LES analysis for auto-ignition induced abnormal combustion based on a downsized SI engine

Applied Energy 191 (2017) 183–192

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

LES analysis for auto-ignition induced abnormal combustion based on a downsized SI engine Jiaying Pan, Haiqiao Wei ⇑, Gequn Shu, Mingzhang Pan, Dengquan Feng, Nan Li State Key Laboratory of Engines, Tianjin University, Tianjin 300072, China

h i g h l i g h t s
Knock and super-knock are studied using LES with detailed chemistry solver.
Classical knocking intensity varies proportionally with spark-ignition timing.
Low-temperature chemical reaction plays important role in super-knock formation.
Developing detonation can be induced by multiple or single hot-spot auto-ignition.
Developing detonation wave directly by single hot-spot produce stronger knocking.

a r t i c l e

i n f o

Article history: Received 20 September 2016 Received in revised form 30 December 2016 Accepted 15 January 2017 Available online 1 February 2017 Keywords: Engine knock Super-knock Auto-ignition Primary Reference Fuel Developing detonation

a b s t r a c t Engine knock and super-knock have become the main barriers to significantly improving engine thermal efficiency. To further study the nature of the abnormal combustion, this work quantitatively investigates engine knock and super-knock using a Large Eddy Simulation framework coupling detailed chemistry solver. Firstly, classical knocking cycles with different knocking intensities have been calculated through adjusting spark-ignition timing. It shows that knocking onset and intensity vary proportionally with the advance of spark-ignition timing, however, super-knock events are not observed under the operation conditions. Then for a given spark-ignition timing, the blends of Primary Reference Fuels are introduced in order to obtain different octane number of mixture, through which super-knock events with stronger knocking intensity are observed. The results show that as the decreases of octane number, knocking onset is significantly advanced due to the enhancement of low-temperature chemical reactivity. Consequently, more auto-ignition centers appear at hot exhaust valve side and even cool intake valve side at very low octane number. But for the knocking intensity, it does not always show a proportional correlation with octane number during super-knock. Further auto-ignition scenarios show that developing detonation wave can be induced by both multiple hot-spots auto-ignition and directly by single hot-spot autoignition, with different reaction front curvatures. However, the later seems to produce much stronger knocking intensity, especially when there are several developing detonation waves during superknock. Therefore, how to effectively regulate local auto-ignition initiation and development seems the key to the avoidance of abnormal combustion in modern engines. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Nowadays highly boosted spark-ignited (SI) engines show great advantages in the improvement of thermal efficiency. However, the risk of abnormal combustion, e.g. classical knock and superknock, may often be encountered when the engine is operated under high-load conditions [1]. Engine knock is generally consid⇑ Corresponding author. E-mail address: [email protected] (H. Wei). http://dx.doi.org/10.1016/j.apenergy.2017.01.044 0306-2619/Ó 2017 Elsevier Ltd. All rights reserved.

ered as a result of end-gas auto-ignition before the arrival of main flame front [2], and super-knock is attributed to a developing detonation due to the resonance between acoustic wave and reaction wave in multi-scale turbulent flows. When super-knock occurs, stochastic auto-ignition reaction fronts may consume surrounding mixture within less than a millisecond, which can cause a knocking intensity beyond 20 MPa [3]. Despite the numerous studies devoted to the abnormal combustion [4–7], there are still many ambiguities associated with the key physical–chemical mechanisms, which hinder the rapid development of modern SI engines.

184

J. Pan et al. / Applied Energy 191 (2017) 183–192

For a long time, the visualization experiment has been considered as an important access to the fundamentals of abnormal combustion thanks to its capability in capturing microcosmic mechanism of combustion processes [8]. Based on an optical hydrogen SI engine, Kawahara et al. [9] found that end-gas autoignition and subsequent pressure wave in knocking combustion could be visualized through a high-speed camera. Using chemiluminescence imaging technology, Vafamehr et al. [10] investigated the competing chemical and physical effects of transient fuel enrichment on heavy knock in an optical SI engine. Through high-speed direct photography, Wang et al. [11] experimentally analyzed the events of pre-ignition and super-knock in a rapid compression machine, and possible mechanism for super-knock formation was obtained. Despite the wonderful visualization work, wide applications of the visualization techniques in knocking combustion are still difficult due to the limitations in both engine systems and measuring methods. Compared with visualization experiments, Computation Fluid Dynamics (CFD), especially Large Eddy Simulation (LES), has become attractive to the study of abnormal combustion [12]. This is because the LES allows for the stochastic behavior of individual cycles and the variations of multiple cycles [13]. Based on a new LES framework coupling flame surface density model and tabulated kinetics, Lecocq et al. [14] found that the local autoignition initiation and development, consistent with in-cylinder pressure history, could be accurately captured during SI engine knock. Using dual heat transfer and two-step reduced scheme, Misdariis et al. [15] carried out a LES work to explore the effect of transient temperature distributions on engine knock, and found that local auto-ignition largely depends on the temperature heterogeneity in cylinder. Several numerical studies on classical knock have been performed, however, there is few LES literature focusing on the formation of super-knock under practical engine conditions. Theoretically both classical knock and super-knock are induced by local sporadic auto-ignition phenomena. Rudloff et al. [16] experimentally investigated the relation between preignition and super-knock in terms of auto-ignition initiation and development. They found that early pre-ignition only tends to vary surrounding thermodynamic conditions, and it is the later gaseous auto-ignition at other regions contributing to the eventual super-knock. Bradley et al. [17] pointed out that the developing detonation within a multitude of hot spots would be very damaging, and as the number of hot spots increases, so do the interactions between adjacent ones. Using 15 LES cycles of a high load/low speed SI engine operating point, Robert et al. [5] analyzed deflagration to detonation mechanism, and found that one or a couple of hot spots are strong enough to induce local temperature increases, which further promotes the coupling between pressure wave and auto-ignition reaction rate and eventually the formation of super-knock. Therefore, it is necessary to further explore physical–chemical mechanism of local hot-spots autoignition during abnormal combustion, the results of which will provide effective approaches to the optimization of engine combustion. The primary objectives of current numerical investigations are to further explore the role of local sporadic auto-ignition and fuel property in both classical knock and super-knock based on a highly downsized SI engine. Classical knocking cycles were obtained through the advance of spark-ignition timing, and super-knock events were achieved by decreasing octane number of Primary Reference Fuels (PRF). Current LES numerical investigations will give insights into the ways how to effectively restrain the occurrence of abnormal combustion in modern engines, such as exhaust gas recirculation and water injection technology [18,19].

2. Methodology and model descriptions 2.1. SAGE detailed chemistry solver To accurately predict the sporadic auto-ignition controlled by chemical kinetics, a SAGE detailed chemistry solver [20] was adopted in CONVERGE CFD code [21]. The SAGE solver calculates reaction rate of each elementary reaction while CFD code solves transport equations. Combined with the algorithm of Adaptive Mesh Refinement (AMR), a detailed mechanism can be used to simulate various combustion regimes, including ignition, premixed and mixing-controlled combustion, etc. Therefore, the SAGE solver is perfect for these numerical simulations related to auto-ignition controlled combustion processes. As described by Turns [22], a multi-step chemical reaction mechanism can be written in: M X

M X

i¼1

i¼1

v 0i;r vi ¼

v 00i;r vi

r ¼ 1; 2; . . . :; R

ð1Þ

where v 0i;r and v 00i;r are the stoichiometric coefficients for reactants and products, respectively, for species i and reaction r, R is the total number of reactions and vi represents species i. Net production rate for species i can be obtained by:

x_ i ¼

R X

v i;r qr ;

i ¼ 1; 2; . . . :; M

ð2Þ

r¼1

where M is the total number of species, v i;r ¼ v 00i;r  v 0i;r . The rate-of-progress variable qr for the rth reaction is:

qr ¼ kfr

M M Y Y v0 v 00 ½X i  i;r  kbr ½X i  i;r i¼1

ð3Þ

i¼1

where [Xi] is the molar concentration of species i, and kfr and kbr are the forward and reverse rate coefficient for reaction r. The governing equations of mass and energy conservation can be solved for a given computational cell:

d½X i  _i ¼x dt P dP _ iÞ dT V dt  m ðhi x ¼ P  dt ð½X  c Þ m p;i m

ð4Þ ð5Þ

_ i net reaction where V is volume, T temperature, P pressure and x rate of species i. The hi and cp;i are the molar specific enthalpy and molar constant-pressure specific heat, respectively. The above equations are solved at each CONVERGE computational time-step, and cell temperature is updated after chemistry calculation is converged. To expedite the detailed chemistry, a minimum cell temperature Tcut and minimum hydrocarbon molefraction HCmin are specified, below which the kinetics are not solved. 2.2. One-equation Eddy viscosity model A one-equation sub-grid scale (SGS) eddy viscosity model [23,24] has been used to solve a sub-filter-scale (SFS) kinetic energy transport equation, and turbulent viscosity is modeled by employing the SFS turbulent kinetic energy and filter size. Due to the good performance in adapting various grid levels, this model has been widely used to resolve complex flow filed. The sub-grid kinetic energy equation is given by:

i @k @k @u @ i þu ¼ sij eþ @t @xi @xi @xn



vt

rk

@k @xi



where the sub-grid kinetic energy is given by:

ð6Þ

185

J. Pan et al. / Applied Energy 191 (2017) 183–192

i u j Þ k ¼ 0:5ðui unj  u

ð7Þ

The sub-grid stress tensor is modeled as:

2 3

sij ¼ 2v tk Sij þ kdij

ð8Þ

The turbulent viscosity, vtk, for the one-equation model is given as:

v tk ¼ C k k1=2 D

ð9Þ

The sub-grid dissipation is given as: 3

e¼

C e k2 D

ð10Þ

Fig. 1. Geometry for the SI engine and monitor points in combustion chamber.

Here the turbulent viscosity and sub-grid dissipation can be tuned by adjusting the constant Ck and Ce.

100 Cal. PRF100 Exp. PRF100 Cal. PRF90 Exp. PRF90 Cal. PRF80 Exp. PRF80 Cal. PRF60 Exp. PRF60 Cal. PRF0 Exp. PRF0

Table 1 Main specifications for the experimental SI engine. Parameters

Value

Compression ratio Bore/Stroke (mm) Connection rode length (mm) Engine speed (rpm) Intake pressure (atm) Fuel Intake Valve Open Intake Valve Close

10.6 77/86 132 1800 1.80 PRFs 7° BTDC 18° ABDC

10

1

0.1

P=4 MPa, φ=1.0 0.01 0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1000 / T

(a) 50

40

Flame speed (cm/s)

In present work, knocking experiments were carried out in a Port Fuel Injection, single-cylinder turbocharged SI engine with four valves. Main specifications for the engine can be found in Table 1. For the simulation, high-quality orthogonal hexahedral meshes were generated through CONVERGE CFD code. To keep a good balance between computation efficiency and resolution, the AMR strategy was employed to automatically adjust meshes, depending on the distributions of local velocity, temperature and species mass fraction. As a consequence, the cell size is dx = 0.1–0.2 mm around spark plug and end-gas region, and it is close to dx = 0.5–1.0 mm for the cells at other regions. Meanwhile, the inhomogeneous temperature distributions in cylinder were considered, with cylinder head and piston temperature at 450 K, intake valves at 639 K and exhaust valves at 784 K, respectively. Further, the maximum Courant-Friedrichs-Lewy (CFL) number was adjusted from 50 to 1.75 once combustion starts, which could guarantee the high resolution of gas dynamics (i.e. pressure wave or shock wave). For the cyclical variations, random seed was used in each individual engine cycle, and eight monitor points (P1-P8), with an interval of 45 degree mutually, were set up at the edge of combustion chamber to capture more local information, as showed in Fig. 1. A reduced Primary Reference Fuel (PRF) mechanism developed by Liu et al. [25] has been adopted here rather than detailed chemical kinetics to improve computation efficiency. This mechanism has been validated and tested against a wide range of engine combustion conditions [26–28]. Fig. 2 shows the comparison of ignition delay time and laminar flame speed between experiments and calculations under various initial conditions. The results indicate that current mechanism is able to accurately predict low-to-high temperature ignition delay, including Negative Temperature Coefficient (NTC) regime, as shown in Fig. 2(a). Despite the overprediction under fuel rich conditions, the laminar flame speed calculated yields satisfactory agreement with experimental data at stoichiometric ratio, as shown in Fig. 2(b).

Ignition delay time (ms)

2.3. Model specifications

30 Cal. PRF0 Exp.PRF0 Cal. PRF50 Exp.PRF50 Cal. PRF85 Exp.PRF85 Cal. PRF100 Exp.PRF100

20

10

P=1.0 atm From counterflow

0 0.6

0.8

1.0

1.2

1.4

1.6

Equivalence ratio

(b) Fig. 2. Validations of (a) ignition delay time and (b) laminar flame speed for different PRF blends [25,29]. Lines denote calculation results and symbols represent experimental data with an error bar below ±5%. The PRFx means the blends of x% isooctane and (1  x%) n-heptane in volume.

It should be noted that in current LES work, classical knock is achieved by advancing spark-ignition timing, as general knocking experiments done [30], while super-knock events are generated through decreasing octane number. The reasons for this research technique are as follows: firstly, lubricating oil et al. is generally regarded as the original source of pre-ignition and thereby superknock, however, it shows that early pre-ignition appears as a normal combustion and subsequent gaseous auto-ignition contributes to the eventual super-knock [16]. Secondly, auto-ignition is mainly

J. Pan et al. / Applied Energy 191 (2017) 183–192

controlled by chemical kinetics, such that super-knock can be achieved through enhancing low-temperature reactivity with low octane number of mixture. Actually, with the tendency of intake boost and engine downsizing, it is expected that the NTC regime shall shift to high temperature region, leading to greater relevance of low temperature for regular gasoline [31].

3. Results and discussions 3.1. LES prediction for engine knock

21

ST= -8 CAD ST= -6 CAD ST= -4 CAD ST= -2 CAD ST= 0 CAD ST= 2 CAD ST= 4 CAD

18

In-cyinder pressure (MPa)

186

15 12 9 6 3

Fig. 3. Comparison of in-cylinder pressure traces between engine experiments and simulations at different spark-ignition timing (ST). ST = 0 CAD denotes the location of Top Dead Center.

0 -10

0

10

20

30

40

50

60

50

60

Crank angle (degree)

(a) 6

Pressure oscillation (MPa)

Fig. 3 shows the results of experimental and calculated incylinder pressure, where the experimental data is from 500 engine cycles with a ST sweep from ST = 2 to 8 CAD and the calculation results are obtained from a ST sweep from ST = 8 to 8 CAD to extend research scope. It is observed that under current downsizing level, engine combustion seems very sensitive to the ST. At a later ST (e.g. ST = 4 CAD), in-cylinder pressure curve is relatively smooth without obvious pressure oscillation, which corresponds to a normal combustion. As the ST is advanced to ST = 0 CAD, pressure peak and pressure rise rate are significantly increased. And the advance of combustion phasing is followed by an apparent pressure oscillation superimposing on in-cylinder pressure traces, which suggests the occurrence of engine knock. Compared with the pressure peak of Pmax = 5.7 MPa at ST = 4 CAD, it is increased up to Pmax = 8.6 MPa when ST = 1 CAD. Fig. 4 shows the traces of instantaneous in-cylinder pressure and pressure oscillation filtered by a band frequency of 5–40 kHz at different STs. It is observed that as the advance of the ST, pressure rise rate and pressure peak are significantly increased, e.g. peak pressure increases from Pmax  6.5 MPa at ST = 2 CAD to Pmax  20.5 MPa at ST = 8 CAD. Correspondingly, the knocking intensity KI, defined as the maximum amplitude of pressure oscillations, is KI = 0.56 MPa at ST = 0 CAD and it reaches KI = 1.50 MPa at ST = 2 CAD, which represents a moderate engine knock. Further advances in the spark-ignition timing (e.g. ST = 6 CAD), the knocking intensity is increased up to KI = 4.0 MPa, characterizing a heavy knocking event. Meanwhile, due to the more compression from regular flame front at an early ST, the knocking onset KO, defined as the instant of sharp increases in the first pressure derivative, is linearly advanced as the ST. However, super-

ST= -8 CAD ST= -6 CAD ST= -4 CAD ST= -2 CAD ST= 0 CAD ST= 2 CAD ST= 4 CAD

4

2

0

-2

-4 -10

0

10

20

30

40

Crank angle (degree)

(b) Fig. 4. (a) Instantaneous in-cylinder pressure and (b) pressure oscillation at different spark-ignition timings.

knock events have not been observed only through varying ST in terms of limited knocking intensity. Fig. 5 further shows the evolutions of flame propagation and end-gas auto-ignition represented by temperature and species mass fraction at ST = 8 CAD. It is observed that when the piston moves to 19.11 CA, there is an auto-ignition center induced by single hot-spot at the location around point P8. Then the auto-ignition reaction front produced propagates outward and consumes surrounding mixture with approximate 1.0 CA, which corresponds to a flame speed of several meters per second. Consequently, there is another homogeneous auto-ignition event at the location around point P2 when the piston moves to 20.61 CA. The homogeneous auto-ignition consumes surrounding mixture with a behavior of quasi constant-volume combustion and leads to an intense overpressure with the amplitude over 5 MPa. Eventually a strong pressure wave is constructed and propagates back and forth in cylinder, resulting in obvious pressure oscillating behavior. It should be noted that the reaction front from homogeneous auto-ignition does not develop into a detonation wave because of the insufficiency in unburned mixture, low flame speed and pressure peak, which are consistent with the conclusions obtained before. From the calculations of classical knocking cycles with different knocking intensities, it is seen that current LES results yield good agreement with experimental data in terms of combustion phasing and pressure level. This further demonstrates the feasibility of cur-

J. Pan et al. / Applied Energy 191 (2017) 183–192

187

Fig. 5. Evolutions of flame propagation and end-gas auto-ignition at spark-ignition timing of ST = 8 CAD.

rent model in predicting auto-ignition induced abnormal combustion. As an extension of research scope, super-knock events are further numerically investigated based on current downsized engine configuration.

3.2. LES prediction for super-knock In order to generate super-knock events, spark-ignition timing is maintained ST = 0 CAD and the octane number (ON) of PRF blends are decreased to enhance low-temperature reactivity of mixture. This approach can significantly shorten ignition delay of mixture and promote the occurrence of auto-ignition under the same thermodynamic conditions. Fig. 6 shows the results of incylinder pressure and corresponding pressure oscillation filtered by a band frequency of 5–40 kHz. Different from classical knock events, several new observations can be obtained in the superknock cases. Firstly, it appears more pronounced for the effect of octane number on knocking onset and knocking intensity in comparison to spark-ignition timing, e.g. when the octane number decreases from ON = 100 to 90, pressure peak and knocking intensity increase to Pmax = 55 MPa and KI = 22 MPa, respectively. Secondly, knocking onset is significantly advanced as the ON decreases due to the enhancement of low-temperature chemical reactivity. However, knocking intensity does not show proportional variations with the ON, and there is a maximum level of KI  60 MPa at ON = 40, as shown in Fig. 7. This indicates that knocking intensity is associated with not only auto-ignition timing, but also subsequent combustion processes, such as multiple hotspots interaction and reaction wave propagation. Furthermore, from the comparisons of ON = 0 and 80 cases, it is observed that super-knock events can still take place without pre-ignition. This suggests that there are no necessary correlations between preignition and super-knock: pre-ignition can induce super-knock at early stage of combustion, but super-knock is not necessarily caused by the pre-ignition events, which are consistent with the results of previous engine experiments [31].

To demonstrate the effects of octane number on lowtemperature chemical reactivity, the evolutions of temperature and CH2O mass fraction have been shown in Fig. 8. It is observed that before knocking occurrence, unburned mixture experiences obvious two-stage auto-ignition, where the first one is featured by cool flame temperature rise and species CH2O accumulation, and the second one by sharp increases in high temperature and rapid CH2O consumption. As the decreases of octane number, both the first and second-stage ignition timings are advanced, with more apparent two-stage auto-ignition behavior. Meanwhile, the second-stage ignition delay is significantly shortened at low octane number conditions because of the contributions from cool flame temperature rise, e.g. the cool flame temperature rise increases from 100 K at ON = 100 to 242 K at ON = 0. Further, when the octane number reaches the level below ON = 60, there is obvious pressure disturbances during the second ignition delay, which indicates the complex auto-ignition interactions and strong gas dynamics during super-knock, as shall be discussed in Section 3.3.

3.3. Auto-Ignition scenarios during super-knock In this section, the combustion evolutions in single super-knock cycle are detailed in terms of flame propagation, auto-ignition initiation and strong pressure wave formation. Fig. 9 shows the ON = 90 case where the temperature is presented by isosurfaces and the species mass fraction and pressure difference by counters. It is observed that after spark ignition, main flame front propagates towards cylinder wall with a stretched and wrinkled turbulent behavior. In this process, unburned mixture experiences some progresses of low-temperature combustion, as shown the accumulations of intermediate species CH2O. Due to the fluctuations of species composition and temperature in reactive turbulent flow, a hot-pot auto-ignition begins to appear at the cylinder wall close to points P5 and P6at 19.90 CA. This auto-ignition reaction front consumes surrounding mixture within 0.2 CA, but there is no obvious over-pressure generation in this process. Subsequently, new

J. Pan et al. / Applied Energy 191 (2017) 183–192

100

1500

Temperature [K]

1200

900

ON 100 ON 90 ON 80 ON 60 ON 40 ON 0

10 1 0.1

By pressure disturbances

0.01

600

1E-3 1E-4

300

CH2O Mass fraction

188

1E-5 0 -20

-10

0

10

20

1E-6 30

Crank angle [degree] Fig. 8. Evolutions of temperature and CH2O mass fraction under different octane number conditions.

20

100

15

80

10

60

5

40

0

20

-5 0

20

40

60

80

Knocking intensity (MPa)

Knocking onset (CAD)

Fig. 6. (a) Instantaneous in-cylinder pressure and (b) pressure oscillation at sparkignition timing ST = 0 CAD and different octane number conditions.

0 100

Octane number Fig. 7. Correlations between octane number and knocking onset and knocking intensity.

hot-spots auto-ignition occurs at the location adjacent to point P2 and P8 at 20.40 CA. However, this auto-ignition event produces very strong local over-pressure with the pressure amplitude of 20 MPa, where the interactions of pressure wave and cylinder wall are considered.

Fig. 10 shows the evolutions of flame propagation and autoignition development at ON = 60. Compared with the case of ON = 90, the auto-ignition timing of current case is significantly advanced due to the enhancement of low-temperature chemical reactivity. When the piston moves to 5.10 CA, several autoignition centers firstly appear under exhaust valves due to the local relatively higher temperature compared with other parts of combustion chamber. As the decreases of octane number, the effects of temperature inhomogeneity on auto-ignition timing will become more obvious, as shall be shown later. However, these early hot-spots auto-ignitions do not result in obvious local overpressure, as shown by the pressure difference at 5.10 CA. As time passed by, several new auto-ignition centers take place along the cylinder wall close to the points of P5 and P6 during 5.30– 5.60 CA. These auto-ignition centers interact with each other and produce very strong burning intensity, as shown by the high levels of OH mass fraction and pressure difference. As a consequence, these auto-ignition centers merge together and produce a fast propagating reaction front moving from exhaust valves side to intake valves side. This smooth reaction front propagates forward at a supersonic speed of approximately one thousand meters per second, which is lower than that of the steady detonation wave under the same thermodynamic conditions. Meanwhile, this auto-ignition front is companied with a strong shock wave with the pressure amplitude over 20 MPa before interacting with cylinder wall, as shown by pressure difference counters at 5.8 CA. All these observations suggest that this auto-ignition front develops into a developing detonation. Further decreases in octane number to ON = 40, there are some new observations during the super-knock, as shown in Fig. 11. Firstly, due to the enhancement of low-temperature chemical reactivity, auto-ignition events occur at not only exhaust valves side, but also intake valves side. Such that there are two reaction fronts from local auto-ignition propagating towards each other. Secondly, the early auto-ignition directly initiate a fast-propagating reaction wave with the maximum flame speed beyond one thousand meters per second, even though it is still lower than the steady detonation wave (approximate 1800 m/s) under the same thermodynamic conditions. The magnitude of the strong pressure wave accompanying auto-ignition reaction front reaches over 20 MPa, which indicates the formation of developing detonation wave. Meanwhile, the interactions of strong pressure waves and cylinder wall will induce more intense local over-pressure, which may also contribute to the damage mechanism of combustion chamber during super-knock.

J. Pan et al. / Applied Energy 191 (2017) 183–192

189

Fig. 9. Evolutions of flame propagation and auto-ignition by temperature, species mass fraction and pressure difference at ON = 90.

Fig. 10. Evolutions of flame propagation and auto-ignition by temperature, species mass fraction and pressure difference at ON = 60.

From the above discussions, we have clearly demonstrated the effects of octane number on local auto-ignition and strong pressure wave formation. However, the roles of different hot-spots auto-

ignition initiation and development in knocking intensity have not been fully revealed. Fig. 12 further shows the comparisons of auto-ignition scenarios of unburned mixture for ON = 60 and 40

190

J. Pan et al. / Applied Energy 191 (2017) 183–192

Fig. 11. Evolutions of flame propagation and auto-ignition by temperature, species mass fraction and pressure difference at ON = 40.

cases, which are shown by the contours and isosurfaces of temperature. For the case of ON = 60, it is observed that there are several auto-ignition centers from multiple hot-spots located along cylinder wall. Because these auto-ignition centers possess the same physical–chemical characteristics (e.g. burning area and autoignition timing), the subsequent auto-ignition reaction fronts interact significantly with each other in the near-wall region. Eventually, a steady auto-ignition reaction front with negative curvature is formed at 5.6 CA. While for the case of ON = 40, initially there is only two or three discretized auto-ignition centers with large difference in physical–chemical characteristics. A fast propagating auto-ignition reaction front is firstly induced by single hotspot and presents a positive curvature. Combined with earlier observations, it can be obtained that the developing detonation wave induced by multiple hot-spots auto-ignitions shows significant impacts on knocking intensity, however, the one directly by single hot-spot auto-ignition leads to much stronger knocking intensity, especially when there are collisions between several developing detonation waves during super-knock. This calculation results are consistent with the optical results of detonation experiments in rapid compression machines [11]. 4. Conclusions In present work, a LES framework coupled with detailed chemistry solver and SGS eddy viscosity model were performed to investigate sporadic auto-ignition induced abnormal combustion in a highly downsized SI engine. Classical knocking cycles fueled with iso-octane were firstly calculated, which further validated the feasibility of current model for the prediction of abnormal combustion. Then for a given spark-ignition timing, the octane number of Primary Reference Fuels (PRF) was varied to obtain different

levels of low-temperature chemical reactivity of mixture. In this way, super-knock events with different knocking intensities were observed. The results of classical knocking cycles show that as the advance of spark-ignition timing, the events of normal combustion to knocking combustion have been identified. Both knocking intensity and knocking onset are varying proportionally with the advance of spark-ignition timing. Further end-gas auto-ignition scenarios show that compared with single hot-spot auto-ignition, a homogeneous auto-ignition can lead to stronger local overpressure and thereby knocking intensity. However, super-knock events have not been obtained only through adjusting sparkignition timing. The results of super-knock cycles show that as the decreases of octane number, knocking onset is significantly advanced due to the enhancement of low-temperature chemical reactivity, as shown by cool flame temperature rise and accumulations of intermediate species. However, knocking intensity does not always show a proportional relation with octane number. The auto-ignition scenarios show that as the decreases of octane number, more auto-ignition centers appear in unburned mixture, and these auto-ignition centers are located at not only hot exhaust valve side, but also cool intake valve side. All these variations contribute to the promotions of knocking intensity of super-knock. Further analysis on different auto-ignition scenarios shows that knocking intensity also depends on complex hot-spots autoignition and subsequent reaction wave development. The developing detonation wave induced by multiple hot-spots auto-ignitions has significant impacts on knocking intensity, however, the one directly by single hot-spot auto-ignition leads to much stronger knocking intensity, especially when there are collisions between several developing detonation waves during super-knock.

J. Pan et al. / Applied Energy 191 (2017) 183–192

191

Fig. 12. Comparisons of auto-ignition initiation and development of unburned mixture between (a) ON = 60 case and (b) ON = 40 case during super-knock.

The present study aims to investigate the effects of engine parameter and fuel property on engine knock and super-knock in terms of local hot-spots auto-ignition. Such a way can facilitate future studies on the regulation of auto-ignition induced abnormal combustion, with further guidance towards the practical situations including fuel properties, thermodynamic conditions and flow characteristics. Acknowledgement This work was supported by Major Research Plan of the National Natural Science Foundation of China (91641203), National Natural Science Foundation of China (Grant No. 51476114). Pan is supported by China Postdoctoral Science Foundation (2016M590201) and 2016 Industry, Education and Research Foundation of Tianjin University.

References [1] Dahnz C, Spicher U. Irregular combustion in supercharged spark ignition engines-pre-ignition and other phenomena. Int J Eng Res 2010;11:485–98. [2] Pan J, Shu G, Wei H. Interaction of flame propagation and pressure waves during knocking combustion in spark-ignition engines. Combust Sci Technol 2014;186:192–209. [3] Peters N, Kerschgens B, Paczko G. Super-knock prediction using a refined theory of turbulence. SAE Int J Eng 2013;6:953–67. [4] Pan J, Shu G, Zhao P, Wei H, Chen Z. Interactions of flame propagation, autoignition and pressure wave during knocking combustion. Combust Flame 2016;164:319–28. [5] Robert A, Richard S, Colin O, Poinsot T. LES study of deflagration to detonation mechanisms in a downsized spark ignition engine. Combust Flame 2015;162:2788–807. [6] Wei H, Feng D, Pan M, Pan J, Rao X, Gao D. Experimental investigation on the knocking combustion characteristics of n-butanol gasoline blends in a DISI engine. Appl Energ 2016;175:346–55. [7] Bellis V. Performance optimization of a spark-ignition turbocharged VVA engine under knock limited operation. Appl Energ 2016;164:162–74.

192

J. Pan et al. / Applied Energy 191 (2017) 183–192

[8] Irimescu A, Merola SS, Valentino G. Application of an entrainment turbulent combustion model with validation based on the distribution of chemical species in an optical spark ignition engine. Appl Energ 2016;162:908–23. [9] Kawahara N, Tomita E. Visualization of auto-ignition and pressure wave during knocking in a hydrogen spark-ignition engine. Int J Hydrogen Energ 2009;34:3156–63. [10] Vafamehr H, Cairns A, Sampson O, Koupaie MM. The competing chemical and physical effects of transient fuel enrichment on heavy knock in an optical spark ignition engine. Appl Energ 2016;179:687–97. [11] Wang Z, Qi Y, He X, et al. Analysis of pre-ignition to super-knock: hotspotinduced deflagration to detonation. Fuel 2015;144:222–7. [12] Zhen X, Wang Y, Zhu Y. Study of knock in a high compression ratio SI methanol engine using LES with detailed chemical kinetics. Energ Convers Manage 2013;75:523–31. [13] Zhou D, Yang W, An H, Li J. Application of CFD-chemical kinetics approach in detecting RCCI engine knocking fuelled with biodiesel/methanol. Appl Energ 2015;145:255–64. [14] Lecocq G, Richard S, Michel JB, Vervisch L. A new LES model coupling flame surface density and tabulated kinetics approaches to investigate knock and pre-ignition in piston engines. P Combust Inst 2011;33:3105–14. [15] Misdariis A, Vermorel O, Poinsot T. LES of knocking in engines using dual heat transfer and two-step reduced schemes. Combust Flame 2015;162:4304–12. [16] Rudloff J, Zaccardi JM, Richard S, Anderlohr JM. Analysis of pre-ignition in highly charged SI engines: Emphasis on the auto-ignition mode. P Combust Inst 2013;34:2959–67. [17] Bradley D, Kalghatgi GT. Influence of autoignition delay time characteristics of different fuels on pressure waves and knock in reciprocating engines. Combust Flame 2009;156:2307–18. [18] Bozza F, Bellis V, Teodosio L. Potentials of cooled EGR and water injection for knock resistance and fuel consumption improvements of gasoline engines. Appl Energ 2016;169:112–25. [19] Wei H, Zhu T, Shu G, Tan L, Wang Y. Gasoline engine exhaust gas recirculation – a review. Appl Energ 2012;99:534–44.

[20] Senecal PK, Pomraning E, Richards KJ. Multi-dimensional modeling of direct injection diesel spray liquid length and flame lift-off length using CFD and parallel detailed chemistry. SAE technical paper 2003-01-1043; 2003. [21] Richards KJ, Senecal PK, Pomraning E. CONVERGE users guide & reference manual (Version2.1.0). Middleton, WI: Convergent Science, Inc.; 2013. [22] Turns SR. An introduction to combustion. McGraw-Hill, Inc.; 1996. [23] Yoshizawa A, Horiuti K. A statistically-derived subgrid-scale kinetic energy model for large eddy simulation of turbulent flows. J Phys Soc Jpn 1985;54:2834–9. [24] Menon S, Yeung PK, Kim WV. Effect of subgrid models on the computed interscale energy transfer in isotropic turbulence. Comput Fluids 1996;25:165–80. [25] Liu YD, Jia M, Xie MZ, Pang B. Enhancement on a skeletal kinetic model for primary reference fuel oxidation by using a semidecoupling methodology. Energ Fuel 2016;26:7069–83. [26] Wang H, Yao M, Reitz RD. Development of a reduced primary reference fuel mechanism for internal combustion engine combustion simulations. Energ Fuel 2013;27:7843–53. [27] Li Y, Jia M, Liu Y, et al. Numerical study on the combustion and emission characteristics of a methanol/diesel reactivity controlled compression ignition (RCCI) engine. Appl Energy 2013;106:184–97. [28] Liu YD, Jia M, Xie MZ, Pang B. Improvement on a skeletal chemical kinetic model of iso-octane for internal combustion engine by using a practical methodology. Fuel 2013;103:884–91. [29] Huang Y, Sung CJ, Eng JA. Laminar flame speeds of primary reference fuels and reformer gas mixtures. Combust Flame 2004;139:239–51. [30] Porpatham E, Ramesh A, Nagalingam B. Investigation on the effect of concentration of methane in biogas when used as a fuel for a spark ignition engine. Fuel 2008;87:1651–9. [31] Pan J, Wei H, Shu G, Chen Z, Zhao P. The role of low temperature chemistry in combustion mode development under elevated pressures. Combust Flame 2016;174:179–93.