lable at ScienceDirect

Journal of Loss Prevention in the Process Industries 39 (2016) 141e151

Contents lists avai

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier .com/locate/ j lp

Effect of pipe configurations on flame propagation ofhydrocarbonseair and hydrogeneair mixtures in a constant volume

Sina Davazdah Emami a, b, Siti Zubaidah Sulaiman b, Rafiziana Md. Kasmani b, **,Mahar Diana Hamid a, *, Che Rosmani Che Hassan a

a Chemical Engineering Department, Faculty of Engineering, University of Malaya, 50603 UM Kuala Lumpur, Malaysiab Energy Management Research, Department of Renewable Energy Engineering, Faculty of Petroleum and Renewable Energy Engineering, UniversitiTeknologi Malaysia, 81310 Skudai, Johor, Malaysia

a r t i c l e i n f o

Article history:Received 4 August 2015Received in revised form22 October 2015Accepted 8 November 2015Available online 30 November 2015

Keywords:BendingTee junctionFlame propagationRate of pressure rise

* Corresponding author.** Corresponding author.

E-mail addresses: rafiziana@petroleum.utm.my (R.um.edu.my (M.D. Hamid).

http://dx.doi.org/10.1016/j.jlp.2015.11.0050950-4230/© 2015 Published by Elsevier Ltd.

a b s t r a c t

It is commonplace in industrial installations to have different piping arrangements for the efficienttransport of substances/materials. However, the processing industry has raised some major concerns interms of safety, due to the accidental gas explosions that have occurred frequently and caused seriousdamage. It is the aim of this study to comprehensively analyse the governing factors involved in flamepropagation inside different pipe configurations. This research investigates confined pipe explosionsusing straight, 90-degree bending, and tee-junction pipes with different obstacle placements. Hydrogen-,ethylene-, propane- and natural gaseair mixtures, over a range of concentrations (equivalence ratio,Ф ¼ 0.6e1.4) have been used. The results show that, while there is no significant difference in themaximum pressure and rate of pressure rise in both tee-pipe arrangements investigated, the bendingpipe consistently produces the worst set of results in terms of maximum pressure and flame speed in gasexplosions, involving the most reactive mixtures. In addition, the detailed records of pressure traces andblast waves show that the duration of flame acceleration, the flame direction and the initial ignition pointdepend on the tee junction placement along the pipe length, resulting to different overall profile of theflame acceleration mechanism.

© 2015 Published by Elsevier Ltd.

1. Introduction

Explosions in the chemical, gas and petroleum industries arestill a significant problem, leading to injuries, death, the destructionof equipment, and downtime. In the chemical, hydrocarbon andplant process industries, we can find a large variety of scenarios inwhich internal gas explosionsdconfined or unconfined explo-sionsdmay occur. Such explosions can be caused by uncontrolledleaks, or simply by accidental purging with air or unpredictablefailures (Grossel, 2010). For underground coal mine, coal dust ex-plosion caused by gas explosion often cause secondary disaster andsuch an accident can bring more severe disasters than single-phasegas explosion. Moving at the speed of sound, pressure waveresulting from gas explosion lifts the deposited coal dust in the air,

Md. Kasmani), mahar.diana@

causing a dust explosion which is more severe than the first one(Beidaghy Dizaji et al., 2014; Bidabadi et al., 2015, 2014; Bidabadiaet al., 2013; Soltaninejad et al., 2015). As a consequence, there is aneed for pipeline protection against the propagation of unwantedcombustion phenomena, such as deflagration to detonation trans-mission (DDT) (including decomposition flames), occurring in theprocess (Blanchard et al., 2010; Grossel, 2010).

In order to ensure that better precautions are taken in relation topipeline gas carriers, it is essential to fully characterize and quantifytheir explosion mechanisms. In particular, knowledge is requiredabout the maximum pressure, the maximum rate of pressure rise(i.e., deflagration index) and the flame speed, which are among themost important parameters for the risk assessment of processhazards and the safer design of process equipment (Hawkes andChen, 2004; Salzano et al., 2012). Studies on flame propagationand explosionmechanisms in the pipes have beenwidely discussed(Blanchard et al., 2010; Chatrathi et al., 2001; Emami et al., 2013;Jianliang et al., 2005; Oh et al., 2001; H Phylaktou et al., 1993;Thomas et al., 2010), but most of these are focused on specific

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151142

experimental configuration and applications (i.e., using eitherbending or straight pipes/tubes). For this reason, holistic studiesshould be performed on flame propagation in different pipe con-figurations, considering the complicated problems involved in theinteraction between fluid dynamics, heat transfer and turbulentcombustion. For instance, Zhu et al. (2010) used a single-bend, U-shaped pipe and a Z-shaped pipe in their experimental setup, toinvestigate the effect of roadway turning on methane-air explosionpropagation. The results showed that, by increasing the number ofturnings, the explosion strength was significantly enhanced, whilethe flame speed and peak overpressure increased dramatically. Inaddition, the values of flame speed and maximum overpressure areenhanced by increasing the concentration of methane in themixture in a horizontal pipe (Zhu et al., 2010). Another study on astraight pipe with a 60 L/D (length/diameter) showed that thepressure and velocity waves accelerate with the increasing reactionrate of methane-air (Zhao et al., 2015). Investigation on flame re-gion distribution and hazard effects in a tube and a tunnel gasexplosion also showed that the flame region is always longer thanthe original gas region in any cases, by only considering the initialconcentration of mixture (Ma et al., 2015).

On the other hand, Razus et al. (2006) have studied the explo-sion pressures of hydrocarbon-air mixtures in closed vessels. Theyshowed that the initial pressure, fuel concentration and heat losseshave a significant effect on the maximum overpressure duringflame propagation. Gu et al. (2000), and Liao et al. (2004), haveinvestigated the flame instability and ratio of the laminar burningvelocity of methane-air and natural gaseair mixtures at differentequivalence ratios. Liu et al. (2015) have studied flame propagationand explosion development in propane-air mixtures, in a 1.16 m3

vessel with central ignition, to evaluate the burning velocity of thefuel by considering the history of flame-front trajectory and pres-sure in the vessel. From their work, flame instability was observedat the equivalence ratio of 1.2 and above, suggesting that the flameinstability is due to the effect of thermal-diffusion instability andhydrodynamic instability. Rich mixtures are known to be moresusceptible to developing surface instabilities (flame cellularity),which can lead to higher burning rates and hence higher flamespeed. This in turn could result in a more severe explosion thanmight otherwise be expected (on the basis of its laminar burningvelocity alone).

On the determination of ethylene explosion severity, a num-ber of experimental and numerical studies on flammableethylene have been conducted (Kumar et al., 2007; Movileanuet al., 2011a, 2011b). In the oxidation of higher hydrocarbons,ethylene is among the key intermediates motivating researchersto apply the numerical methods for kinetic modelling, in order tofind a suitable mechanism for ethylene oxidation at a wide rangeof temperatures, pressures, and equivalence ratios (Bergthorsonand Dimotakis, 2007; Egolfopoulos et al., 1991; Jomaas et al.,2005). However, for the fuel transfer process, only a limitednumber of studies have been carried out. An investigation, con-ducted by Thomas et al. (2010), showed that the transition todetonation in pure ethylene can sustain a detonation by adecomposition reaction at pressures greater than atmosphericpressure. They also reported that the initial pressure does notplay an important role in increasing the overall pressure; how-ever, the initial pipe wall temperature, and (possibly) the mixturehumidity, could affect the overall flame propagation mechanism,as similarly observed by Blanchard et al. (2011) and Ma et al.(2015). They showed that there is no possibility for this partic-ular gas to promote shock waves since its over-driven detona-tions were not strong enough to enable confident measurementof its velocity. They concluded that, after DDT, flame speed de-creases during the transition process in both straight and bent

pipes with a 159 mm diameter. However, the maximum flamespeed was observed at approximately 80% of the straight pipelength, and at approximately 70% if obstacles were present inthese two configurations. This phenomenon gives a strong indi-cation that the reflected pressure waves from the closed-endpipe do have a significant effect in slowing down the flamefront during the flame propagation.

On the other hand, Xiao et al. (2011) have carried out anexperimental study on half-open and closed horizontal ducts, andfound that premixed hydrogeneair mixtures undergo differentphases of flame shapes, indicating pronounced characteristicscompared to other gaseous fuels. Research on the estimation ofshock waves in hydrogen-oxygen mixtures in a 12-m diametervolume has also shown that wave intensification from a small[initial] amount of energy could create secondary combustion ex-plosion centres, whose parameters exceed the values predicted bythe Chapman-Jouguet condition (Petukhov et al., 2009). This con-tradicts the normal assumption that detonation is stimulated by asignificant power effect. However, experimental researchers reporta number of common findings, including the fact that when anexplosion uses hydrogen fuels, DDT has the potential to be achievedat a magnitude of greater severity, compared to hydrocarbon fuels(Heidari and Wen, 2014; Thomas et al., 2010).

The numerical simulation and experimental study of flamepropagation in a duct with a 90-degree curved section has alsoconsistently reported a good agreement with the basic physicalphenomena, such as the tulip flame, flame shedding, pressureevolution trends, flame propagation speed trends, and vortexdevelopment in the bend (Emami et al., 2013; Zhou et al., 2006).This previous work has discussed and illustrated the fact that theunburned mixture flow development in the bend was marked byan embedded transient secondary flow, in the form of two ormorestream-wise vortices (Zhou et al., 2006). Hu et al. (2009) have alsoshown that, with the increase of the equivalence ratio, the laminarburning velocity increases for fuel-lean mixture combustion anddecreases in the case of fuel-rich mixture combustion. Thelaminar burning velocity also intensifies the increases of initialtemperature and pressuredan observation that similar in-vestigations using different vessels have also consistently sup-ported (Bauwens et al., 2012; Dahoe, 2005). However, what hasbeen lacking is a comprehensive study of the governing parame-ters involved in flame propagation in different pipe configura-tions, by considering the physics and dynamics of the flame andpressure development of hydrocarbonseair and hydrogeneairexplosions in a wide range of equivalence ratios. For this reason,the physics and dynamics of explosion development is investi-gated in different pipe configurationsdi.e., closed-ended straight,90-degree and tee pipelinesdwhere the behaviour of the flamepropagation and explosion characteristics of the fuel-air mixturesare observed in order to establish an appropriate vessel and pipingdesign for safer application. An analytical analysis of the resultsobtained in this study is presented with respect to overpressure,rate of pressure rise, flame speed of premixed gas mixture ex-plosions, and the effect of the ignition position.

2. Methodology

In this study, straight, 90-degree, and tee pipes with a constantvolume (0.042m3) were used, with a 0.1 m initial diameter (refer toFig. 1). Both pipe ends were closed. The pipes were made up of anumber of segments, ranging from 0.5 to 1 m in length, boltedtogether with gasket seals between the connections and blindflanges at both ends. The flammable mixtures were ignited by anelectrical spark, which gave 16 J of energy to the gas explosion tests.For this study, four pipe configurations were considered,

Fig. 1. Scheme of pipes: (a) straight; (b) 90-degree; (c) and (d) tee.

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151 143

designated as A-D, as shown in Fig. 1 C and D are tee junctions withdifferent obstacle locations.

The flame speed measurements were recorded using exposed-junction, mineral-insulated, K-type thermocouples, which werepositioned along the centre line of the pipe. The flame speed datawere generated from the flame arrival times (marked as an abruptchange in the thermocouple output). The flame arrival time in thepipe was taken to be the first point at which the reading began torise. With regard to the thermocouples in the pipe, this was hin-dered by a pre-compression wave ahead of the flame (and theassociated high flow velocity around the thermocouple), whichgave rise to two distinct gradients on the thermocouple trace. Inthis case, the point at which the second (steeper) gradient becameapparent was taken as the flame arrival time. The pressure withinthe pipe was monitored using an array of piezoresistive pressuretransducers along the outer wall, denoted as P in Fig. 1. The datagenerated were collected using a 34-channel transient datarecorder (by NI CompactDAQ). To obtain more information aboutthe accuracy and reproducibility of the detected parameters, pleaserefer to the research study by Wang et al. (2007). The ignition

positions and the location of each of the sensors are presented inTable 1.

For the fuel-air mixtures, NG-air, C2H4-air, C3H8-air and H2-airmixturesdwith equivalent ratios (f) of 0.6, 0.8, 1, 1.2 and 1.4dwereused in this study. In this experiment, the partial pressure methodwas employed for the mixture preparations, by controlling theinitial concentration of each injected gas via a digital pressuregauge with the accuracy ±0.01 mbars. The injected gases were leftfor 10 min in each run to achieve a homogeneous composition. Foraccuracy and reproducibility, each explosion was repeated at leastthree times on each mixture.

3. Results and discussion

3.1. The effect of maximum overpressure on the pipe configuration

Fig. 2 illustrates the recorded maximum overpressures for allfuels. The pressure traces are presented with respect to times. Forhydrocarbon explosions, typical explosion profiles were shown,giving a highest maximum pressure value at the bending region

Table 1Position of each pressure transducer and thermocouple from the ignition points.

Ignition position (m)

A B C G

Pressure transducers P1 0.32 0.32 0.26 0.26P2 1.35 1.35 1.18 1.18P3 2.02 2.02 2.53 2.16P4 2.58 2.58 3.42 3.08P5 3.33 3.33 3.9 4.07P6 4.16 4.16 4.03 2.77P7 e 4.44 4.52 3.26

Thermocouples T1 0.26 0.26 0.45 0.45T2 0.705 0.75 1.26 1.26T3 1.065 1.61 1.65 1.65T4 1.485 2.17 2.75 2.06T5 2.14 2.62 3.42 3.13T6 3.2 3.48 3.99 3.9T7 4.345 4.26 4.12 2.73T8 e e 4.38 2.99

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151144

when ignited at the rear positions. From the hydrogeneair pressureresults, it was found that the maximum overpressure of a hydrogenexplosion is greater than that of hydrocarbon fuels: twice as high atall pipe configurations. Surprisingly, the maximum pressure ofhydrogeneair mixtures experienced its highest value (~7.2 bars) atthe bending region, rather than the tee junction position for boththe longer obstacles position and the equal distance position(2.22 bars and 2.67 bars, respectively).

From previous investigations, the burning area was found toincrease when the flame front travelled through the 90-degreebend (Clanet and Searby, 1996; Zhou et al., 2006). Consequently,the flame speed also increases, thus giving a higher overpressurewhen compared to similar experiments carried out in a straightpipe and tee junction (Blanchard et al., 2010; Cheng et al., 2009;Emami et al., 2013; H Phylaktou et al., 1993). Fig. 2d shows theoverpressure development along the pipe at different positions forhydrogeneair mixture. It is clearly shown that the two over-pressure peaks were observed at two different positionsdfirst, atthe bending position, and second, near the endwall (0.9m from thepipe's end wall)dfor the 90-degree pipe explosions. There is nodoubt that the first pressure rise is due to the influence of anobstacle: in this case, the presence of the 90-degree bend in theclosed pipe. Strong reflection and diffraction acoustic/shock wavesenter into the reaction zone, enhancing the burning rate andincreasing the flame speed. This phenomenon causes the over-pressure to increase by about 7.2 barg (at f ¼ 1). It is a known factthat the presence of an obstacle in the pipewill randomize the flow,thus increasing the flame speed and overpressure when comparedto the straight pipe/tube (Phylaktou et al., 1993). However in aclosed pipe/tube, the end wall also acts as an obstacle with a pro-pensity to initiate flame perturbation, and hence, affects the ex-plosion behaviour (Liberman et al., 2010; Thomas et al., 2010; Zhuet al., 2012). Fig. 3a also shows the pressure development asinfluenced by the distance from the ignition position, for ethylene-air mixture. It shows that the pressure decayed after the bendingposition, which is inconsistent with other fuel explosion profiles.For propane and natural gas (NG) explosions, the pressure roseafter the bend/obstacle, either in the 90-degree or tee pipes. It issuspected that the quenching effect of the preferential-diffusionmechanism promoted the flame dissipation at the pipe wall, giv-ing a lower overall adiabatic temperature inside the pipe. Accordingto Aung et al. (2002), the turbulent flame propagation decreasesprogressively with the increase of the preferential-diffusion region.The preferential-diffusion concept is relative to the flameestretchinteraction. Thus, it can be suggested that, when the preferential-

diffusion effect is apparent, it tends to retard the flame distortionthrough excessive flame stretching and causes the flame to partiallyquench (Bradley et al., 2008). Due to the various flame instabilitymechanisms involved in the flame propagation (e.g., hydrodynamicinstability, thermal diffusion, Darrieus-Landau), turbulent flamealso develops as a result of interactions between the flame frontand the acoustic waves (Bradley et al., 2008; Ciccarelli andDorofeev, 2008; Gamezo et al., 2008; Liberman et al., 2010). How-ever, the findings confirm that the maximum overpressure attainedat the bending position for most fuels (except for propane and NG)are as shown in Fig. 3a, b, c and d. This inconsistent trend can beexplained based on fuel reactivity itself. The diffusivity of propaneis significantly smaller than its thermal conductivity, allowing thefuel to be preferentially heated more rapidly at the preheated zone(Aspden et al., 2011). For propane, the highest maximum pressurewas observed at x ¼ 2.02 m (P3), before the 90-degree bend posi-tion shown in Fig. 3c. At the same time, the acoustic/shock wave-dwhich was diffracted and reflected due to the bendingeffectdtravelled back toward the hot flames to amplify the burningrate. Due to the slower burning rate and heat losses to the wall, theflame tends to quench after point P3. However, a second acceler-ation was observed at a distance of 4.16 m from the ignition posi-tion (P5), due to the end wall's effect relative to the flame-reflectivewave interaction. For NG-air mixtures, the pressure at P3 wasapproximately 0.41 barg, as exhibited in Fig. 2. The second peakoverpressure was shown to be located at P5 (after the bendingposition), due to the flameewave interaction, which was compa-rable to the other fuels. This could suggest that the effect of bendingwas more pronounced in the NG-air explosions. As expected, themethane-air mixture exhibited the lowest pressure developmentalong the pipe, because of its lower reactivity behaviour.

The results clearly show that, in the straight pipe (refer to Aconfiguration in Fig. 1), the trend is consistent for all fuels, with themaximum overpressure occurring at a distance of x ¼ 2.02 m (P3)from the ignition position. The associated maximum overpressure,resulting from the highest burning rate, is due to the flame cellu-larity/wrinkling, i.e., the distortion of the flame surface increasingthe flame area and, hence, enhancing the burning velocity. In thisinstance, the tulip formation attenuates the turbulence, due to itsvortex creation, and the reflective wave promotes a strong inter-action between the fast flame and turbulence, which increases theflame speed, followed by a pressure rise. Ma et al. (2015) reportedsimilar observation on the flame acceleration of methane/air in atube. However, at x > 2.02 m (P4 onwards), the trend was incon-sistent for reactive fuels. For instance, hydrogeneair gave a gradual

Fig. 2. Maximum overpressure vs. the recorded points.

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151 145

pressure development, yet there were no similar observations forethylene-air, propane-air or methane-air explosions, which dis-played an abrupt pressure drop to around two to five times lowerthan the overpressure at x ¼ 2.02 m. This inconsistent trend can berelated to the nature of fuel reactivity. This will be discussed furtherin the next section.

Within the examined range of hydrocarbonseair and hydro-geneair concentrations, the value of the maximum overpressurewas almost the same for both tee pipe configurations (C and D),only it was attained at different locations. As Fig. 2 shows, thehighest overpressure for hydrogeneair mixture was observed atthe tee junction region (P7) for the C configuration, whereas P5gave the maximum peak pressure for the tee pipe D. As seen, the

propagated flames in tee pipes C and D experienced a higher ac-celeration rate after the tee region, leading to a higher pressure atthe end pipe. As mentioned earlier, the flames for the variousmixtures were susceptible to flame wrinkling, particularly for themore reactive fuels, due to the diffusional-thermal instability ef-fect. Consequently, this wrinkled flame causes an increase in themass burning rate, giving a higher overall flame speed and over-pressure intensity (Kim et al., 2014). From Fig. 3d, we can see thatit took a shorter amount of time for the hydrogen flames to reachthe maximum overpressure, towards the end pipe, giving a netresult of a higher pressure rise and higher flame speed when theexplosion occurred inside the pipe C configuration. On the otherhand, when the mixtures were exploded inside the tee pipe D

Fig. 3. Maximum flame speed vs. the recorded points.

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151146

configuration, the flame propagation of this particular mixturewas inconsistent with the others. The appearance of multiple peakoverpressures indicates that, at a shorter distance between igni-tion point and obstacles, the flame experiences a strong interac-tion with the transverse pressure waves from the tee junctioneffect, inducing more turbulence and, consequently, attenuatingthe acceleration of the burning of the unreacted mixtures, whichwere trapped along the tee junction region. This again, will in-crease the combustion rate and thus, increase the flame speed andoverpressure. The findings of this study are in good agreementwith previous studies (Kim et al., 2014). It appears that

hydrocarbonseair flames have the most minimal fluctuationsalong the tee pipes, both in terms of overpressure and flame speed(Figs. 2 and 3). However, the relationship between the overallmaximum pressure and its explosion severity on both tee pipeconfiguration, can be substantiated by the unburned gas velocity,Sg profile, as shown in Fig. 4. The Sg for both the tee pipe config-urations was at an average of 208.18 m/s, indicating that there isno possibility of these particular gases presenting shock wavessince their overdriven detonations were not strong enough(Blanchard et al., 2011). This statement is also in agreement withprevious investigations on the flame acceleration of hydrocarbon

Fig. 4. Unburned gas flame speed vs. the recorded points.

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151 147

fuels (Cho and He, 2007; Kumar et al., 2007; Movileanu et al.,2011a, 2011b; Prince and Williams, 2012). Moreover, the plottedblast waves of the flames at multiple locations (given in Fig. 5) alsosuggest a linearity of flame acceleration for the hydrocarbonseairmixtures in tee pipes. It is interesting to note that the duration offlame acceleration, the flame direction and the initial ignitionpoints all depended on the tee junction placement along the pipelength, resulting to different overall profiles on the flame accel-eration mechanism in general.

From the analysis and discussion, it can be said that flamepropagation and pressure development pose a greater severity in90-degree and straight pipe explosions, when compared to tee pipeexplosions.

3.2. Flame speed inside the pipes

Fig. 3a, b, c and d illustrate the flame speed as a function of flamearrival points, represented by the thermocouple positions (denotedby T1eT8 in Fig. 1). For the straight pipe, natural gas, propane andethylene gave consistent flame speed profile: i.e., the maximumflame speed occurred at x ¼ 1.0 m from the ignition position (T3).However, two peaks were observed in the hydrogen flame speedfigures, occurring at x ¼ 1.48 m (T4) and 3.20 m (T7) from theignition position, as shown in Fig. 3d. It should be noted that thelaminar burning velocity (SL) of hydrogen, at a stoichiometricconcentration (F ¼ 1.0), is approximately 3.15 m/s (Harris, 1983),and the adiabatic spherical flame speed is 28 m/s (Alekseev et al.,

Fig. 5. Blast waves.

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151148

2014). The measured flame speeds were much higher than SL, andat least seven times higher than the adiabatic flame speed. Thesignificant increase in flame speed could be due to the self-acceleration of the flame resulting from the thermal-diffusive in-stabilities of the flame-front mechanism. The first peak reached aspeed of 172.4 m/s, an increment of approximately 13 times theinitial speed, indicating a very high turbulence experienced in thepipe. This flow creates turbulence ahead of the flame, inducing theflame to speed up rapidly. Coupled with the formation of flamewrinkling, flame folding and vorticity generation, the rapid flamepropagation would draw the flame expansion preferentially in thedirection of the end of the pipe. Furthermore, the reflective wavefrom the end of the pipe would attenuate the mass burning rate atthe maximum, resulting in a rise to the second peak at ~1000 m/s,

at T7 from the ignition position. It can be suggested that the flamespeed development, seen between the ignition point and a distanceof x ¼ 2.14 m (T5), is due to the unsteady flow caused by the tulipflame phenomenon, the quenching effect and the flameewaveinteraction in confined pipes (Ciccarelli and Dorofeev, 2008). Thehigh flame speed for hydrogeneair were due to the flame's self-acceleration, resulting from the interaction between the shockwave, the fast flame and the flame front during the onset of avortex, in accordance with the baroclinic effect (Petchenko et al.,2007). Such vortices induce the stretching of the flame surfacearea and cause the flame to cellularize/distort, intensifying theturbulence flame. To the contrary, for ethylene-air, propane-air andmethane-air mixtures, the flame speed had decreased slightly atx ¼ 1.5 m (T5), before propagating at a constant velocity towards

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151 149

the end pipe. It is suspected that the flame starts to quench,resulting in a weaker flame turbulisation.

The recorded flame speed in the 90-degree pipe indicated thatthe presence of bendingwould promote the flame stretch around it,intensifying the turbulence vortex and thus increasing the flamesurface area through the strong mixing of burned and unburnedgas. As Fig. 3d shows, the peak flame speed of hydrogeneair wasapproximately 1384 m/s: a steep increase of up to 4.8 times beforeentering the bend. Meanwhile, ethylene-air, propane-air andmethane-air experienced maximum flame speed of about 496 m/s,407 m/s and 339 m/s, respectively, before the bending position.Furthermore, it also shows that the second acceleration occurred at3.49 m (T6) from the ignition point, giving an increase of flamespeed to 1311 m/s, 487 m/s, 407 m/s and 310 m/s for hydrogeneair,ethylene-air, propane-air and methane/air, respectively, beforereaching the end wall. The increased flame speed at this positionwere approximately twice as high as the flame speed at distance2.62 m (T5). This is possibly caused by the complicated interactionbetween flame acceleration, the quenching effect and flame shape,which affects themass-burning rate (Phylaktou and Andrews,1991;Phylaktou et al., 1993). It is worth noting that the interaction of hotflame with pressure waves, which are reflected from the end wall,plays an important role in contributing to the second acceleration.

For the tee pipe, the flame speed was lower than those of thestraight and 90-degree pipes, due to the free paths available afterthe tee region. An investigation of parallel pipes also showed asimilar observation (Zhu et al., 2012). However, there is no doubt ofthe explosion severity posed by the tee pipe configuration. The datafrom both tee pipe configurations (Fig.1c and d) show that a shorterdistance of the tee junction from the ignition position gives a higherseverity magnitude, compared to a longer distance of the teejunction placement. However, at a longer distance of tee junction,different flame speed profiles were recorded. The higher flamespeed observed at T3 (~500 m/s) was due to the dynamic effectexperienced by the flame propagation; thermo diffusivity, the rapidmixing of the induced turbulence, fast flame downstream, a stronginteraction between reflective waves and fast flames from the endpoints, on the other, can cause retonation and detonation, as dis-cussed by Qiao et al. (2005). When the tee junctionwas placed at anequal distance along the pipe, as shown in Fig. 1d, the flame ac-celeration of the fuel gases gave the lowest flame speed intensity,since the flames accelerate at almost the same distance after the teejunction region. In this instance, the interaction between thereflective waves and fast flames from the end points was insignif-icant; this phenomenon has been fairly discussed in the literature(Hassan et al., 2012; Kolbe and Baker, 2005).

3.3. Maximum rate of pressure rise

The maximum rate of pressure rise at all equivalence ratios andblast waves at stoichiometric concentration for all fuels are pre-sented in Figs. 5 and 6. As shown, the associated maximum rates ofpressure rise from the ignition point were recorded at the end ofthe second of the three parts of the straight pipe (P3), and at thebend point in the 90-degree pipe (P4). However, it was localized atdifferent points in the tee pipes, with respect to the tee junctionallocation. As mentioned earlier, flame cellularity causes distortionto the flame surface and this gives a larger area and thus, increasingthe ratio of burning velocity. Consequently, the tulip formationreduces the turbulence due to the vortex creation, and the reflectivewave leads to a strong interaction between the fast flame andturbulence, increasing the flame speed as well as the rate of pres-sure rise (Bougrine et al., 2014; Merlin et al., 2012). This conditioncan be substantiated by Fig. 5. The fluctuation rate at each pressuretransducer point suggests that the flame has a tendency to become

cellular as it is progressively stretched, particularly in bending andtee junction regions; giving it more room to grow larger and thus,increasing the mass-burning rate. The rate of pressure rise dataappear to be more repeatable than the maximum overpressureresults, and show that the hydrogen fuel produced a more severeexplosion for all the configurations tested. For instance, it took theshortest time for the hydrogeneair to reach maximum over-pressure in all configurations. This was not only an effect of thekinetic reaction for each mixture, but also of the dynamic proper-ties of the flame (i.e. diffusional-thermal instability also has a sig-nificant influence on this phenomenon). This is the most importantfactor affecting the burning velocity and heat release rate duringflame propagation (Movileanu et al., 2013); consequently, it isassociated with flame wrinkling. This wrinkled flame causes anincrease in the mass-burning rate, giving a higher overall flamespeed and overpressure intensity (Kim et al., 2014). This was thereason for the highest and lowest rates of pressure rise observed inhydrogeneair and propane-air explosions, respectively, and it wasshown to be more significant in the 90-degree bend pipe.

4. Conclusion

A comprehensive study of the governing parameters involved inthe flame propagation of hydrocarbonseair and hydrogeneairmixtures, in a wide range of equivalence ratios and in different pipeconfigurations, was carried out in this current research. The find-ings can be summarized as below:

1) From the hydrogeneair pressure and flame speed results,hydrogen explosions gave both higher maximum overpressureand flame speed values compared to those of the hydrocarbonfuels (approximately twice as high for all studied pipe configu-rations). However, the recorded data also showed that theassociatedmaximum overpressure and flame speed of ethylene-air were the highest among all the hydrocarbon-air fuels, fol-lowed by NG-air and propane-air, respectively, since the diffu-sivity of propane is significantly smaller than its thermalconductivity, which allows the fuel to be preferentially heatedmore rapidly at the preheated zone compared to other hydro-carbon fuels.

2) It is interesting to note that the maximum pressure of hydro-geneair mixtures experienced its highest value (~7.2 bars) at thebending region; compared to the straight pipe and both teejunction configurations (designated as C and D). Therefore, pipebends pose the highest explosion severity in terms of the rate ofpressure rise when compared to other pipe obstacles.

3) There are two distinct overpressure peaks recorded at twodifferent positions in the bent pipe explosions for all studiedmixtures: first, at the bending position, and second, near the endwall (0.9 m from the pipe end wall). The presence of a bend canbe observed to attenuate flame stretching and intensify theturbulence. This condition promotes further strong mixing ofreflection and diffraction acoustic/shock waves, entering intothe reaction zone, enhancing the burning rate and increasingthe flame speed.

4) The trend for all fuels is consistent in the straight pipe (Aconfiguration), with the maximum overpressure occurring at adistance of x ¼ 2.02 m (P3) from the ignition position. Theassociated maximum overpressure and flame speed resultingfrom the highest burning rate is due to the flame cellularity/wrinkling, i.e., the distortion of the flame surfaces to becomelarger, hence increasing the burning velocity.

5) Within the examined range of hydrocarbonseair and hydro-geneair concentrations, the values for the maximum over-pressure and maximum flame speed were almost the same for

Fig. 6. Rate of pressure rise vs. the recorded points.

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151150

both tee pipe configurations, only they were attained atdifferent locations. Moreover, the appearance of multiple peakoverpressures indicates that, at a shorter distance from the ob-stacles, the flame experiences a strong interaction with thetransverse pressure waves resulting from the tee junction effect,inducing more turbulence and, consequently, attenuating theacceleration of the substantial unburned gases trapped aroundthe tee junction region. This results in a rise in flame speed andthus, enhances the maximum pressure and rate of pressure rise.

Acknowledgement

The authors would like to express the gratefulness and appre-ciation to PPP grant (Project No: PG105-2013A), FRGS grant (ProjectNo: FP013e2014B) and UMRG grant (Project No: RP015-2012H)

from University of Malaya (UM). We also would like to show ourgratitude to Universiti Teknologi Malaysia for the research grant(No; QJ130000 2542 03H41) that have developed the explosion testfacility.

References

Alekseev, V.A., Christensen, M., Konnov, A.A., 2014. The effect of temperature on theadiabatic burning velocities of diluted hydrogen flames: a kinetic study usingan updated mechanism. Combust. Flame (0). http://dx.doi.org/10.1016/j.combustflame.2014.12.009.

Aspden, A.J., Day, M.S., Bell, J.B., 2011. Lewis number effects in distributed flames.Proc. Combust. Inst. 33 (1), 1473e1480. In: http://dx.doi.org/10.1016/j.proci.2010.05.095.

Aung, K.T., Hassan, M.I., Kwon, S., Tseng, L.K., Kwon, O.C., Faeth, G.M., 2002. Flame/stretch interactions in laminar and turbulent premixed flames. Combust. Sci.Technol. 174 (1), 61e99. http://dx.doi.org/10.1080/713712909.

Bauwens, C., Chao, J., Dorofeev, S., 2012. Effect of hydrogen concentration on vented

S.D. Emami et al. / Journal of Loss Prevention in the Process Industries 39 (2016) 141e151 151

explosion overpressures from lean hydrogeneair deflagrations. Int. J. HydrogenEnergy 37 (22), 17599e17605.

Beidaghy Dizaji, H., Faraji Dizaji, F., Bidabadi, M., 2014. Determining thermo-kineticconstants in order to classify explosivity of foodstuffs. Combust. Explos. ShockWaves 50 (4), 454e462. http://dx.doi.org/10.1134/S0010508214040145.

Bergthorson, J.M., Dimotakis, P.E., 2007. Premixed laminar C 1eC 2 stagnationflames: experiments and simulations with detailed thermochemistry models.Proc. Combust. Inst. 31 (1), 1139e1147.

Bidabadi, M., Beidaghy Dizaji, H., Faraji Dizaji, F., Mostafavi, S., 2015. A parametricstudy of lycopodium dust flame. J. Eng. Math. 92 (1), 147e165. http://dx.doi.org/10.1007/s10665-014-9769-3.

Bidabadi, M., Dizaji, F., Dizaji, H., Ghahsareh, M., 2014. Investigation of effectivedimensionless numbers on initiation of instability in combustion of moistyorganic dust. J. Central South Univ. 21 (1), 326e337. http://dx.doi.org/10.1007/s11771-014-1944-1.

Bidabadia, M., Mostafavia, S., Dizajib, H.B., Dizajia, F.F., 2013. Lycopodium Dust ameCharacteristics Considering Char Yield.

Blanchard, R., Arndt, D., Gr€atz, R., Poli, M., Scheider, S., 2010. Explosions in closedpipes containing baffles and 90 degree bends. J. Loss Prev. Process Ind. 23 (2),253e259.

Blanchard, R., Arndt, D., Gr€atz, R., Scheider, S., 2011. Effect of ignition position on therun-up distance to DDT for hydrogeneair explosions. J. Loss Prev. Process Ind.24 (2), 194e199.

Bougrine, S., Richard, S., Colin, O., Veynante, D., 2014. Fuel composition effects onflame stretch in turbulent premixed combustion: numerical analysis of flame-vortex interaction and formulation of a new efficiency function. Flow, Turbul.Combust. 93 (2), 259e281.

Bradley, D., Lawes, M., Liu, K., 2008. Turbulent flame speeds in ducts and thedeflagration/detonation transition. Combust. Flame 154 (1e2), 96e108. http://dx.doi.org/10.1016/j.combustflame.2008.03.011.

Chatrathi, K., Going, J.E., Grandestaff, B., 2001. Flame propagation in industrial scalepiping. Process Saf. Prog. 20 (4), 286e294.

Cheng, Z., Bai-quan, L., Qing, Y., Xian-zhong, L., Chuan-jie, Z., 2009. Influence ofgeometry shape on gas explosion propagation laws in bend roadways. ProcediaEarth Planet. Sci. 1 (1), 193e198.

Cho, H.M., He, B.-Q., 2007. Spark ignition natural gas enginesdA review. EnergyConvers. Manag. 48 (2), 608e618. http://dx.doi.org/10.1016/j.enconman.2006.05.023.

Ciccarelli, G., Dorofeev, S., 2008. Flame acceleration and transition to detonation inducts. Prog. Energy Combust. Sci. 34 (4), 499e550. http://dx.doi.org/10.1016/j.pecs.2007.11.002.

Clanet, C., Searby, G., 1996. On the “tulip flame” phenomenon. Combust. Flame 105(1e2), 225e238. http://dx.doi.org/10.1016/0010-2180(95)00195-6.

Dahoe, A., 2005. Laminar burning velocities of hydrogeneair mixtures from closedvessel gas explosions. J. loss Prev. process ind. 18 (3), 152e166.

Egolfopoulos, F., Zhu, D., Law, C., 1991. Experimental and Numerical Determinationof Laminar Flame Speeds: Mixtures of C 2-hydrocarbons with Oxygen and Ni-trogen. Paper presented at the Symposium (International) on Combustion.

Emami, S.D., Rajabi, M., Hassan, C.R.C., Hamid, M.D.A., Kasmani, R.M., Mazangi, M.,2013. Experimental study on premixed hydrogen/air and hydrogenemethane/air mixtures explosion in 90 degree bend pipeline. Int. J. Hydrogen Energy 38(32), 14115e14120. http://dx.doi.org/10.1016/j.ijhydene.2013.08.056.

Gamezo, V.N., Ogawa, T., Oran, E.S., 2008. Flame acceleration and DDT in channelswith obstacles: effect of obstacle spacing. Combust. Flame 155 (1e2), 302e315.http://dx.doi.org/10.1016/j.combustflame.2008.06.004.

Grossel, S.S., 2010. Deflagration and Detonation Flame Arresters, vol. 9. John Wiley& Sons.

Gu, X.J., Haq, M.Z., Lawes, M., Woolley, R., 2000. Laminar burning velocity andMarkstein lengths of methaneeair mixtures. Combust. Flame 121 (1e2), 41e58.http://dx.doi.org/10.1016/S0010-2180(99)00142-X.

Harris, R.J., 1983. The Investigation and Control of Gas Explosions in Building andHeating Plant. E&F N Spon Ltd, New York.

Hassan, Y., Mousseau, V., Sanyal, J., Durbin, S., Baglietto, E., Lucas, D., ..., Dreisbach, J.,2012. Technical Session Summaries.

Hawkes, E.R., Chen, J.H., 2004. Direct numerical simulation of hydrogen-enrichedlean premixed methaneeair flames. Combust. Flame 138 (3), 242e258. http://dx.doi.org/10.1016/j.combustflame.2004.04.010.

Heidari, A., Wen, J., 2014. Flame acceleration and transition from deflagration todetonation in hydrogen explosions. Int. J. Hydrogen Energy 39 (11), 6184e6200.

Hu, E., Huang, Z., He, J., Miao, H., 2009. Experimental and numerical study onlaminar burning velocities and flame instabilities of hydrogeneair mixtures atelevated pressures and temperatures. Int. J. Hydrogen Energy 34 (20),8741e8755. http://dx.doi.org/10.1016/j.ijhydene.2009.08.044.

Jianliang, Y., Wei, M., Yajie, W., 2005. Experiment to suppress Gas explosion in pipewith structure of multi-layer wire mesh. Nat. Gas. Ind. 25 (6), 116.

Jomaas, G., Zheng, X., Zhu, D., Law, C., 2005. Experimental determination of coun-terflow ignition temperatures and laminar flame speeds of C 2eC 3 hydrocar-bons at atmospheric and elevated pressures. Proc. Combust. Inst. 30 (1),

193e200.Kim, W.K., Mogi, T., Dobashi, R., 2014. Effect of propagation behaviour of expanding

spherical flames on the blast wave generated during unconfined gas explosions.Fuel 128, 396e403.

Kolbe, M., Baker, Q., 2005. Gaseous Explosions in Pipes. Paper presented at theASME 2005 Pressure Vessels and Piping Conference.

Kumar, K., Mittal, G., Sung, C., Law, C., 2007. An Experimental Investigation onEthylene/oxygen/diluent Mixtures: Laminar Flame Speeds with Preheat andIgnition Delays at High Pressures. Paper presented at the 5th US CombustionMeeting.

Liao, S., Jiang, D., Gao, J., Huang, Z., 2004. Measurements of Markstein numbers andlaminar burning velocities for natural gaseair mixtures. Energy & fuels 18 (2),316e326.

Liberman, M.A., Ivanov, M.F., Kiverin, A.D., Kuznetsov, M.S., Chukalovsky, A.A.,Rakhimova, T.V., 2010. Deflagration-to-detonation transition in highly reactivecombustible mixtures. Acta Astronaut. 67 (7e8), 688e701. http://dx.doi.org/10.1016/j.actaastro.2010.05.024.

Liu, Q., Zhang, Y., Niu, F., Li, L., 2015. Study on the flame propagation and gas ex-plosion in propane/air mixtures. Fuel 140, 677e684.

Ma, Q., Zhang, Q., Li, D., Chen, J., Ren, S., Shen, S., 2015. Effects of premixed methaneconcentration on distribution of flame region and hazard effects in a tube and atunnel gas explosion. J. loss Prev. process ind. 34, 30e38. http://dx.doi.org/10.1016/j.jlp.2015.01.018.

Merlin, C., Domingo, P., Vervisch, L., 2012. Large eddy Simulation of turbulent flamesin a Trapped Vortex Combustor (TVC)eA flamelet presumed-pdf closure pre-serving laminar flame speed. Comptes Rendus M�ecanique 340 (11), 917e932.

Movileanu, C., Razus, D., Oancea, D., 2011a. Additive effects on explosion pressureand flame temperature of stoichiometric ethyleneeair mixture in closed ves-sels. Rev. Roum. Chim. 56, 11e17.

Movileanu, C., Razus, D., Oancea, D., 2011b. Additive effects on the burning velocityof ethyleneeair mixtures. Energy & fuels 25 (6), 2444e2451.

Movileanu, C., Razus, D., Oancea, D., 2013. Additive effects on the rate of pressurerise for ethyleneeair deflagrations in closed vessels. Fuel 111, 194e200.

Oh, K.-H., Kim, H., Kim, J.-B., Lee, S.-E., 2001. A study on the obstacle-inducedvariation of the gas explosion characteristics. J. loss Prev. process ind. 14 (6),597e602.

Petchenko, A., Bychkov, V., Akkerman, V. y., Eriksson, L.-E., 2007. Flameesoundinteraction in tubes with nonslip walls. Combust. Flame 149 (4), 418e434.http://dx.doi.org/10.1016/j.combustflame.2007.02.003.

Petukhov, V., Naboko, I., Fortov, V., 2009. Explosion hazard of hydrogeneair mix-tures in the large volumes. Int. J. Hydrogen Energy 34 (14), 5924e5931.

Phylaktou, H., Andrews, G.E., 1991. The acceleration of flame propagation in a tubeby an obstacle. Combust. Flame 85 (3e4), 363e379. http://dx.doi.org/10.1016/0010-2180(91)90140-7.

Phylaktou, H., Foley, M., Andrews, G., 1993. Explosion enhancement through a 90�

curved bend. J. loss Prev. process ind. 6 (1), 21e29.Prince, J.C., Williams, F.A., 2012. Short chemical-kinetic mechanisms for low-

temperature ignition of propane and ethane. Combust. Flame 159 (7),2336e2344.

Qiao, L., Kim, C., Faeth, G., 2005. Suppression effects of diluents on laminar pre-mixed hydrogen/oxygen/nitrogen flames. Combust. Flame 143 (1), 79e96.

Razus, D., Movileanu, C., Brinzea, V., Oancea, D., 2006. Explosion pressures ofhydrocarboneair mixtures in closed vessels. J. Hazard. Mater. 135 (1), 58e65.

Salzano, E., Cammarota, F., Di Benedetto, A., Di Sarli, V., 2012. Explosion behavior ofhydrogenemethane/air mixtures. J. loss Prev. process ind. 25 (3), 443e447.http://dx.doi.org/10.1016/j.jlp.2011.11.010.

Soltaninejad, M., Dizaji, F., Dizaji, H., Bidabadi, M., 2015. Micro-organic dust com-bustion considering particles thermal resistance. J. Central South Univ. 22 (7),2833e2840. http://dx.doi.org/10.1007/s11771-015-2815-0.

Thomas, G., Oakley, G., Bambrey, R., 2010. An experimental study of flame accel-eration and deflagration to detonation transition in representative processpiping. Process Saf. Environ. Prot. 88 (2), 75e90.

Wang, W.-B., Li, J.-Y., Wu, Q.-J., 2007. The design of a chemical virtual instrumentbased on LabVIEW for determining temperatures and pressures. J. Anal.Methods Chem. 2007.

Xiao, H., Wang, Q., He, X., Sun, J., Shen, X., 2011. Experimental study on the be-haviors and shape changes of premixed hydrogeneair flames propagating inhorizontal duct. Int. J. Hydrogen Energy 36 (10), 6325e6336.

Zhao, Z., Zhenyuan, J., Haizhu, L., 2015. Characteristics of gas explosion flow fields incomplex pipelines. Int. J. Min. Sci. Technol. 25 (1), 157e164.

Zhou, B., Sobiesiak, A., Quan, P., 2006. Flame behavior and flame-induced flow in aclosed rectangular duct with a 90� bend. Int. J. Therm. Sci. 45 (5), 457e474.http://dx.doi.org/10.1016/j.ijthermalsci.2005.07.001.

Zhu, C., Lin, B., Jiang, B., 2012. Flame acceleration of premixed methane/air explo-sion in parallel pipes. J. loss Prev. process ind. 25 (2), 383e390.

Zhu, C., Lu, Z., Lin, B., Jiang, B., 2010. Effect of variation in gas distribution on ex-plosion propagation characteristics in coal mines. Min. Sci. Technol. (China) 20(4), 516e519.