, and the corresponding CRN structure could be generated again by changing the division criteria. The temperature and pressure calculated by the volume-weighted average method and the mass-weighted average method differed in some reactors. However, the NOx values predicted by the two methods were basically identical. The CFD-CRN method predicted NOx emissions more accurately than the Fluent NOx post-processing method. CFD-CRN had a maximum forecast error of 32%, while Fluent NOx post-processing had a maximum prediction error of 96%. The greatest errors in the NOx forecast results of CRN models with different CFD grid numbers and reactor numbers were 5% and 2%, respectively, based on the premise of selecting appropriate division criteria to reasonably build a CRN. The CRN template could be used to predict the NOx emission under nearby working conditions within the acceptable error range. In the lean premixed burner, when heat loss was allocated to the wall area, the NOx values were generally higher than when it was allocated to different reactors according to volume weight. However, the predicted results of the two allocation methods were opposite in the micromix burner. [Conclusions] The CRN automatic generation program may automate the CRN’s division, construction, and solution. Taken temperature, velocity and geometric parameters as the criteria, it can generate well-structured CRN. With fewer grids and reactors, the CRN model can estimate NOx emission accurately. When the combustion temperature is high, considering heat loss and distributing it to different chemical reactors can improve the accuracy of the NOx prediction substantially. The same CRN model may be reused again to accurately predict NOx emissions under similar working conditions.
  • Comparison of NOx numerical models for methane/air combustion simulations
    SUN Jihao, LUO Shaowen, ZHAO Ningbo, YANG Huiling, ZHENG Hongtao
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 623-632. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.022
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Correct usage of models for NOx combustion simulations can considerably reduce the computational time compared to directly coupling the detailed chemical mechanisms. Several NOx models are available: the NOx postprocessing model, decoupled detailed mechanism model, and adding NOx transport equations in flamelet generated manifold (FGM) model. However, their differences and applicability remain unclear, so choosing a model for a particular work is challenging. Therefore, the differences and applicability of these models must be verified under different situations, particularly for diffusion combustion, premixed combustion, and real combustors (partly premixed combustion). [Methods] In this study, numerical simulations were performed on a diffusion jet flame (Sandia flame D), premixed swirl flame (Cambridge swirl flame SW3), and partly premixed flame (an industrial gas turbine combustor) to thoroughly understand the differences and applicability of these three models. The turbulence and combustion models were the realizable standard k-ε model and the flamelet-generated manifold model, respectively. The turbulence and combustion models were verified against the experimental results; furthermore, the NOx (including NO, NO2, and N2O) distribution and formation characteristics, as well as NOx emissions, were compared and discussed with the experimental results. For the NOx postprocessing model, O and OH radicals were treated as partial equilibrium consumption, and the turbulence-combustion interaction was modeled as β-PDF (β-probability density function, PDF) consumption. For the decoupled detailed mechanism model, the species (excluding NOx), pressure, velocity, and temperature distributions were obtained using numerical simulation and held constant, and then NOx chemistry was solved. For the added NOx transport equation model, the three NOx transport equations of NO, NO2, and N2O were added to the PDF table to calculate NOx (including NO, NO2, and N2O). During the computation of NOx transport equations, only NO, NO2, and N2O were solved, and the remaining species, such as O, OH, and CH, were directly read from the PDF table. [Results] (1) For diffusion combustion, premixed combustion, and the gas turbine combustor, NO accouned for more than 95.00% of the total NOx behind the flame (at the burned-out zone), the amount of N2O was relatively small, and the amount of NO2 was negligible. (2) The NOx postprocessing model could accurately simulate the NOx formation position near the flame (at the reacting zone) and the NOx generation rate behind the flame; however, this method underestimated the NOx concentration and NOx generation rate at the flame position. Moreover, the NOx postprocessing model couldn't reproduce the phenomenon of the initial increase in the N2O concentration near the flame and then its decrease. (3) The added NOx transport equation model showd the best accuracy for the NOx generation position, NOx concentration, and NOx formation rate near the flame, but it underestimated the NOx generation rate behind the flame. (4) The decoupled detailed mechanism model showd the worst accuracy in NOx simulation and couldn't correctly predict the NOx formation characteristics of the three studied cases. [Conclusions] The decoupled detailed mechanism model may not be suitable for NOx simulation under some conditions. To capture NOx formation and distribution characteristics, the postprocessing model and added NOx transport equation model can be used. However, the postprocessing model may not provide quantitative results, particularly in diffusion flames. The added NOx transport equation model may be suitable under most conditions.
  • Surrogate model of combustor flow mixing process
    GENG Junjie, WANG Xingjian, LI Jialu, FEI Teng, QI Haiying
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 633-641. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.030
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Numerical simulation of a gas turbine combustor is an important step in its design process. Due to the complexity of the physical and chemical processes, the calculation cost is high. The calculation cost can be reduced by constructing surrogate model of combustor. This paper focuses on the key steps in the construction of a surrogate model suitable for cold gas flow and the mixing process of combustor. Furthermore, this paper proposes a surrogate model for the central nozzle and the subsequent combustor space of a heavy-duty gas turbine. [Methods] The construction of the surrogate model includes several key steps: design of experiments (DOE), numerical simulation, dimensionality reduction, and an interpolation process. Two parameters are selected as the input parameters for the surrogate model: The fuel mass flow rate Gf and the combustor inlet air pressure p2. Latin hypercube sampling is used in the DOE to determine 12 operating conditions for computational fluid dynamics (CFD) simulations, and the results are used to build the surrogate model. Proper orthogonal decomposition is used for dimensionality reduction, wherein a set of basis functions and corresponding coefficients are extracted. The basis functions reflects the main characteristics of the combustor flow field. Moreover, the data dimensionality is reduced from the number of grid nodes of the combustor to the number of basis functions, which do not exceed the number of operating conditions. The Kriging model is used to interpolate the coefficients of the basis function with the input parameters of the surrogate model. Four verification conditions are set up to determine the accuracy of the surrogate model through a comparison of the surrogate model results with the CFD simulation results. The outlet cross section of the central nozzle and the longitudinal section of the combustor are selected to compare multiple key parameter distributions, including axial velocity, radial velocity, tangential velocity, CH4 concentration, turbulent kinetic energy, and pressure. The vector operations are used to compare the distributions of various parameters, which can simultaneously reflect the differences in the numerical and spatial distributions of various parameters. [Results] The results showed that the error in most parameters was ~1%. The results also revealed that the construction method of the surrogate model could be applied to cold high-speed and high-turbulence strong rotational flow and fuel/air mixing. The accuracy was higher than the international average level, and the application criteria of the construction method were proposed. The influences of interpolation methods, sample numbers, and basis function numbers on the accuracy of the surrogate model were analyzed. The accuracy of SM was higher than extrapolation. Increasing the number of sample operations and basis functions could improve the accuracy of the surrogate model, but also increased the computational cost. [Conclusions] The SM construction method (POD & Kriging) is suitable for the cold gas flow and mixing process in the combustor. The paper lays the foundation for subsequent research on the construction method of combustor SM, which includes combustion reactions and geometric structure changes.
  • The applicability study of Gao-Yong turbulence model to boundary layer transitions
    SUN Yifan, ZHU Wei, WU Yuxin, QI Haiying
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 642-648. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.010
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Transition is one of the most significant progressions in fluid mechanics. Accurate boundary layer transition prediction is also essential for a complete understanding of the production of turbulent boundary layers and the optimization of airfoil shapes for industrial applications. Turning turbulence models can accurately forecast transition by including an empirical turning criterion, a low Reynolds number viscous flow field characteristic, or an intermittency component for flow in the transition zone. As for the conventional turbulence models using Reynolds averaging method, such as the k-ε and the k-ω models, accurately forecasting this phenomenon is tough since the transition process is poorly characterized or described. [Methods] A functional correction relation for the G-Y model is suggested and put to the test in the calculations to increase its precision and usefulness for forecasting boundary layer development and transitional features across flat plates. The coefficient of substance Cs had the problem of discontinuous distribution and intermittent values in the original model. This work provided a correlation formula between Cs and the drift vector Reynolds number ReT to solve this problem. The drift vector Reynolds number ReT is built using the G-Y model's drift velocity and vector to describe the strength of local turbulent pulsations. With the aid of physical understanding, the Cs-ReT relationship is produced, resulting in a continuous distribution of the real degree coefficient with ReT variation, close to 0 at the location of mild turbulent pulsation and close to 2/3 at the location of strong turbulent pulsation. The turbulent viscosity is bound by the size of the real degree coefficient at the point of weak turbulent pulsation by solving the discrete distribution's initial problem in this manner. [Results] A solver for the modified G-Y turbulence model was made using the open-source program OpenFOAM. The CFD numerical simulations and validation were then performed for the tests involving the T3A and T3B transitions as well as the conventional flat plate boundary layer. The G-Y model's computational findings showed that: (1) The formula of Cs and ReT considerably increased the G-Y model's accuracy for boundary layer instances and gave the model the ability to forecast transitions. The G-Y model accurately predicted the transitional positions of the two boundary layer experiments, T3A and T3B. (2) Results of the modified G-Y model were in good agreement with the experimental and theoretical values for the distribution of surface friction coefficients before and after the turn. The relative error of the friction coefficients in the segment just entering turbulence was only 3%. (3) The G-Y model accurately replicated the transition from laminar to turbulent flow, which caused the velocity profile of the boundary layer to change from fullness to loss. [Conclusions] The findings show that the improved G-Y model has clear advantages over the established k-ω model in terms of its ability to accurately simulate the boundary layer transition process of a flat plate. This model can be used to investigate the properties of boundary layer transitions in greater depth.
  • Prediction of the pollutant generation of a natural gas-powered coaxial staged combustor
    SUN Jihao, SONG Ying, SHI Yunjiao, ZHAO Ningbo, ZHENG Hongtao
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 649-659. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.014
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] As pollutant emission is an important technical index of gas turbines, pollutant emission prediction has become one of the active research topics. However, the irregular strong turbulent combustion process in the combustion chamber of natural gas turbines causes chaotic pollutant generation, and the characteristics of low-emission combustion are extremely complex. The influence law of various geometric factors on pollutant generation characteristics is not clear. Moreover, the common pollutant prediction methods have certain limitations. For example, the numerical simulation method needs to be combined with a complex dynamic mechanism, resulting in a long calculation time. Therefore, this paper proposes to apply a neural network to the prediction of gas turbine pollutant emissions and develop a new method for the rapid and accurate prediction of pollutant emissions. [Methods] Computational fluid dynamics-based numerical simulation was used to study the influence of typical structural factors, such as the number of first-stage swirling flow, the number of second-stage swirling flow, and the fractional area ratio, on pollutant generation in the gas turbine combustion chamber, and to elucidate the variation trends of pollutant generation for different structures. The data were divided into a training set and a test set. Four structural parameters, namely the first-level swirl number, the second-level swirl number, the graded area ratio, and the graded axial distance of the combustion chamber head, were defined as input variables; the NOx and CO emissions at the combustion chamber outlet were defined as output variables for neural network training calculation; and then the radical basis function (RBF) neural network prediction model was established. The model structure was determined as 4-22-2. [Results] The results showed that for the studied coaxial graded combustor, the increase in the swirl number will lead to the increased and backward movement of the vortex core in the return zone, and the increase of the graded area ratio will lead to an increase in the equivalent ratio in the center of the return zone, which will increase the intensity of chemical reactions in the combustor, the maximum temperature, and the NOx emission. The CO emission in the combustion chamber was not sensitive to the typical structural parameters of the combustion chamber head, and the CO emission at the combustion chamber outlet exhibited little change with the variation of different structural parameters, such as swirl number, fractional area ratio, and fractional axial distance. The established combustion chamber emission RBF neural network prediction model could accurately and rapidly predict the combustion chamber outlet emission under different structural parameters. The maximum prediction error of NOx emission was 12.28%, and the average error was 4.58%; the maximum prediction error of CO emission was 2.75%, and the average error was 0.97%. [Conclusion] In this study, the characteristics of gas turbine pollutant generation are analyzed via numerical simulation, and the results prove that the neural network prediction model can effectively predict the characteristics of gas turbine pollution emission with good feasibility and high accuracy.
  • Optimization investigation of low NOx emissions of natural gas radially staged combustor
    TIAN Yuan, GENG Junjie, SUN Yifan, QI Haiying
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 660-669. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.032
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] To solve the problem of excessive NOx emissions of the existing natural gas radially staged combustor, the design principle of parameter matching between multiple nozzles at the head of the combustion chamber is explored, and an optimized prediction formula for NOx emission performance of the combustor is obtained. [Methods] This study adopts numerical simulation and experimental data analysis methods. The different loads and diameters of the dilution holes of the combustor are numerically simulated. Through the numerical analysis of the nonuniformity of the premix and the equivalent ratio of the original design, a structural improvement strategy and an optimization scheme to reduce the dilution holes' diameter were proposed. The influence of the improvement in the dilution holes' diameter on NOx emissions and that of the modification scheme on the total pressure loss and other performance parameters were investigated. The relationship between parameter matching of multiple nozzles and NOx emission under different combustion modes was investigated. Because of the deficiency of the current NOx emission prediction formula, based on the experimental and numerical data, the prediction formula of NOx emission under partial load is improved. [Results] Results showed that the original low-emission design of the combustor was unreasonable, and the NOx emission value was six times that of the limit value of 41 mg/Nm3 at 15% O2. The standard scheme of changing the dilution holes' diameter was reasonable and feasible, and the dilution holes' diameter Djet=44.45 mm can ensure that the standard of NOx was determined and verified by the experiment. The resulting changed in the total pressure loss and other performance parameters were within the acceptable range. In the proposed dilution holes' diameter modification scheme, the change in the dilution holes' area from large to small was 33.1%, 47.1%, and 59.5%, which exceeded 30% of the original design, and the head structure was not redesigned. A warning that the combustor may face the risk of wall overtemperature during long-term operation was proposed. The equivalence ratio and nonuniformity of the premix, two key parameters used to characterize lean premix, were sufficient and necessary conditions to effectively control the NOx emission. The central nozzle was the main factor in determining the emission level. Because the central area was in the high equivalent ratio area, the fuel amount of a single nozzle was high, premix uniformity was poor, and diffusion combustion of heavy-duty fuel occurred. Notably, the low-emission design criterion of the multi-nozzle matching relationship in the natural gas radially staged combustor investigated in this study was that the fuel amount of the single annular nozzle was equal to the fuel amount of the central area in combustion mode 2. [Conclusions] Based on the experimental and numerical data in this study, an improved NOx emission prediction formula that can accurately reflect the partial load behavior of this type of combustor is proposed. Compared with the test data of the two combustors, the prediction formula has sufficient accuracy, and the maximum relative error is 8.73%. This research provides a theoretical basis for the subsequent test debugging and design improvement of the combustor.
  • Modeling and simulation of two-phase turbulent combustion in aeroengine combustors
    MO Yi, CHEN Fan, XU Xiaoyan, JIAO Zhe, WEI Gang, LIN Hongjun, XIAO Wei, WANG Fang, REN Zhuyin
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 670-680. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.005
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] As the energy-producing component of aeroengines, the combustor is the core area of fuel atomization, oil and gas mixing, and chemical reaction. Its design directly affects the overall performance of the engine. The structure of an aeroengine combustor is complex, allowing for a series of complicated physical and chemical processes. [Methods] The application of numerical simulation is of great significance in shortening the development cycle of the combustor while reducing test experiments and risks in design. In this paper, we conduct a bottom-up study on framework design, model integration, software development, test validation, and engineering application of the self-developed software platform. First, we design a hierarchical simulation software by analyzing the common numerical algorithm of an individual physical model and optimizing the secondary development interface of the model code. The software framework can be divided into three levels from bottom to top: unstructured grid high-performance parallel programming framework, particle-fluid computing layer, and advanced physical models and methods. The software framework has a reasonable data structure and highly scalable function interface, which guarantees the independence and high maintainability of each model and supports the R&D team in realizing the efficient integration of different types of physical models. Second, for the complex two-phase turbulent combustion process in the combustor, ten physical models suitable for simulating engine combustors, such as fuel atomization, wall oil film, evaporation, and turbulent combustion models, are integrated. Four hierarchical test cases of three-stage swirl, gas-phase swirl, simple cylinder and model combustor configurations are constructed, and the coupling consistency of multiple models is studied and improved. Based on the work related to the framework, model, and validation, a parallel adaptive unstructured grid combustor two-phase turbulent combustion numerical simulation software (CBTLES), which can run efficiently on modern mainstream high-performance computers, was developed. Finally, to test the engineering applicability of CBTLES, two-phase turbulent combustion simulation in the annular main combustor and afterburner of a large turbofan engine are conducted. [Results] The simulation results showed that: 1) thousand-core parallel efficiency reached 104.50%, while the ten thousand-core parallel efficiency reached 70.92%, indicating that CBTLES has good parallel scalability for hundreds of millions of grid-scale annular combustor cases. 2) Qualitative simulation results of unsteady two-phase turbulent combustion were consistent with physical phenomena, indicating that CBTLES has engineering coupling simulation ability with a typical two-phase physical model. 3) With typical working conditions, the quantitative errors of the total pressure recovery coefficient and outlet temperature distribution coefficient of the main combustor were 1.2% and 9.7%, respectively, while the errors of the total pressure recovery coefficient and average outlet temperature of the afterburner were 5.6% and 0.9%, respectively, revealing that CBTLES has acceptable engineering simulation accuracy. Generally, CBTLES realizes a breakthrough from 0 to 1 through framework design, model integration, software development, test verification, and engineering application in this paper. [Conclusions] The engineering simulation results of a typical annular main combustor and afterburner show that the simulation efficiency, function, and accuracy of CBTLES meet practical engineering requirements. Simultaneously, it also reveals that physical models integrated into CBTLES realize the key transformation from basic theory to engineering applications.
  • A prediction method for wall cooling efficiency of combustor chamber based on neural network
    LIU Jiangfan, GE Bing, LI Shanshan, LU Xiang
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 681-690. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.019
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] The effusion cooling technique is applied to hot sections in the modern heavy-duty gas turbine combustor, for example, the liner provides efficient protection from high-temperature gas using a small amount of cooling air. However, establishing a theoretical model or empirical correlations for regression and prediction of wall cooling efficiency is difficult. Because it is affected by many factors, including cooling hole type, patterns, and other flow parameters. The prediction methods based on computational fluid dynamics (CFD) techniques also have the disadvantages of long-term consumption and resource occupation. This paper proposes a prediction method based on artificial neural networks, namely back propagation (BP) and radial basis function (RBF) networks, that reach high accuracy for field reconstruction and cooling efficiency prediction. [Methods] This study collects simulation data from the design process of an industrial combustor and analyzed all dimensions of the wall adiabatic film cooling efficiency data set from the working conditions. Artificial feature selection, mini-batch, and normalization are used to preprocess the data. Important feature regions A and B are manually extracted by calculating the standard deviations of the cooling efficiency matrix considering different cooling air pressure loss conditions combined with CFD analysis. Then, the training set, the validation set, and the test set are divided based on cross-validation. For the field reconstruction, the BP network and the RBF network are selected. For its fast convergence and high precision, the BP network adopts the L-M algorithm. For the prediction, the BP network and generalized regression neural network (GRNN) are selected. The BP network adopts a Bayesian regularization algorithm to avoid overfitting. The GRNN network outperforms the RBF network for prediction and also takes less time for grid searching. For regression, mean square error (mse) and network prediction accuracy (accu) are defined to measure variance and deviation, respectively. [Results] The results showed that: 1) Through artificial feature selection and mini-batch, the samples in region A were compressed by ten times while perfectly extracting cooling efficiency characteristics. 2) For the field reconstruction, under the non-reactive condition 3, the field reconstruction accuracy of the optimized BP and RBF networks reached 92.2% and 95.5%, respectively. Under the reactive condition 10, the same networks reached 90.5% and 92.0% respectively. Compared with the BP network, the RBF network had better global approximation ability than the BP network. 3) For the prediction, under the reactive condition 10, the prediction accuracy of the BP and GRNN networks reached 88.3% and 86.8%, respectively, both lower than the field reconstruction accuracies. [Conclusions] The rapid field reconstruction and high prediction for wall adiabatic film cooling efficiency are realized by proposing a universal data preprocessing technique for the combustor wall cooling efficiency data set and an optimization technique for the BP and RBF networks. In summary, the field reconstruction accuracy reaches above 90%, and the prediction accuracy reaches above 85%, meeting the actual engineering requirements in the modern industrial gas turbine combustor design.
  • Investigation of combustor-forced instability under different cooling boundaries
    LU Xiang, JIA Yuliang, JI Yongbin, GE Bing, ZANG Shusheng
    Journal of Tsinghua University(Science and Technology). 2023, 63(4): 691-696. https://doi.org/10.16511/j.cnki.qhdxxb.2023.25.021
    Abstract ( ) Download PDF ( ) HTML ( )   Knowledge map   Save
    [Objective] Suppression of combustion oscillations is critical to the design of aero engines and gas turbine combustion chambers to avoid materials and structural damage caused by structural resonance and to prevent the reduction of combustion effectiveness. Previous research has shown that adjustments in common wall cooling methods, such as effusion cooling, can affect the acoustic properties of combustion chambers and can be considered in the design to inhibit instability. However, these previous studies mainly operate under non-reaction conditions, ignoring the mutual interference between flow, flame heat release, coolant flow, and sound oscillation. Therefore, the results deviate from actual engine operating states and need to be verified under reaction conditions. [Methods] This study investigated the influence of effusion and impingement cooling boundaries on forced oscillation characteristics. A swirl-stabilized model combustor test rig with a boundary of effusion/impingement cooling structure was built, and its acoustic characteristics are measured by microphone sensors. The influence of effusion and impingement cooling boundaries on the pressure pulsation with an excitation frequency of 80～210 Hz of a model combustion chamber is compared using experimental and acoustic numerical methods under the same reacting conditions of a ratio equivalent to 0.61. In the acoustic simulations, we examine the acoustic principle of why the effusion cooling boundary has a better performance in sound absorption compared with the impingement cooling boundary. In addition, the blowing ratio of effusion cooling is then changed to 0~8 to determine the influence of coolant mass flow change on its sound absorption ability. [Results] (1) Experimental and numerical investigations indicated that the effusion cooling boundary has better sound absorption ability in the pressure oscillation than the impingement cooling boundary, and the amplitude of sound pressure pulsation in the combustion chamber decreased at the largest percent of 37.7% at an excitation frequency of 110 Hz. (2) Changes in the upstream excitation frequency revealed its relationship to the sound absorption capacity of the effusion cooling boundary. Within a pulsation frequency band of 90~170 Hz, the absorption ability of the effusion cooling structure can lead to lower pressure pulsation amplitude. However, when the excitation frequency is <90 Hz or >170 Hz, the sound absorption capacity sharply decreased, and the pressure oscillation amplitude of the effusion-cooled combustor became similar to the impingement cooling boundary. (3) Changes in the blowing ratio (BR) in the effusion cooling boundary are proved to influence its sound absorption capacity. The effusion cooling boundary achieved the best sound absorption ability when BR is 0, owing to the pulsation energy consumption ability of the tiny holes. However, with the increase in BR, the pressure oscillation amplitude increased because of the noise caused by the coolant passing through the tiny holes, while the sound transmission capacity decreased owing to increased back pressure. [Conclusions] The effusion cooling boundary shows a better performance in suppressing combustion instability, especially in the frequency band of 90~170 Hz, and can achieve better sound absorption ability through its BR adjustment.
Postal Code: 100084
Tel: 010-62788108/62771385/62792976/62792994　
E-mail：xuebaost@tsinghua.edu.cn
Powered by Beijing Magtech Co. Ltd.
Visited
    Total visitors:
    Visitors of today:
    Now online:
　
Copyright © Journal of Tsinghua University(Science and Technology)