Abstract
Theoretical framework and high-fidelity direct numerical simulations (DNS) on preignition and knock encountered in internal combustion (IC) engines are briefly reviewed in this chapter. The theoretical framework is presented first, and systematically followed by one-dimensional and multi-dimensional simulations. A number of high-fidelity simulations under realistic IC-engine conditions were used to demonstrate the super-knock mechanism in the presence of the complex chemistry-turbulence interaction. The ratio of ignition delay time to eddy-turnover time, \(\tau _\mathrm{ig}/\tau _t\), and the ratio of the most energetic length scale of turbulence and temperature fluctuations, \(l_t/l_T\), affect the knock intensities are discussed. Different statistical metrics extracted from the multi-dimensional simulations to predict knock intensity are also presented. The quantitative determination of strong and weak ignition modes using the predicted Sa-based metrics is also cover in this chapter.
Access provided by Autonomous University of Puebla. Download chapter PDF
Similar content being viewed by others
Keywords
1 Introduction
Modern downsized and boosted internal combustion (IC) engines provide higher thermodynamic efficiencies and cleaner combustion. However, these engines operated under extreme conditions of high load suffer from a higher possibility of abnormal preignition process, a premature auto-ignition induced by hot spots leading to ignition front propagation [1,2,3,4,5,6,7]. The mechanism and the source of such irregular preignition events are still not well-understood [8]. The early flames developed from preignition kernels increase the temperature and pressure of the end-gas that may induce a stronger end-gas auto-ignition, leading to extremely high-knock intensities, and even super-knock under high-load and/or high-temperature conditions. As representatively shown in Fig. 1 [6], super-knock cycle (S2) is characterized by high-pressure amplitudes and intensities due to localized detonation development that may lead to severe structural damage to the engine components. Therefore, a reliable criterion to predict irregular combustion phenomena is of critical importance [1, 3, 4, 6, 9,10,11,12].
A large spectrum of turbulence length and time scales combined with extremely short time scales of complex combustion process in detonation simulations pose a significant challenge in reproducing the phenomena using computational simulations, which require extremely fine spatial and temporal resolutions to capture the highly intermittent localized characteristics. This chapter reviews our recent efforts to investigate the fundamental issues of the knock events using high-fidelity direct numerical simulations (DNS) with the capability of fully resolving all temporal and spatial scales and the complex interaction of thermochemistry and turbulence [3, 4, 13, 14]. First-principle DNS is capable of unraveling the complex interplay between turbulence and chemical reactions to provide a better understanding of the mechanism of detonation development encountered in modern combustion devices under extreme high-load operating conditions, and to develop a reliable predictive model for real-world industrial applications [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31].
In the following, a theoretical framework on super-knock development encountered in internal combustion (IC) engines is briefly reviewed, followed by high-fidelity DNSs in the subsequent sections. A systematically parametric set of two-dimensional (2-D) and three-dimensional (3D) DNS simulations are conducted at engine-relevant conditions [1,2,3,4, 32]. Effects of various parameters such as length and time scales of turbulence and scalar fluctuations on the occurrence and intensity of detonation are examined and a scaling relation for knock prediction is proposed [33,34,35,36,37,38,39,40,41,42,43].
2 Developing Detonation Regime Diagram
2.1 Theoretical Background
2.2 Zeldovich’s Theory on Reactivity Gradient
According to Zeldovich’s theory [44], a spontaneous ignition front is defined as a successive ignition events in the presence of spatial reactivity gradients. Its propagation speed, \(S_\mathrm{sp}\mathrm{{}}\), is determined as
where \(\tau _\mathrm{ig}\) is the ignition delay time of the local reactant mixture as schematically shown in Fig. 2, defined as the time to reach the maximum heat release as illustrated in Fig. 3.
Based on the relative ratio of \(S_\mathrm{sp}\) to the laminar flame speed, \(S_\mathrm{sp}/S_L\), and to the sound speed, \(S_\mathrm{sp}/a\), of the local mixture as illustrated in Figs. 2 and 3, Zeldovich [44] classified four different ignition regimes of a reaction front propagation in a combustible gaseous mixture:
-
1.
weak ignition: In the presence of large spatial variations in \(\tau _\mathrm{ig}\), \(S_\mathrm{sp} \sim S_L\), the local ignition establishes a slow-speed thin-front deflagration (see Figs. 2c and 3).
-
2.
spontaneous ignition: For a mixture with a smaller level of \(\tau _\mathrm{ig}\) variations, \(S_\mathrm{sp} \gg a\), the mixture ignites spontaneously and \(S_\mathrm{sp}\mathrm{{}}\) exhibits a rapid ignition front propagation (see Figs. 2a and 3).
-
3.
detonation: If \(S_\mathrm{sp}\mathrm{{}}\) is sufficiently large and comparable to the sound speed, \(S_\mathrm{sp} \sim a\) as visually shown in Fig. 2b, the combustion wave couple with the acoustic wave may lead to detonation development characterized by extremely high-peak amplitudes of pressure oscillations.
-
4.
homogeneous ignition: In the asymptotic limit of zero gradient, \(S_\mathrm{sp}\mathrm{{}}\) becomes infinity, resulting in the purely homogeneous ignition mode.
2.3 \(\varepsilon \!-\!\xi \) Regime Diagram
Gu and Bradley [45] further mapped different propagation modes into a regime diagram characterized by two key non-dimensional parameters, \(\xi \) and \( \varepsilon \), which are defined as:
where \(\varepsilon \) is defined as the ratio of the residence time, \(r_\mathrm{hs}/a\), of the acoustic wave within the hot spot with a radius, \(r_\mathrm{hs}\), to the excitation time, \(\tau _e\), in which most of the chemical energy is released. \(\tau _e\) is typically defined as the time interval between 5% of the peak heat release rate (HRR) and the peak HRR as schematically illustrated in Fig. 3.
Based on the range of \(\xi \), five different modes of ignition front propagation within a hot spot were identified [45, 46]:
-
1.
\(\xi = 0\) for instant thermal explosion,
-
2.
\( 0< \xi < \xi _l\) for spontaneous (supersonic) ignition where the autoignitive wave is ahead of acoustic wave,
-
3.
\( \xi _l< \xi < \xi _u\) developing and developed detonation, through the thermoacoustic coupling process referred to as the shock wave amplification by coherent energy release (SWACER) mechanism [47, 48]),
-
4.
\( \xi _l \ge \xi < a/S_L\) subsonic autoignitive deflagration where \(S_L\) is the laminar burning velocity,
-
5.
\( \xi _l \ge \xi < a/S_L\) laminar burning deflagration propagating at the speed comparable to the laminar flame speed.
Figure 4 shows the \(\xi \!-\! \varepsilon \) peninsula with the upper and lower limit, \(\xi _u\) and \(\xi _l\), respectively, between which the developing detonation regime is identified for a stoichiometric 50%H\(_2\)–50%CO/air mixture at 5 atm [45, 46]. Note that the upper and lower \(\xi \) limits of the C-shaped curve strongly depend on the initial conditions of \(T_0\), \(P_0\), and \(\phi _0\), fuel types, and the kinetic model used in simulations [49, 50]. In general, the developing detonation regime is expected when \(\xi \) ranges between 1 and O(10) quantity [49, 50]. As representatively shown in Fig. 5, due to the NTC behavior of two-stage ignition fuels (e.g., DME), they are much more prone to detonation development (gray region with roughly \(1< \xi < 10\)) than single-stage ignition fuels (e.g., ethanol) for the same initial conditions of temperature, pressure, equivalence ratio, temperature gradient.
2.4 Quantitative Analysis
Luong et al. [3] proposed three a posteriori statistical metrics to quantitatively assess the knock intensity, defined as follows:
where \(\dot{q}\) is the HRR; \(P_e\), \(P_\mathrm{CJ}\) and \(P_\mathrm{VN}\) correspond to the homogeneous constant-volume equilibrium pressure, the Chapman–Jouguet pressure, and the von Neumann pressure, respectively. The summation is operated over all the cells of the computational domain and integrated over time. As such, \(F_i\) (\(i = {e, CJ, VN}\)) measures the fraction of HRR associated with higher pressures exceeding the corresponding \(P_i\), evaluated at the initial conditions of \(T_0\), \(P_0\), and \(\phi _0\).
According to Fig. 4 [45], the developing detonation regime is bounded by the upper and lower \(\xi \) limits of the C-shaped curve, \(\xi _u\) and \(\xi _l\), respectively, suggesting that the detonation intensity depends strongly on the statistical quantities extracted from the spatial distribution of \(\xi \) such as its volume-average \(\mathrm {\overline{{\xi }}}\), probability density function. To this end, \(F_{D}\), defined as the volume fraction of \(\xi _u< \xi < \xi _l\), is proposed to predict knock intensity:
which serves as an a priori predictive indicator of the subsequent detonation development while \(F_H\) is an a posteriori statistical assessment of knock intensity. Note that both \(F_D\) and \(F_H\) range from zero to unity. \(F_D\) was found to have a strong correlation with the fraction of the total heat release associated with detonation extracted from the full DNS data, \(F_H\) [3]. The mixture is found to be more prone to strong detonation as \(F_D\) approaches unity if \(\varepsilon _p \gtrapprox 1\). For each initial condition, \(F_D\) together with the volume-average, \(\mathrm {\overline{{\xi }}}\), and probability density function, \(P(\xi )\), are computed to quantify the knock intensity [3].
In the next sections, the above predictive criteria are discussed using 1D simulations first, followed by selected 2D and 3D simulations with and without turbulence in order to characterize the spontaneous ignition and detonation development.
2.5 One-Dimensional Detonation Development
The developing detonation \(\varepsilon \!-\!\xi \) diagram [45, 46] has been used in many 1-D studies to investigate the interaction between flame propagation and end-gas auto-ignition, and the effects of boundary conditions, domain size, hot-spot location, hot-spot interaction, and cool flames on subsequent detonation formulation [6, 49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71]. The effects of the bulk-mixture gradient on developing detonation were also studied by Sow et al. [60] and Nogawa and Terashima [72]. In the following, two representative 1-D cases with distinct detonation development characteristics are discussed.
The first case is shown in Fig. 6a features a deflagration to detonation transition (DDT) event where a hot spot is developed into deflagration, gradually consuming the unburned mixture, and finally accelerating to make a transition to detonation. Specifically, the initial temperature profile is designed such that \(\xi \) ranges from 0 to 1000 within the hot spot (see Fig. 6). Figure 7 shows the temporal evolution of the temperature and pressure profiles. For such a strong ignition source, autoignition immediately forms a detonation front within the hot spot length, which is subsequently attenuated quickly upon leaving the hot spot (lines 1–4 in Fig. 7). Due to the low reactivity of the unburned mixture at \(T_0=\) 1000 K, the heat release rate within the reaction zone is not fast enough to sustain the detonation front such that it decays quickly and transits back into deflagration at \(x \sim 10\) mm. As more fuel/air mixture is gradually consumed by deflagration, the temperature and pressure of the remaining unburned mixture ahead of the flame are increased due to the compression heating effect (lines 5–7 in Fig. 7). Due to the increased reactivity of the unburned mixture together with the flame acceleration, the deflagration front transits into detonation at \(x \sim 20\) mm and rapidly consumes the rest of the unburned mixture (lines 8–12 in Fig. 7). Much higher pressure peaks are observed during this period. These processes are typically referred as to deflagration to detonation transition (DDT).
The second case (i.e., \(T_0\) of 1200 K with \(T_\mathrm{peak}\) of 1264 K in Fig. 6) features a direct detonation initiation in which the reaction front and the pressure wave are perfectly coupled to form an extremely high shock wave. The initial hot spot is designed for \(\xi \) to range between 3 and 6 (see Fig. 6). Due to the high reactivity of the unburned mixture with \(T_0\) of 1200 K, the hot spot immediately gives rise to a detonation front within the hot spot length, thereby rapidly consuming the unburned mixture (see Fig. 8). Once the detonation is established, the rapid heat release rate in the reaction zone is able to sustain the steady detonation front that propagates and consumes the rest of the unburned mixture.
The above cases show that the two parameters, \(\xi \) and \(\varepsilon \), within the hot spot and the bulk mixture determines the nature of the early detonation development. Figure 9 shows the compiled results of a series of 1-D simulations for various combinations of these parameters. For a fixed \(T_0\) of 1200 K and \(\phi \) of 1, the 1-D simulations were performed by varying \(T_\mathrm{hs}\) and \(r_\mathrm{hs}\) such that both \(\xi \) and \(\varepsilon \) are varied accordingly. As seen in Fig. 9, the maximum pressure above the equilibrium pressure (i.e., > 1.1\(P_e\)) for \(0.2< \xi < 20\). The peak \(P_{\max }\) occurs at \(\xi \sim 6\) where the perfect coupling between the reaction front and the acoustic wave evolving into a stable detonation shock wave. The range of \(\xi \) was used to determine the volume fraction of \(\xi \) and \(F_D\) to predict the amplitude of knock intensity in multi-dimensional cases.
2.6 2-D and 3-D Simulations of Super-Knock Under Realistic IC-engine Conditions
The basic characterization in 1D problem is now extended to multi-dimensional conditions in the presence of spatial fluctuations in temperature and velocity fields. Parametric 2-D and 3-D simulations were systematically performed to validate the criteria proposed in Fig. 2.4. The distribution of \(\xi \) and \(\varepsilon \) were varied by changing \(T'\) over a wide range from 0.5 to 20 K, and \(l_T\) of 1, 2, 5, and 10 mm. For each initial condition, the full detailed information of spatial distribution of \(\xi \) and the corresponding statistical quantities of \(\xi \) can be computed. Figure 10 shows the spatial distribution of \(\xi \) of four cases with different \(T'\) and \(l_T\) that result in no-knock, mild-knock and super-knock [3].
Figure 11 shows a representative 2-D contour of temperature field with \(T_0\) of 1200 K, \(T'\) of 15 K, and \(l_T\) of 5 mm. According to Peters et al. [53, 73], the temperature field is subdivided into the dissipation elements over which the temperature profile is monotonic, ranging from the peak to the trough of each dissipation element. For isotropic temperature field fluctuations, Peters et al. [53] found \(r_\mathrm{hs}\) comparable with the mean distance of the dissipation elements, \(l_\mathrm{DE}\), and approximated as \(l_\mathrm{DE} \!\approx \! 2\lambda _T\), where \(\lambda _T\) is the Taylor mixing scale [53]. In the multi-dimensional problems, \(l_\mathrm{DE}\) is believed to be a minimum size of a hot spot of unburned reactants required so that it allows a sufficient run-up distance of detonation development [12, 53].
Based on the Taylor mixing scale, Towery et al. [12] proposed a statistical model to evaluate the predictive \(\varepsilon _p\) for highly compressible homogeneous isotropic turbulence fluctuations. \(\varepsilon _p\) is defined as
where a and \(\tau _\mathrm{{exo}}\) are the mass-weighted sound speed and exothermicity time, respectively (see Ref. [12] for details). In Refs. [3, 4], at the the initial conditions of \(T_0 = 1200\) K, \(P_0 = 35\) K, and \(\phi _0 = 1\), \(\varepsilon _p = 1\) corresponds to the initial temperature field of \(l_T = 3.3\) mm. The cases with \(\varepsilon _p <1\) (i.e., \(l_T\) of 1 and 2 mm with \(\varepsilon _p\) of 0.3 and 0.6) are less likely to develop detonation while the cases with \(\varepsilon _p\) of 1.2, 1.4, and 3.8 for \(l_T\) of 4 mm, 5 mm, and 10 mm, respectively, are more likely to develop strong detonation if \(1< \xi < 10\).
For a uniform mixture with only thermal fluctuations, the predictive \(\xi _p\) is scaled as [11, 12]
where the statistical mean temperature gradient is further approximated by \(|\widetilde{\nabla T}| \cong T'/ \lambda _T\), with the Taylor mixing length scale, \(\lambda _T \cong (l_T \alpha /T')^{0.5}\) [11]. These predictive parameters, \(\xi _p\), and \(\varepsilon _p\), will be validated in the subsequent sections.
2.6.1 Baseline Cases with No Turbulence
In this subsection, three baseline (no turbulence) cases with \(l_T\) of 1 mm, 2 mm, and 5 mm denoted as BL1, BL2, and BL5, respectively, are chosen for discussion. The probability density functions (PDF) of \(\xi \), the temporal evolution of the normalized \(P_{\max }\), and the contours of pressure at the time of the peak of \(F_H\) for three baseline cases are shown in Figs. 12 and 13. According to the scaling of \(\xi _p\) above, with a fixed \(T'\) of 15 K, increasing \(l_T\) from 1 to 5 mm decreases \(\xi _p\) which is consistent with the PDF of \(\xi \) shifting toward the developing detonation regime (see Fig. 12). Consequently, BL5 is predicted to be more prone to developing detonation as compared to BL1 and BL2.
Consistent with the PDFs of \(\xi \) and \(\varepsilon _p = 0.3\), \(P_{\max }\le P_\mathrm{VN}\) and the isocontours of pressure in Figs. 12 and 13, respectively, confirm that BL1 features a spontaneous ignition process. BL2 with \(F_\mathrm{VN} \approx 0.07\) is characterized as a mild detonation process as also quantitatively predicted by \(\varepsilon _p = 0.6\). Unlike BL1 and BL2, a strong detonation process is observed for BL5, quantified by \(F_\mathrm{VN} \approx 0.2\) and \(P_{\max } \gg P_\mathrm{VN}\), which is accurately predicted by \(\varepsilon _p = 1.6\) for BL5.
The statistical quantities, \(F_H\), \(F_D\) and \(\mathrm {\overline{{\xi }}}\), of all the baseline cases are plotted in Fig. 14. Three main points are noted from Fig. 14. First, \(F_D\) has a better agreement with \(F_H\) for \(\varepsilon _p > 1 \) (i.e., cases with \(l_T > 4\) mm in Fig. 14b). Second, the magnitude of \(F_H\) for 3-D cases is comparable with the corresponding 2-D cases as seen in Fig. 14a despite the peak \(F_\mathrm{VN}\) of the 3-D cases slightly lower than that of the 2-D cases [3]. Third, Fig. 14 also reveals a good correlation between \(\mathrm {\overline{{\xi }}}\) and the knock intensity characterized by the \(F_H\) magnitude. The correlation indicates that the conditions with \(\mathrm {\overline{{\xi }}}\) ranging between 0.5 and 30 experience developing detonation and the strongest detonation intensity (\(F_H\) approaches unity) occurs at \(\mathrm {\overline{{\xi }}}\) \(\sim \) 6 with \(\varepsilon _p > 1\). Specifically, the maximum values of \(F_H\) at \(\mathrm {\overline{{\xi }}}\) \(\sim \) 6 are approximately 0.18, 0.22, 0.55, and 0.75 for \(l_T\) of 1 mm, 2 mm, 5 mm, and 10 mm, respectively. These results verify that the prediction of \(\varepsilon _p\) for knock propensity is consistent with \(F_H\), the detonation intensity is greatly alleviated with decreasing \(l_T\) (mild-to-no detonation for \(l_T < 3.3\) mm with \(\varepsilon _p < 1\)).
2.6.2 Effect of Turbulence
The previous subsection demonstrated that no detonation was observed for BL1 while BL5 exhibited a strong detonation process because a minimum run-up distance is required for developing detonation as accurately predicted by \(\varepsilon _t\). The results suggest that decreasing (increasing) the length scale of temperature fluctuations (hence the dissipation element length) is a key parameter to alleviate (promote) knock intensity. In this subsection, the additional effect of turbulent velocity fields on knock intensity is also investigated [4].
The isocontours of pressure and HRR of two representative 2-D simulations and a 3-D simulation in Figs. 15, 16 and 17 are chosen for discussion. With the same turbulent velocity fluctuation, \(u'\) of 83.3 m/s, and root-mean-square temperature fluctuation, \(T'\) of 15 K, \(l_T/l_e\) and \(\tau _{ig}/\tau _t\) were found to be two key parameters to determine knock propensity.
First, increasing turbulent intensity by increasing \(u'\) and fixing \(l_e = 5\) mm is not effective in reducing knock intensity since with \(\tau _\mathrm{ig}/\tau _t\) of 1.0, turbulence is not fast enough to modify the autoignition and the subsequent detonation process as visually seen in Fig. 16. Rather, the main effect of turbulent eddies is to weakly wrinkle and distort the detonation fronts. Second, a finer turbulence structure, \(l_e\) of 1 mm \(\ll l_T = 5\) mm, and \(\tau _\mathrm{ig}/\tau _t\) of 5.0, effectively attenuates the detonation intensity as visually seen in Figs. 16 and 17. High turbulence intensity together with a short mixing-time scale significantly increases the amount of reactants burned by corrugating turbulent flames rather than by the developing detonation phase. Fast and small turbulence structures also induce smaller and finer localized detonation waves during the developing detonation process. Third, for the 3-D cases with a slightly lower \(F_{H}\), a consistent agreement between the 2-D and 3-D cases was observed in terms of \(P_{\max }\), the duration of detonation process, and the level of pressure oscillations, \(P'\) [4].
Contrary to the case with \(l_e =1\) mm \(\ll l_T = 5\) mm being able to attenuate detonation intensity, a fast turbulent mixing, \(\tau _\mathrm{ig}/\tau _t\) of 5, together with \(l_e = 5\) mm \(\gg l_T =1\) mm, promotes detonation development as seen in Fig. 18. Turbulence with a larger eddy size effectively spreads the temperature field over a wider area, stretching out the run-up distance, and thus facilitates the developing detonation [4].
3 Weak and Strong Ignition Modes
3.1 Theoretical Background
Based on Zeldovich’s theory and validated by high-fidelity DNS data, Sankaran et al. [15] proposed a criterion to delineate the weak or strong ignition modes, referred to as the Sankaran number (Sa) herein, which is based on the ratio of laminar flame speed, \(S_L\), to the ignition front speed, \({{S_\mathrm{sp}}} = |\nabla \tau _{\mathrm{ig}}|^{-1}\) [44].
where \(\beta \) is a weighting factor of 0.5 to account for some delay in developing a deflagration from the ignition kernel due to the interactions of thermal and radial diffusion between the neighboring regions. Sa = 1 serves as the boundary between the strong (Sa < 1) and the weak (Sa > 1) ignition.
In practical combustion devices, the detailed information of the spatially resolved scalar fluctuations is not available, and thus the prediction for in a real engine should be based on the statistical quantities such as the mean and variance of fluctuations in RANS or LES simulations. Im et al. [11] conducted a theoretical scaling analysis by incorporating the turbulent effect into the original Sankaran criterion to derive the turbulent ignition regime diagram as shown in Fig. 19.
For compositionally homogeneous reactant mixtures with turbulent velocity and temperature fluctuations, the diagram classifies ignition modes into three categories: weak exhibiting a dominant slow-speed deflagration mode of fuel consumption, reaction-dominant strong, and mixing-dominant strong exhibiting a dominant volumetric-ignition mode of fuel consumption. Pal et al. [74] and Luong et al. [1] conducted a series of DNS simulations and found consistent results with the predictions by the turbulent ignition regime diagram. The ignition criterion [11] was further extended to accommodate both temperature and concentration fluctuations and validated for the mixtures with NTC and non-NTC behaviors [1].
The gradient of ignition delay time, \(\nabla \tau _{\mathrm{ig}}\), for a mixture with a uniform composition and pressure is solely due to temperature variations expanded by the chain rule as [11]:
where \(\text {d}\tau _{\mathrm{ig}}/\text {d}T\) represents the ignition delay sensitivity to temperature as representatively shown in Fig. 20b, and \(\nabla T\) is the temperature gradient of the bulk mixture [11, 74]. In turbulent conditions, the temperature gradient is approximated by the statistical mean temperature gradient, \(\widetilde{|\nabla T|}\), which is estimated based on the variance of the temperature, \(T'\), fluctuations and the Taylor mixing scale, \(\lambda _T\), as
where the Taylor mixing scale is approximated as
with Re\(_l = u'l/\nu \) being the turbulent Reynolds number, \(\nu \) the kinematic viscosity of the bulk mixture gas, \(u'\) the root mean square (RMS) turbulent velocity fluctuation, and l the integral scale of velocity field.
Further scaling analysis yields the expression for the predicted Sankaran number, Sa\(_p\), [11] as follows:
where \(\mathrm{{Da}}_l\) = \(\tau _{t}\mathrm{}\)/\(\tau _{ig}\mathrm{}\) is the ignition Damköhler number, defined as the ratio of the turbulence integral time scale, \(l/u'\), to the ignition delay time, and \(\tau _f\mathrm{{}}\) = \(\alpha /{S^2_L}\) is the characteristic flame time, where \(\alpha \) is the thermal diffusivity of the initial bulk mixture. K is referred to as the normalized thermal ignition sensitivity.
In the presence of both temperature and equivalence ratio (concentration) fluctuations, \(\tau _\mathrm{ig}\) is a function of T and \(\phi \) such that \(|\nabla \tau _\mathrm{{ig}}|\) is expanded by using the chain rule as flows:
where \(\nabla \phi \) and \(\partial {\tau _\mathrm{ig}} / {\partial \phi }\) are the equivalence ratio gradients, and the ignition delay sensitivity to equivalence ratio, respectively [1]. Assuming that the temperature and equivalence ratio fields are correlated with the turbulent flow field, it results in a comparable integral length scale, \(l_T \sim l_{\phi } \sim l_e\), and \(\widetilde{\nabla \phi } \sim \phi '/\lambda _T\) as approximated in the temperature case.
As such K in Fig. 14 is extended as:
where the sign of s is to account for the NTC behavior and the correlation of T-\(\phi \) distribution. s is positive if T and \(\phi \) fluctuations are uncorrelated, thereby being additive to \(|\nabla \tau _\mathrm{{ig}}|\) that synergistically promotes the deflagration mode regardless of the NTC regime. s adopts a negative (positive) sign for a negative \(T - \phi \) correlation if the product of \({\partial {\tau _\mathrm{ig}}} / {\partial T}\) and \({\partial {\tau _\mathrm{ig}}} / {\partial \phi }\) is positive (negative) that the combined effects of T and \(\phi \) on K become subtractive (additive).
3.2 Prediction of Weak/strong Combustion Modes
Many fuels with and without the NTC behavior were chosen to investigate the effect of temperature and equivalence ratio fluctuations on the combustion modes by varying the initial mean temperatures and different fluctuation levels [1, 20, 27, 28, 30, 75,76,77]. The initial conditions and physical parameters are chosen to be relevant to IC engine conditions [1, 33,34,35,36,37,38,39,40,41,42,43]. The DNS solution were processed to verify the predictive accuracy of the ignition criteria.
Figure 20 shows the homogeneous ignition delay time of DME/air mixtures and ethanol/air mixtures and the sensitivity of ignition delay time to temperature that is needed to compute Sa\(_p\). DME exhibits the negative-temperature coefficient (NTC) regime in which \(\tau _\mathrm{{ig}}\) increases with increasing temperature such that its \(\tau _\mathrm{{ig}}\) varies non-monotonically with temperature. As such, within/near the NTC regime, the variation of d\(\tau _\mathrm{ig}/\text {d}T\) for DME is small compared to that of single-stage fuels (e.g., ethanol), which in turn leads to a small \(|\nabla \tau _\mathrm{{ig}}|\) for the same temperature fluctuation level. It suggests that within/near the NTC regime, the mixture with \(T'\) only is more susceptible to spontaneous ignition as quantitatively predicted by Sa\(_p\) in Fig. 21 [2].
Figure 22 shows a typical initial condition used in the simulations. These DNS cases were adopted to perform the statistical analysis of \(\mathrm{{Sa}}\) such that the volume-average (mean) Sa, \(\mathrm {\overline{{S}}a}\), the probability density function of Sa, \(P(\mathrm {Sa})\), and the volume fraction of the regions with Sa <1, \(F_{\mathrm{{Sa,}}S}\), extracted from the spatial Sa distribution are obtained. In addition, \(\mathrm{{Sa}_p}\) for each simulation was also computed based on the information of the initial mean bulk mixture condition such as \(T_0\), \(p_0\), and \(\phi _0\), and the RMS values, \(T'\) and \(\phi '\) as representatively shown in Fig. 21 as a function of temperature.
The isocontours of HRR for some representative DME cases are shown in Fig. 23. Spontaneous ignition characterized by volumetric ignition is dominant even with \(T'\) of 60 K at \(T_0\) of 900 K due to a small variation of \(\tau _{ig}\mathrm{}\)with temperature within/near the NTC regime as shown in Fig. 20). On the contrary, deflagration propagation becomes dominant for the cases with \(T_0\) of 680 and 1045 K outside the NTC regime. Similar results were found for n-heptane [20, 29,30,31, 75, 76, 78, 79].
Contrary to the \(T'\)-only cases, in the presence of both T and \(\phi \), especially when they are negatively correlated, \(T_0\) of 900 K results in a much lower HRR (due to the synergistic effect of \(T'\) and \(\phi '\)) than those of \(T_0\) outside the NTC regime (due to the subtractive effect of \(T'\) and \(\phi '\) for the cases with \(T_0\) of 770 and 1045 K), as seen in the bottom row of Fig. 23.
Quantitative assessment of combustion modes for the previous cases was also conducted. For each initial field of a DNS case, the exact distribution of Sa within the domain and its associated statistical quantities can be computed. For direct validations with the predictive quantities such as \(F_S\) and \(\mathrm {\overline{{S}}a}\), the fractional contributions of weak and strong ignition to the total heat release were extracted from the detailed DNS solution.
Figure 22a-b shows representative initial fields of temperature and equivalence ratio distribution. Based on these detailed DNS information, the spatial distribution of \(\tau _{ig}\mathrm{}\) and Sa, and the corresponding probability density function (PDF) of Sa, \(P(\mathrm {Sa})\) are computed in Figs. 22c–d and 24, respectively. The volume fraction of the regions attributed to strong ignition was also estimated as
which was proposed as a quantitative metric to predict the fractional cumulative heat release rate associated with strong ignition directly computed from DNS solution fields, \(F_{\mathrm{Da},S}\) as [20, 27, 28, 30, 75, 76]
where the summation is operated over the total number of computational cells in the DNS domain, \(\dot{q}\) is the HRR, and Da is Damköhler number defined as the ratio of the local reaction and diffusion terms for a selected scalar variable [16, 17, 20, 28, 30, 71, 75, 76, 80]. Strong ignition is dominant if \(F_{\mathrm{{Sa,}}S}\) \(\sim \) 1. Strong and weak ignition is delineated by a threshold value, Da\(_\mathrm{{0}}\), of order unity that depends on the chosen initial condition. The significance of the Sa-based metric is that the prediction of combustion modes is based purely on the initial conditions.
Consistent with the qualitative observations in the previous section, \(\mathrm {\overline{{S}}a}\) and \(F_{\mathrm{{Sa,}}S}\) for the DME cases of \(T_0\) at 680 K, 900 K, and 1045 K are found to be, respectively, 4.2 & 0.20, 0.5 & 0.92, and 1.3 & 0.4. The cases with \(T_0\) of 680 K and 1045 K have \(\mathrm {\overline{{S}}a}\) greater than one, suggesting a weak-ignition dominance, while for the case with \(T_0\) of 900 K, a strong ignition is expected. The predictive accuracy of \(\mathrm {\overline{{S}}a}\) and \(F_{\mathrm{{Sa,}}S}\) is further verified by the magnitude of \(F_{\mathrm{{Da,}}S}\) as shown in Fig. 25.
Figure 25 shows a strong correlation between \(\mathrm {\overline{{S}}a_n}\)= \(\mathrm {\overline{{S}}a}\)/(1+ \(\mathrm {\overline{{S}}a}\)) and \(F_{\mathrm{{Sa,}}S}\) regardless of \(T_0\), \(T'\), \(\phi '\), and fuel types. A good correlation between \(\mathrm{{Sa}_p}\) and \(F_{\mathrm{{Sa,}}S}\) was also demonstrated, which is attributed to the agreement between the predicted \(\mathrm{{Sa}_p}\) and the volume-average \(\mathrm {\overline{{S}}a}\). Based on this universal correlation, the magnitude of either \(\mathrm {\overline{{S}}a}\) or \(\mathrm{{Sa}_p}\) can be used to quantitatively infer \(F_{\mathrm{{Sa,}}S}\), which characterize the combustion intensity. In general, a higher \(\mathrm {\overline{{S}}a}\) corresponds to a lower \(F_{\mathrm{{Sa,}}S}\), which translates into a lower combustion intensity. In particular, \(\mathrm {\overline{{S}}a_n}\) = 0.5 serves as a transition point from SI to WI dominance with 60% heat released by spontaneous ignition mode (\(F_{\mathrm{{Sa,}}S}\) = 0.6). Spontaneous ignition dominance (\(F_{\mathrm{{Sa,}}S}\) = 1.0) occurs at \(\mathrm {\overline{{S}}a}\) < 0.3. On the contrary, weak ignition is expected to be dominant at \(\mathrm {\overline{{S}}a}\) > 1.0. In summary, either \(\mathrm {\overline{{S}}a}\) or \(\mathrm{{Sa}_p}\) can be used to quantitatively predict the combustion modes over a wide range of initial conditions and different fuel types.
4 Summary
Recent developments in the theoretical study to predict detonation development in IC engine conditions were summarized. A brief review was given on the theoretical framework, starting from the original Zeldovich theory, and the regime diagrams per Bradley [46] and Im [11], and various proposed predictive criteria by Luong et al. [3, 4]. A large collection of DNS simulation data were used for a priori and a posteriori test for the validation of different predictive criteria based on the initial solution field, and its extension to large turbulent conditions with statistical average and variances has been suggested.
In real combustion engines, additional complexities arise from the front-to-wall and front-to-front interactions which may further enhance or suppress the occurrence and intensity of the detonation event. This is of strong practical significance and is currently being investigated by large-scale simulations [3, 4].
References
Luong MB, Hernández Pérez FE, Im HG (2020) Prediction of ignition modes of NTC-fuel/air mixtures with temperature and concentration fluctuations. Combust Flame 213:382–393
Luong MB, Hernández Pérez FE, Sow A, Im HG (2019) Prediction of ignition regimes in DME/air mixtures with temperature and concentration fluctuations. AIAA SciTech 2019 Forum https://doi.org/10.2514/6.2019-2241
Luong MB, Desai S, Hernández Pérez FE, Sankaran R, Johansson B, Im HG (2020) A statistical analysis of developing knock intensity in a mixture with temperature inhomogeneities. Proc Combust Inst 37
Luong MB, Desai S, Hernández Pérez FE, Sankaran R, Johansson B, Im HG (2006) Effects of turbulence and temperature fluctuations on knock development in an ethanol/air mixture. Flow Turbul Combust
Wang Z, Liu H, Reitz RD (2017) Knocking combustion in spark-ignition engines. Prog Energy Combust Sci 61:78–112
Kalghatgi GT, Bradley D (2012) Pre-ignition and ‘super-knock’ in turbo-charged spark-ignition engines. Int J Engine Res 13:399–414
Kalghatgi GT (2015) Developments in internal combustion engines and implications for combustion science and future transport fuels. Proc Combust Inst 35:101–115
Figueroa-Labastida M, Badra J, Elbaz AM, Farooq A (2018) Shock tube studies of ethanol preignition. Combust. Flame 198:176–185
Meyer JW, Oppenheim AK (1971) On the shock-induced ignition of explosive gases. Symp (Int) Combust 13:1153–1164
Lutz AE, Kee RJ, Miller JA, Dwyer HA, Oppenheim AK (1989) Dynamic effects of autoignition centers for hydrogen and c1, 2-hydrocarbon fuels. Symp (Int) Combust 22:1683–1693
Im HG, Pal P, Wooldridge MS, Mansfield AB (2015) A regime diagram for autoignition of homogeneous reactant mixtures with turbulent velocity and temperature fluctuations. Combust Sci Technol 187:1263–1275
Towery CAZ, Poludnenko AY, Hamlington PE (2020) Detonation initiation by compressible turbulence thermodynamic fluctuations. Combust Flame 213:172–183
Hernández Pérez FE, Mukhadiyev N, Xu X, Sow A, Lee BJ, Sankaran R, Im HG (2018) Direct numerical simulations of reacting flows with detailed chemistry using many-core/GPU acceleration. Comput Fluids 173:73–79
Desai S, Yu JK, Song W, Luong MB, Hernández Pérez FE, Sankaran R, Im HG (2020) Direct numerical simulations of reacting flows with shock waves and stiff chemistry using many-core/GPU acceleration. Comput Fluids, In revision
Sankaran R, Im HG, Hawkes ER, Chen JH (2005) The effects of non-uniform temperature distribution on the ignition of a lean homogeneous hydrogen-air mixture. Proc Combust Inst 30:875–882
Chen JH, Hawkes ER, Sankaran R, Mason SD, Im HG (2006) Direct numerical simulation of ignition front propagation in a constant volume with temperature inhomogeneities: I. fundamental analysis and diagnostics. Combust Flame 145:128–144
Hawkes ER, Sankaran R, Pébay P, Chen JH (2006) Direct numerical simulation of ignition front propagation in a constant volume with temperature inhomogeneities: II. parametric study. Combust Flame 145:145–159
Bansal G, Im HG (2011) Autoignition and front propagation in low temperature combustion engine environments. Combust Flame 158:2105–2112
Gupta S, Im HG, Valorani M (2011) Classification of ignition regimes in HCCI combustion using computational singular perturbation. Proc Combust Inst 33:2991–2999
Yoo CS, Lu T, Chen JH, Law CK (2011) Direct numerical simulations of ignition of a lean \(n\)-heptane/air mixture with temperature inhomogeneities at constant volume: parametric study. Combust Flame 158:1727–1741
Yu R, Bai X-S (2013) Direct numerical simulation of lean hydrogen/air auto-ignition in a constant volume enclosure. Combust Flame 160:1706–1716
El-Asrag HA, Ju Y (2013) Direct numerical simulations of exhaust gas recirculation effect on multistage autoignition in the negative temperature combustion regime for stratified HCCI flow conditions by using H\(_2\)O\(_2\) addition. Combust Theory Model 17:316–334
El-Asrag HA, Ju Y (2014) Direct numerical simulations of NO\(_{x}\) effect on multistage autoignition of DME/air mixture in the negative temperature coefficient regime for stratified HCCI engine conditions. Combust Flame 161:256–269
Bhagatwala A, Lu T, Chen JH (2014) Direct numerical simulations of HCCI/SACI with ethanol. Combust Flame 161:1826–1841
Bhagatwala A, Sankaran R, Kokjohn S, Chen JH (2015) Numerical investigation of spontaneous flame propagation under RCCI conditions. Combust Flame 162:3412–3426
Bansal G, Mascarenhas A, Chen JH (2015) Direct numerical simulations of autoignition in stratified dimethyl-ether (DME)/air turbulent mixtures. Combust Flame 162:688–702
Yoo CS, Luo Z, Lu T, Kim H, Chen JH (2013) A DNS study of ignition characteristics of a lean \(iso\)-octane/air mixture under and SACI conditions. Proc Combust Inst 34:2985–2993
Luong MB, Luo Z, Lu T, Chung SH, Yoo CS (2013) Direct numerical simulations of the ignition of lean primary reference fuel/air mixtures with temperature inhomogeneities. Combust Flame 160:2038–2047
Kim SO, Luong MB, Chen JH, Yoo CS (2015) A DNS study of the ignition of lean PRF/air mixtures with temperature inhomogeneities under high pressure and intermediate temperature. Combust Flame 162:717–726
Luong MB, Yu GH, Chung SH, Yoo CS (2017) Ignition of a lean PRF/air mixture under RCCI/SCCI conditions: a comparative DNS study. Proc Combust Inst 36:3623–3631
Yu GH, Luong MB, Chung SH, Yoo CS (2019) Ignition characteristics of a temporally evolving \(n\)-heptane jet in an \(iso\)-octane/air stream under RCCI combustion-relevant conditions. Combust Flame 208:299–312
An Y, Jaasim M, Vallinayagam R, Vedharaj S, Im HG, Johansson B (2018) Numerical simulation of combustion and soot under partially premixed combustion of low-octane gasoline. Fuel 211:420–431
Mittal G, Sung C-J (2006) Aerodynamics inside a rapid compression machine. Combust Flame 145:160–180
Wang Y, Rutland CJ (2005) Effects of temperature and equivalence ratio on the ignition of \(n\)-heptane fuel spray in turbulent flow. Proc Combust Inst 30:893–900
Kokjohn SL, Musculus MPB, Reitz RD (2015) Evaluating temperature and fuel stratification for heat-release rate control in a reactivity-controlled compression-ignition engine using optical diagnostics and chemical kinetics modeling. Combust Flame 162:2729–2742
Tang Q, Liu H, Yao M (2017) Simultaneous measurement of natural flame luminosity and emission spectra in a RCCI engine under different fuel stratification degrees. SAE Int J Engines 10:2017–01–0714
Schießl R, Maas U (2003) Analysis of endgas temperature fluctuations in an SI engine by laser-induced fluorescence. Combust Flame 133:19–27
Kaiser SA, Schild M, Schulz C (2013) Thermal stratification in an internal combustion engine due to wall heat transfer measured by laser-induced fluorescence. Proc Combust Inst 34:2911–2919
Söderberg F, Johansson B, Lindoff B (1998) Wavelet analysis of in-cylinder ldv measurements and correlation against heat-release. SAE paper 107:444–456
Funk C, Sick V, Reuss DL, Dahm WJA, Turbulence properties of high and low swirl in-cylinder flows, SAE paper (2002) 2002–01–2841
Miles PC (2008) Turbulent flow structure in direct-injection, swirl-supported diesel engines. In: Flow and combustion in reciprocating engines, experimental fluid mechanics, Springer, pp 173–256
Akkerman V, Ivanov M, Bychkov V (2009) Turbulent flow produced by piston motion in a spark-ignition engine. Flow Turbul Combust 82:317–337
Petersen BR, Ghandhi JB (2010) High resolution scalar dissipation and turbulence length scale measurements in an internal combustion engine. SAE Int J Engines 3:65–83
Zeldovich YB (1980) Regime classification of an exothermic reaction with nonuniform initial conditions. Combust Flame 39:211–214
Gu X, Emerson D, Bradley D (2003) Modes of reaction front propagation from hot spots. Combust Flame 133:63–74
Bradley D, Morley C, Gu XJ, Emerson DR (2002) Amplified pressure waves during autoignition: Relevance to CAI engines. SAE Technical Paper (2002) 2002–01–2868
Lee JH, Knystautas R, Yoshikawa N (1978) Photochemical initiation of gaseous detonations. Acta Astronautica 5:971–982
Lee JHS (2008) The Detonation Phenomenon. Cambridge University Press
Su J, Dai P, Chen Z (2020) Detonation development from a hot spot in methane/air mixtures: effects of kinetic models. Int J Engine Res 146808742094461
Gao Y, Dai P, Chen Z (2020) Numerical studies on autoignition and detonation development from a hot spot in hydrogen/air mixtures. Combust Theo Modell 24:245–261
Bates L, Bradley D, Paczko G, Peters N (2016) Engine hot spots: modes of auto-ignition and reaction propagation. Combust Flame 166:80–85
Bates L, Bradley D (2017) Deflagrative, auto-ignitive, and detonative propagation regimes in engines. Combust Flame 175:118–122
Peters N, Kerschgens B, Paczko G (2013) Super-knock prediction using a refined theory of turbulence. SAE Int J Engines 6:953–967
Dai P, Chen Z, Chen S, Ju Y (2015) Numerical experiments on reaction front propagation in n-heptane/air mixture with temperature gradient. Proc Combust Inst 35:3045–3052
Yu H, Chen Z (2015) End-gas autoignition and detonation development in a closed chamber. Combust Flame 162:4102–4111
Dai P, Qi C, Chen Z (2017) Effects of initial temperature on autoignition and detonation development in dimethyl ether/air mixtures with temperature gradient. Proc Combust Inst 36:3643–3650
Terashima H, Matsugi A, Koshi M (2017) Origin and reactivity of hot-spots in end-gas autoignition with effects of negative temperature coefficients: Relevance to pressure wave developments. Combust Flame 184:324–334
Pan J, Wei H, Shu G, Chen R (2017) Effect of pressure wave disturbance on auto-ignition mode transition and knocking intensity under enclosed conditions. Combust Flame 185:63–74
Wei H, Chen C, Shu G, Liang X, Zhou L (2018) Pressure wave evolution during two hotspots autoignition within end-gas region under internal combustion engine-relevant conditions. Combust Flame 189:142–154
Sow A, Lee BJ, Hernández Pérez FE, Im HG (2019) Detonation onset in a thermally stratified constant volume reactor. Proc Combust Inst 37:3529–3536
Pan J, Wei H, Shu G, Chen Z, Zhao P (2016) The role of low temperature chemistry in combustion mode development under elevated pressures. Combust Flame 174:179–193
Pan J, Dong S, Wei H, Li T, Shu G, Zhou L (2019) Temperature gradient induced detonation development inside and outside a hotspot for different fuels. Combust Flame 205:269–277
Desai S, Sankaran R, Im HG (2019) Unsteady deflagration speed of an auto-ignitive dimethyl-ether (DME)/air mixture at stratified conditions. Proc Combust Inst 37:4717–4727
Desai S, Sankaran R, Im HG (2020) Auto-ignitive deflagration speed of methane (CH\(_4\)) blended dimethyl-ether (DME)/air mixtures at stratified conditions. Combust Flame 211:377–391
Robert A, Richard S, Colin O, Martinez L, De Francqueville L (2015) LES prediction and analysis of knocking combustion in a spark ignition engine. Proc Combust Inst 35:2941–2948
Robert A, Richard S, Colin O, Poinsot T (2015) LES study of deflagration to detonation mechanisms in a downsized spark ignition engine. Combust Flame 162:2788–2807
Ali MJM, Luong MB, Sow A, Hernández Pérez FE, Im HG (2018) Probabilistic approach to predict abnormal combustion in spark ignition engines. SAE paper 2018–01–1722
Wei H, Chen C, Zhou H, Zhao W, Ren Z (2016) Effect of turbulent mixing on the end gas auto-ignition of n-heptane/air mixtures under IC engine-relevant conditions. Combust Flame 174:25–36
Chen L, Wei H, Chen C, Feng D, Zhou L, Pan J (2019) Numerical investigations on the effects of turbulence intensity on knocking combustion in a downsized gasoline engine. Energy 166:318–325
Zhong L, Liu C (2019) Numerical analysis of end-gas autoignition and pressure oscillation in a downsized SI engine using large eddy simulation. Energies 12:3909
Zhang T, Sun W, Wang L, Ju Y (2019) Effects of low-temperature chemistry and turbulent transport on knocking formation for stratified dimethyl ether/air mixtures. Combust Flame 200:342–353
Nogawa T, Terashima H (2020) Effects of globally stratified temperature distributions and NTC characteristics on end-gas combustion modes. Combust Sci Technol 1–25
Wang L, Peters N (2006) The length-scale distribution function of the distance between extremal points in passive scalar turbulence. J Fluid Mech 554:457–475
Pal P, Valorani M, Arias PG, Im HG, Wooldridge MS, Ciottoli PP, Galassi RM (2017) Computational characterization of ignition regimes in a syngas/air mixture with temperature fluctuations. Proc Combust Inst 36:3705–3716
Luong MB, Lu T, Chung SH, Yoo CS (2014) Direct numerical simulations of the ignition of a lean biodiesel/air mixture with temperature and composition inhomogeneities at high pressure and intermediate temperature. Combust Flame 161:2878–2889
Luong MB, Yu GH, Lu T, Chung SH, Yoo CS (2015) Direct numerical simulations of ignition of a lean \(n\)-heptane/air mixture with temperature and composition inhomogeneities relevant to HCCI and SCCI combustion. Combust Flame 162:4566–4585
Zhang J, Luong MB, Hernández Pérez FE, Han D, Im HG, Huang Z (2019) Exergy loss of dme/air mixtures and ethanol/air mixtures with temperature and concentration fluctuations under HCCI/SCCI conditions: a DNS study, submitted to proceedings of the combustion institute (2019)
Luong MB, Yu GH, Chung SH, Yoo CS (2017) Ignition of a lean PRF/air mixture under RCCI/SCCI conditions: chemical aspects. Proc Combust Inst 36:3587–3596
Luong MB, Sankaran R, Yu GH, Chung SH, Yoo CS (2017) On the effect of injection timing on the ignition of lean PRF/air/EGR mixtures under direct dual fuel stratification conditions. Combust Flame 183:309–321
Echekki T, Chen JH (2003) Direct numerical simulation of autoignition in non-homogeneous hydrogen-air mixtures. Combust Flame 134:169–191
Acknowledgements
This work was sponsored by King Abdullah University of Science and Technology and used the resources of the KAUST Supercomputing Laboratory.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
About this chapter
Cite this chapter
Luong, M.B., Im, H.G. (2022). Direct Numerical Simulation of Preignition and Knock in Engine Conditions. In: Gupta, A.K., De, A., Aggarwal, S.K., Kushari, A., Runchal, A.K. (eds) Advances in Energy and Combustion. Green Energy and Technology. Springer, Singapore. https://doi.org/10.1007/978-981-16-2648-7_14
Download citation
DOI: https://doi.org/10.1007/978-981-16-2648-7_14
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-16-2647-0
Online ISBN: 978-981-16-2648-7
eBook Packages: EnergyEnergy (R0)