A catastrophic aluminium-alloy dust explosion

A catastrophic aluminium-alloy dust explosion in China G.Li a,*,H.-X.Yang a,C.-M.Yuan a,R.K.Eckhoff a,ba Fire&Explosion Protection Laboratory,Northeastern University,Shenyang,Chinab Dept.Physics and Technology,University of Bergen,Bergen,Norwaya r t i c l e i n f oArticle history:Received1July2015Received in revised form11November2015Accepted15November2015 Available online2December2015Keywords:Accidental dust explosions Secondary dust explosions Aluminium product polishing Waste metal dustsDustfilters a b s t r a c tOn August2,2014a catastrophic dust explosion occurred in a large industrial plant for polishing various aluminium-alloy parts in Kunshan,China.The explosion occurred during manual polishing of the sur-faces of aluminium-alloy wheel hubs for the car industry.75people lost their lives immediately and another185were injured.Subsequently,71of the seriously injured also died,which increased the total loss of lives to146.The direct economic loss of was351million yuan.This is probably one of the most serious dust explosion catastrophes known apart from some very major coal dust explosion disasters in coal mines.Based on measurements of explosion parameters of dust samples collected on the explosion site and on-site investigations and interviews,it was concluded that a series of consecutive explosions was initiated in one of the external dustfilters.Then it propagated into the main building via the dust extraction ducting and further onto the secondfloor.At the same time the propagating in-house dust flame was sucked into the ducts leading to seven other external dustfilters,which also exploded.On the basis of investigations on site after the explosion and subsequent laboratory experiments and data analyses it was concluded that the explosion was most probably initiated by self-ignition of contaminated aluminium-alloy dust in the dust collecting barrel below the external bagfilter unit in which the initial primary explosion took place.General ignorance of the potential risk of dust explosions in industries producingfine metal dusts as a low-mass waste by-product is the most probable basic root cause of this catastrophic accident.Therefore, avoiding accumulation of deposits of such dusts indoors by good regular housekeeping and other means is regarded the most effective and practical way of loss prevention due to metal dust explosions in such plants.In addition,explosion isolation between dust collecting systems and workshops appears to be another important measure towards minimizing the consequences of such explosions.©2015Elsevier Ltd.All rights reserved.1.IntroductionAs pointed out by Eckhoff(2005)the threat of accidental dust explosions remains a persistent hazard in the process industries.Abbasi and Abbasi(2007)reviewed the main dust explosions that have occurred world-wide during the period1911e2004.It may appear that during the last decade only two other major catastrophic dust explosions have occurred worldwide,which are comparable to the one discussed in the present paper as regards losses of lives.Thefirst occurred in USA on February6,2008.A catastrophic dust explosion andfire devastated the Imperial Sugar refinery plant in USA,causing14deaths and38injuries(CSB,2009).The other occurred in China on Feb.24,2010in the maize starch production plant of Qianhuangdao Lihua Starch.According to State Administration of Work Safety(2014)21employees lost their lives and47others were injured in this disaster.The catastrophic dust explosion accident to be discussed in the present paper is probably one of the most serious industrial dust explosions in modern history apart from some very major coal dust explosion disasters in coal mines.The site of the explosion was a plant named Zhong Rong Metal Products Co(ZRMP),located in Kunshan,China.The plant was mainly polishing surfaces of various metal parts.In the morning of August2,2014a catastrophic aluminium-alloy dust explosion swept through a factory building where aluminium-alloy wheel-hubs for the car industry were polished manually.75people lost their lives more or less imme-diately and another185were injured.Subsequently,71of the seriously injured also lost their lives,which increased the total loss*Corresponding author.E-mail address:ligang@(G.Li).Contents lists available at ScienceDirectJournal of Loss Prevention in the Process Industries jo urnal homepag e:/locate/jlp/10.1016/j.jlp.2015.11.0130950-4230/©2015Elsevier Ltd.All rights reserved.Journal of Loss Prevention in the Process Industries39(2016)121e130of lives to 146.The direct economic loss of was 351million yuan.This accident once again has reminded the world of the importance of acknowledging the hazard of industrial dust explo-sions,and also to acknowledge the need for continually updating the knowledge and awareness required for preventing such catas-trophes in the future.2.The plant that was struck by the explosion 2.1.Sketch of the workshop buildingThe explosion occurred in a two-storey reinforced concrete-frame-structure work-shop building of length 44m (from north to south)and width 24m (from east to west).The two storeys comprised a basement with concrete floor and a first floor above it.The height between floor and ceiling was 4.5m in both storeys.The total floor area was 2112m 2.The roof of the building (and hence also of the first floor)was a light steel beam structure covered by coloured steel panels.The external and internal walls of the building were of brick.The two floors were connected by open stair cases at each end (north and south)of the building.On both sides of the eastern wall there was a 4m by 4m steel-panel sliding door leading to the outside.The floor of the first storey (i.e.the basement of the building)was of concrete.2.2.Process equipment layoutThe 32polishing production lines (16lines on the basement and 16lines on the first floor)were arranged in parallel in the south-north direction.Along each line there were 12working stations,as indicated in Fig.1.On the day before the accident,a total of 29production lines were in operation,13on the basement and 16in the first floor.348workers were on duty.Polishing operations were conducted manually as shown in Fig.2.Electric grinding guns were the main tools.According to the finish smoothness required different grades of grinding heads and/or emery papers were used.2.3.Dust collection system and processing equipmentAccording to the documents investigated,8sets of dust collec-tion systems were installed in March 2006outside the main factory building to serve the polishing processes on the two floors.Therefore,every two production lines shared a set of de-dusting system.Every two pieces of dust extraction ducts were merged by a T-joint into one main 450mm F main duct leading to the bagfilter.At each work location along a production line there was a 500mm Â200mm,which meant that each dust extraction system collected the dust from the 48work locations.According to the design of the bag filters in the dust collectors the bags were supposed to be cleaned by mechanical shaking at intervals.However,after the explosion accident survivors told that,due to breakdown of the electric motor,the automatic shaking systems had been out of operation for a long time.Instead workers had to shake the bags manually every morning before commencing work.This process was called “shaking ash ”.The air flow for each of the 8main dust extraction lines was produced by a suction fan mounted on the clean side of the bag filter unit,and all the 8air flows were joined in a main discharge duct to the atmosphere.No special requirements addressing a possible dust explosion hazard had been enforced on selection and installation of all the electrical equipment used in the plant.Neither the dust collectors or the dust extraction ducting,nor all electrical sockets and power distribution cabinets had been adequately earthed.3.The explosion catastrophe 3.1.ExtentIn the morning of August 2,2014,at about 7:34a.m.,when normal polishing activity has been going on for half an hour,a strong explosion occurred.A survivor stated that he was polishing his second hub at the moment of the first explosion.A video camera located outside another factory building about 500m away from the building that was stuck by the explosion revealed a sequence of several explosions lasting for about 5e 7s,including a distinct series of 8successive explosions.These 8explosions were identi fied on the video recording as violent “mushroom-shaped ”dust/smoke clouds being blown up from each of the 8dust collectors.All the windows in the first floor of the building that exploded were shattered and blown to the outside,and the window frames were completely deformed.Two thirds of the southern wall of the building collapsed,as illustrated in Fig.3.Window frames were blown out of the eastern wall,and two air conditioners located at this wall got partly detached from the wall,as shown in Fig.4.The blasts also lifted some of the light steel-panel roof of the second floor,and broke all the windows.Surprisingly,the glass was found scattered on the ground floor inside the building.Almost all process equipment in the workshop was destroyed as shown in Fig.5.Fig.1.Production lines with workstations.Fig.2.Workers on duty at the work stations.G.Li et al./Journal of Loss Prevention in the Process Industries 39(2016)121e 130122The dust extraction ducts inside the building were basically intact,but the duct connected to the 1#filter suffered serious damage and got detached from the T-joint.As Fig.6shows,a duct branch within the workshop was broke into three parts with the middle part of it having been torn apart.All the dust filters (1#,3#,5#and 7#)collecting dust from the first floor were more severely damaged than those (2#,4#,6#and 8#)collecting dust from the second floor.In particular dust filter #1was blown entirely apart.A signi ficant detail is that the dust col-lecting barrel of this filter had its bottom completely blown out,as shown in Fig.7.The blue arrow shows the corroded hole in the barrel to be discussed later in this paper.3.2.Loss of life,injuries,and financial lossesOn the first day of the catastrophic explosion 75workers lost their lives,whereas 185were injured.Most of the dead corpses carbonized,and all the injured persons had symptoms of respira-tory tract burns.During the days and weeks after the catastrophe another 71workers died,which means that finally the total loss of lives amounted to 146,whereas the total number of injured workers was 114.The direct financial losses caused by this explo-sion disaster amounted to 351million yuan.4.Analysis of the explosion development 4.1.Identi fication of the exploding materialThe investigation revealed that the aluminium dust generated during polishing of aluminium alloy hubs was the onlypossibleFig.3.Collapsed and destroyed part of southernwall.Fig.4.Destroyed eastern wall and airconditioners.Fig.5.Destroyed process lines on the second floor.Fig.6.Detailed view of damagedducts.Fig.7.Dust collecting barrel of 1#filter with its bottom blown off.G.Li et al./Journal of Loss Prevention in the Process Industries 39(2016)121e 130123material that could have caused the explosion accident.The explosive properties of clouds in air of aluminium dust have been investigated by many authors,e.g.Nifuku and Katoh(2001) and Marmo et al.,(2004).In NFPA(2008)aluminium dust is ranked among the most explosible metal dusts.Some data collected by CSB(2003)indicates that nearly25%of all dust explosions in the United States between1980and2005involved metal dusts,and that aluminium dusts accounted for the majority of these events. Metal dusts account for about19%of all dust explosions worldwide.4.2.Identification of the site of explosion initiationExamination of the explosion site revealed that the wall next to the southern-side stairwell leading to the secondfloor(close to the #1dust collector)showed clear traces of a large number of particles having been ejected upwards from the bottom,as shown in Fig.8.This would indicate that the blast wave swept from thefirst floor to the secondfloor along this wall.The fact that the northern side stair railing collapsed from south to north proved that the blast wavefirst propagated from the south to north along thefirstfloor. This agrees with the assumption that the starting point of the in-door propagation from south to north was the production line corresponding to the#1filter,because the dust extraction duct of filter#3close to the#1filter duct suffered a shift from south to north,as shown in Fig.9.Also,several of the“spiro”ducts in the dust extraction and ventilation systems on thefirstfloor showed signs of having been moved from south to north.The closer an item was to the#1filter ducts,the more obvious was this northwards displacement.As already said above,the indoor dust extraction duct connected to the #1dust collector was broken into3separate sections at theflanges, as indicated in Fig.6.The end of the pipeline suffered the most serious damage away from the dust collector to east,as shown in Fig.10.This kind of displacement was not found for the other dust extraction ducts.It was concluded,therefore that the explosion originated in the#1 dust collector outside the western wall of the building,and spread from there to the interior of the1stfloor of the workshop,causing eastward displacement of dust extraction ducts.The1#dustfilter suffered the most serious damage among the8 dust collectors.The cubicalfilter enclosure was completely torn open,and the conical hopper was partly damaged.Theflange connecting the hopper to the dust collecting barrel below was torn apart,and the barrel was separated from the hopper,as shown in Figs.11and12.4.3.Analysis of path of blast wave propagationAs already said above,an external video monitor revealedaFig.8.Vertical upwards traces of aluminium dust particles ejected along the southernwall of thebuilding.Fig.9.#3duct moved to thenorth.Fig.10.End of#1duct moved to theeast.Fig.11.Totally damaged#1dustfilter.G.Li et al./Journal of Loss Prevention in the Process Industries39(2016)121e130124series of successive explosions extending over about 6s,from the appearance of the first “mushroom ”-shaped cloud at dust filter #1till the appearance of the last cloud at one of the other filters.According to the evidence found on site the blast wave that was initiated in filter #1propagated,followed by the propagating dust flame,upstream inside the dust extraction duct and entered the first floor working lines,from where it exited into the workrooms via the dust extraction hoods above the working tables,and eventually ignited the dispersed deposited dust on the floor near both the #1and #2production lines,causing major escalating secondary explosions.Then these explosions and associated dust flames propagated the entire length from north to south along the first floor and got sucked into the dust extraction ducts leading to the 3#,5#,and 7#dust filters,triggering 3further major filter explosions.The most probable blast wave propagation path on the first floor is indicated in Fig.13.While sweeping through the first floor,the blast wave,driven by the secondary explosions had also propagated through the south-ern and northern stairs upwards to the second floor.There it destroyed part of the second floor roof and dispersed the dust deposited on the production lines,the floor etc.causing another extensive secondary dust explosion sweeping through the entire secondary floor.Finally the flame was sucked into 2#,4#,6#and 8#dust filters via the second floor production lines,which caused the four further major filter explosions with the “mushroom ”shaped dust clouds being blown to the outside from the filter enclosures.It should be noted that the dust extraction ducts connected to the 1#dust collector were torn into three parts at the connecting flanges.Each of the three ducts has 8sucking holes located sym-metrically along both sides of the duct.The layout of the ducting system connected to the 1#bag filter unit is shown in Fig.14.The thickness of all duct walls was about 1mm.The venting was provided by two opposite 100mm F openings for every 1m duct length.This would give quite generous venting,although the vent area was somewhat less than that required by NFPA (2008a).In the actual explosion only the middle section of the duct was torn apart by the explosion overpressure in the duct,as shown in Fig.6.A likely reason for the maximum pressure occurring in the middle section could be acceleration of the de flagrating flame in the duct due to flow-induced turbulence,causing pressure piling and enhanced explosion pressure.Under given circumstanced such de flagrations can transit to detonation-like explosions,and this possibly occurred midway along the duct.5.Further analysis5.1.Conditions for dust explosion occurrenceAs pointed out by Amyotte (2013)it is well known that 5basic conditions have to be ful filled simultaneously for a dust explosion to occur:presence of a combustible dust in a finely divided form as a cloudpresence of an oxidant atmosphere presence of an ignition sourceconcentration of dust in the cloud within the explosive range (between the minimumexplosive and maximum explosive concentration)some degree of con finement of the dustcloudFig.12.Torn receiving hopper of #1dust filter.Fig.13.Most probable course of blast wave propagation on the first floor.yout with dimensions of the dust extraction ducting connected to the 1#bag filter unit.G.Li et al./Journal of Loss Prevention in the Process Industries 39(2016)121e 1301255.2.Characteristics of aluminium-alloy dust from hub finishing processThe particle diameter distribution of a sample of aluminium-alloy dust collected from the explosion site was measured using a laser-based particle size analyser.The results are given in Table 1.According to the documents from the Zhong Rong Metal Prod-ucts Co.the chemical composition of the wheel hubs that were finished in the plant on the day of the explosion was as shown in Table 2.This table also shows the chemical composition of the metals in the collected dust sample (Table 1).The collected dust also contained some other components (e.g.polishing paste and sandpaper fragments)but these are not included in Table 2.As can be seen,the percentages by mass of Si and Fe in the collected dust were signi ficantly higher than in the hubs.Explosion parameters for the collected dust samples were determined in the Fire &explosion protection laboratory at Northeastern University,Shenyang,China.Before determining the explosion parameters of the dust the impurities (traces of polishing paste and sandpaper fragments)were removed.The results ob-tained are shown in Table 3.The three first parameters were determined in a 20L closed standard explosion bomb,the other two in other standardized tests.5.3.Formation of the initial explosive dust cloudThe total amount of dust generated per day during the hub polishing was rather small.According to survivors the dust barrels at the bottom of the dust collectors were normally emptied twice a week,and the mass of dust collected in one barrel would then not be more than about 30kg.The volume of the dust fan that operated 12h a day for one filter was about 22,300standard m 3per hour.Therefore the average dust concentration in the clouds entering the filters was about 0.1g/m 3.As shown in Table 3this is less than one percent of the minimum explosive concentration (MEC)of the sample of the collected dust.Consequently,even if there has been an ignition source in the dust cloud flowing through the dust extraction system,a dust explosion could not have been initiated inside the flowing cloud.The dust concentration in the bag filter would have a pro-nounced maximum during the regular bag cleaning sequences of the mechanical vibrating system.But as mentioned before,the vibrating system of the filter close to which the explosion most probably started was out of operation at the time of the explosion because of breakdown of an electric motor.Therefore the workers had to clean the bags by shaking the bag holder every morning before commencing work.The explosion occurred about 30min after the production had started.Fig.15indicates the zone in the “dirty ”side of the filter enclosure,where excessive,explosive dust concentrations would occur during the falling down of the sticked dusts from the filter-bag.5.4.Identi fication of a likely ignition sourceAt the outset one has to consider several possible sources for ignition of the dust cloud on the “dirty ”side in the dust filter enclosure,including:Friction sparkElectrostatic discharge Electrical fault Thermite reactionSelf-heating in deposit of aluminium-alloy dust from hub polishing5.4.1.Friction sparkWhen using an emery wheel rotating at high speed to polish hard metals of metal alloys a beam of bright visible friction sparks is normally generated.Such sparks are burning particles of highTable 1Particle diameter distribution of sample of aluminium-alloy dust.Cumulative volume %of particles smaller than D D3a D10D50D90D97Dav Particle diameter (m m)5.27.619455923aD3¼5.2means that the volume of particles in the dust having particle diameters smaller than 5.2m m was 3%of the total volume.Table 2Chemical composition (%by mass)of metals in a raw hub and in dust sample.ElementAl Si Mg Fe Ti Raw hub 91.2e 92.3 6.5e 7.50.20e 0.40&0.20&0.20Dust sample88.410.20.110.550.0Table 3Explosion parameters for the dust sample.Parameters MEC (g/m 3)P max (bar,g)K St(bar $m $s À1)MIE (mJ)MITC ( C)Results40112507e 11540MEC is the minimum explosive dust concentration.P max is the maximum explosion pressure (closed explosion bomb of 20L).K St is the maximum rate of rise of explosion pressure.MIE is the minimum ignition energy of the dust cloud (electric spark).MITC is the minimum ignition temperature of clouds of thedust.Fig.15.Possible zone for formation of explosive dust clouds in the filter enclosure during manual bag cleaning.G.Li et al./Journal of Loss Prevention in the Process Industries 39(2016)121e 130126temperatures.Li et al.,(2013)found that such sparks from friction between Ti and an emery wheel can easily ignite a methane/air mixture.However,when using an aluminium alloy instead of ti-tanium no ignition was observed.Eckhoff(2003)describes dust cloud ignition experiments by metal particle sparks from short single impacts and pointed out that ignition by such impacts may seem less likely than believed by many in the past,unless the sparking metal is titanium or zirconium.When an aluminium-alloy wheel hub from the explosion site was polished in the laboratory by using a hand held electric grinder, no sparks could be seen by the naked eye against a black back-ground.The temperatures of both the friction spot and hot particles were recorded by a VH-420Thermal Imager(low band LW, 7.5e14m m)produced by Infratec in Germany.The friction spot was found to be the hottest area where the highest temperature value was133.5 C,as indicated in Fig.16.Even if the temperature of the hot particles produced by this polishing process had been consid-erably higher than maximum temperature found in this test,it is difficult to envisage that the particle would be an ignition source when they eventually,after a long period of travel and cooling, reached the explosive dust cloud in the dustfilter.5.4.2.Electrostatic dischargeThe phenomenon of electrostatic charge accumulation on pro-cess plant,powders and dust particles is well documented in the literature.The amount of charge generated depends on many things,including the relative humidity of the air within the process equipment,the particle size,chemical composition of the particles, and the nature of contacting surfaces.There are several types of electrostatic discharges,viz.corona discharges,brush discharges, propagating brush discharges,bulk surface discharges and spark discharges.Apart from corona and brush discharges,all the elec-trostatic discharges mentioned can in principle give rise to dust explosions in air.The minimum ignition energy of the dust sample collected on site,as determined in a standard laboratory test,was7e11mJ as shown in Table3.This is a comparatively low value,and at the outset one cannot neglect the possibility of an electrostatic spark discharge being the ignition source in the present case.However, due to the high electric conductivity of the wheel hub dust it would probably be difficult to accumulate enough electrostatic charge by tribo electric processes to produce even a spark discharge of only about10mJ.Moreover,it had been raining continuously for several days before the explosion accident,and light rain persisted even right up to the moment of the explosion.The relative humidity of the air was97%at the average ambient temperature of29 C.Considering these factors,an electrostatic discharge is unlikely to have been the ignition source of the present explosion disaster.5.4.3.Electrical faultA faulty electrical circuit,a damaged electrical wire or another electrical component can cause sparking or arcing that can defi-nitely ignite explosive dust clouds.However,in the dustfilter in which the explosion most probably started there were no other electrical components than the defect and out-of-operation electric motor intended for automatic periodic shaking of thefilter bags. Furthermore,this motor was on the“clean”side of thefilter.Therefore,an electrical fault is not considered a likely ignition source in the present case.5.4.4.Thermite reactionWhen aluminium or aluminium alloys gets in contact with iron oxide at a high temperature,a strong exothermal reaction can result,releasing sufficient energy to ignite mostflammable mix-tures.This exothermal reaction is called a“thermite reaction”and the reaction equation is:Fe2O3þ2Al/Al2O3þ2Feþheat(1) Survivors of the explosion catastrophe discussed in the present paper said that they usually polished their blunt iron scissors at the same work table as they normally polished the wheel hubs.As seen in Table2,the element analysis of a sample of the dust from the actual hub polishing operation showed an iron content of0.55%by weight.However,nearly2/3of this was from other sources than the aluminium-alloy hubs.In the case of the temperature rise caused by self-heating re-action,small amount of ferric oxide and aluminium in the polished aluminium dust could in principle give rise to a thermite reaction that would liberate some heat and increase the likelihood of self-ignition in the dust deposit.5.4.5.Self-heating/self-ignition in the deposit of aluminium-alloy dust in the dust-collecting barrel below the dustfilterAs shown in Table1the particles produced by polishing the aluminium-alloy hubs were relatively small,with the median (volume based)diameter being only19m m.Due to some organic material added in the polishing process the deposited dust became quitefluffy,as indicated by Fig.17.Tests showed that the poured bulk density of thisfluffy dust was only about520kg/m3.This would indicate that the heat conductivity of deposits of this dust would also be considerably lower than that of deposits of pure aluminium-alloy dust.Many metals can react exothermally with water.Inside a deposit of metal dust/powder the initial reaction rate may be very slow,but if the heat dissipation into the surrounding parts of the dust deposit is sufficiently low,the self-heating reaction can accelerate right up to self-sustained exothermal combustion.Birks and Alabi(1986) investigated the special problem of self-ignition in piles of direct-reduced iron when the piles were exposed to water.They found that auto-ignition starts with aqueous“corrosion”reactions and becomes completely developed when stable high temperature oxidation isestablished.Fig.16.Thermal image of hand hold grinder polishing an aluminium wheel hub.G.Li et al./Journal of Loss Prevention in the Process Industries39(2016)121e130127。