Document NEVzVJo04zbE9dKNngLGrQyDQ

The Dow Chemical Company 2301 N. Brazosport Blvd. Freeport, lexas 77541-3257 Hydrocarbons Tech Center Reactive Chemicals Handbook INTRODUCTION Following is the 1995 edition of the Hydrocarbons Tech Center Reactive Chemicals Handbook which has been compiled over recent years utilizing know-how based upon many years of Dow manufacturing experience. Every effort has been made to provide useful and accurate information so as to enhance training for new and experienced Hydrocarbons personnel. Both basic chemistry and incident summaries are included. Contributions of additional information to existing or new sections of the handbook are welcome and encouraged. Reactive chemicals incidents occur many times due to operating outside of design conditions, other times due to inadequate design stemming from an inadequate knowledge of the process. This handbook is designed to eliminate reactive chemicals incidents through education so read on and remember............. * what you dont know CAN hurt you! Brian Allison Hydrocarbons Tech Center 5/26/95 ^ A 0306 CONFIDENT I 1. Acetylene/Acetylides 2. Aluminum and Halogenated Hydrocarbons 3. Ammonia 4. Caustic Soda 5. Chlorine 6. Ethylene Decomposition 7. Ethylene-Hydrogen Reaction - Rust Catalyzed 8. Fire and Explosion 9. Insulation and Spontaneous Fires 10. Molecular Sieve 11. Peroxide Forming Chemicals 12. Pypophoric Polymers 13. MAPD in C3 Splitters 14. NOx Containing Process Streams 15. Loss Case Histories in Pressurized Ethylene Systems 16. Chemical Compatability Charts 00 A 030e>4 5 CONFI^NTTAl 00 A 030646 C0NFIDFNTIA1 Hydrocarbons Reactive chemicals Hazards ACETYLENE/ACETYLIDES Acetylene and numerous acetylenic compounds have a history of being able to decompose with explosive violence. As a group, these materials are known to be sensitive to heat, impact and electrical or frictional sparks. Pure Acetylene will decompose into its elements at a pressure on the order of one atmosphere. Compressing acetylene becomes a dangerous operations unless special techniques are used. When liquified or solidified, acetylene becomes an explosive of some commercial importance. They are one of the few commercial explosives which contain no oxygen or nitrogen. The vapors of these acetylenic compounds are very flammable. Their vapor-air explosions can and do propogate from deflagrations to detonations over a wide range of compositions. Acetylene reacts with copper to form the explosive copper acetylide. Please see the attached CRI report on this subject. Incidents Union Carbide - Texas City: A butadiene finishing column on total reflux exploded. It was concluded that vinylacetylene reached a concentration of approximately 60% between trays 10 and 15 because light fractions were leaking overhead. Temperature and pressure was sufficiently high enough to cause decomposition. Precautions Distillation columns containing acetylenic compounds should not be placed on total reflux. A bottoms draw-off must be maintained. These systems must be carefully studied to avoid unstable concentrations. DO A 030647 CONFIDENTIAL HOW CONFIDENTIAL INFORMATION R & D REPORT DOW CHEMICAL U.S.A. SAD REPORTS SHOULD REMAIN ON THE PREMISES OF THE DOW CHEMICAL COMPANY . i * >f p* RESEARCH 1. DEVELOPMENT "potential acetylide formation in copper s*ample LINES *44 l* AIIONaIQiM h I.I'EINI CUOl LAD 587 U* T fc i/kIN'.W December 31, 1980 k_ * N V 1. E "* 1l 1 1 1 -L l 1 4 paces IN FULL REPORT Owe LUSIONS: IMARY AND CONCLUSIONS -..plosive copper acetvlides nay form on copper and on brasses which contain 50% copper or more, -hen surfaces are exposed to acetylene under certain conditions, "lean metals require moisture for acetylide formation. None vs formed in dry acetylene. Oxidizing agents, acidic conditions and nitro gen inhibit acetylide formation. Studies have shown that up to 300 ppm concentration of acetylene m contact with pure copper tubing over a period of 12 years does not produce a hazardous condition. Mso, a copper tubing sample line analysis indicated a low or safe amount of acetylides in a line containing on the average of 1000 ppm acetylene and n use for several years. These studies indicate that copper may be safely used in light hydrocarbon systems containing up to fractional percent con centrations of acetylenes. department files 'STRrBUTION: RAD ADMINISTRATION i-ruTO*i #fpntT INDEX - b COPIES DO A CT-suo COWFT Diifribution lift is ctminiitd an -2- IN'TKODUCTION It is known that acetylene forms explosive compounds with copper. These compounds are called copper acetylides, and are liable to decomposition under thermal or mechanical shock. Because of the interest in safety questions related to the use of copper and.'its alloys in sampling systems handling acetylene mixtures, a literature 'study has been done. The infor mation needed is concerned with conditions under which acetylides may be formed on metallic copper and copper alloy surfaces. It is necessary to consider the possibility of such formation not only on the clean metal surfaces, but also when these surfaces have become corroded or contaminated by exposure to acid, alkali, moisture and atmospheric attack. The Chemistry of Acetylides - Acetylene (HCSCH), is unique among hydro carbons in that the lone hydrogen attached to triply bound carbon is acidic and replaceable by metals. When acetylene gas is passed into a solution of cuprous ammonium hydroxide, a reddish brown precipitate of cuprous acetylice is produced. Both acetylenic hydrogens are replaceable by univalent copper atoms; an alkyne of the type RCICT1 gives a monocopper derivative. Silver acetylides are similarly precipitated from an ammonical solution of ver nitrate.2* ylides are also prepared by the action of metals or metal compounds on o jon at high temperatures. Alkali and alkaline-earth acetylides are relatively stable, but most heavy metal derivatives are thermodynamically unstable and may explode when dry. They are readily decomposed by dilute acid and release acetylene under such conditions.^ Cuprous acetylide is an amorphous red powder which is an explosion hazard when shocked or exposed to heat. It reacts vigorously with Br2 C12' .\gNO3. Acetylides explode readily and are one of the few commercial explosives which contain no oxygen or nitrogen. Their explosion produces no gas but simply is an effect of the large amount of heat, instantaneously produced. They are in a class with the fulminates and the azides as primary detonants. Because these materials are so sensitive to shock and temperature, they must be handled with extreme care. They must be kept cool and if they are to be stored, should be kept wet.4 Review of the Literature - Scheiber and Reckleben investigated the action of crude and purified acetylene on copper and a number of its alloys. They found that pure dry acetylene gave no action within a period of 20 months, r test specimens being unaltered in weight or appearance. Pure wet acet ne caused 1.6% increase in weight in the case of a copper specimen Crude wet gas caused the copper to blacken and to increase in weight within 6 months. No explosive material was detected, however, nor was it possible to detect -* 2 than small amounts of acetylide.5 DOW CONFIDENTIAL INFORMA^ljC^Ur\R\ZC.0 030649 dfnttal -3- occasional srxjntn neons decomposition of in cylinders is initiated by Cu->C2 formed in the valve springs tinder the influence of soldering flux residue. Pressure-gage springs from acetylene cylinders (one from an exploded one) were investigated. By x-ray^analysis or by electron diffrac tion the following corrosion products could be found in the bourdon springs: CuCl, 3 Cu(OH)2'CuCij, Cu20, 4Zn(OH)2 'ZnCl, and SiC>2* Elementary "arbon and Fe2C>3 could also be identified through color and chemical Pena vior. In a seven month study, the formation of Cu2C2 upon the surface of pure copper with both pure dry C2H2, and with moist acetylene was found to be extremely slow. The thickness of the acetylide layer reached 1.6u with moist acetylene and 0.16 to 0-4u with dry acetylene.*^ The basic problem -it- handling acetylene in copper sample tubing is the formation of copper icetyiides on brass fittings, valves, lines, etc. The presence of NH2, K-iO, and CO-) further acetylide formation. It is probable that copper car bonate contributes to the formation also. Basic copper compounds are parr uiarly reactive. Oxidizing, strong acid conditions, and nitrogen : bit formation. er acetylides are only explosive when dry. The amount and explosive ness of the acetylides decrease with decreasing copper content of the alloys. If work has to be done on equipment known to have these deposits, the equipment should be thoroughly wetted down if it cannot be cleaned up immediately. TTie best material for wetting down equipment and keeping it wet is a calcium chloride brine. The best and most economical method for removal of copper acetylides from equipment is by cleaning with dilute hydrochloric acid. In-place acid cleaning will eliminate the hazard of topper acetylides explosions upon equipment dissassembly or removal. Experimental It has been shown at the Texas Division Light Hydrocarbons I Plant that a V4" copper sample line over 200* long and containing an average of 1000 ppm acetylene held a low or safe amount of acetylides after several years in service. In addition, laboratory data has been generated from a section of 1/j" copper sample line from the Louisiana Division's Light Hydrocarbons I Plant which had been in service approximately 12 years. Highest concentrations of ace tylene in the system ranged from 200 to 300 ppm. The concentrations of acetylides detected were extremely low and ranged from 0.06 ppm to 0.09 -4- Rcferences 1) Brameld, V. E., Clark, M. T., Seyfang, A. P., pp- 316-53 (1947). Soc. Chem. Ind. 66, 2) Morrison, R. T., and Boyd, R. N., "Organic Chemistry,11 3rd Ed., Allyn and BAcon, Inc., Boston, Mass., 1977, pp. 259-261. 3) Hadley, E- H., "McGraw-Hill Encyclopedia of Science and Technology", Volume i, McGraw-Hill, Inc., New York, 1977, p. 46. 4) sax, N.I., "Dangerous Properties of Industrial Materials, "Van Nostrand Rhinhold Company, Atlanta, Ga., 1979, p. 339. 5) Brameld, V. E. , Clark, M- T., Seyfang, A. P., pp. 346-53 (1947). Soc. Chem Ind 66, 5) "eitknecht, w. , and Hugi Carmes, L., Schweiz. Arch. Anoew. v?iss. 0. Tech. 23, pp. 32Q-338 (1957). 7) Polyakor, N. N., Khim Prom., 1954, pp. 457-462. 9) Rutledge, T. F., "Acetylenic Compounds; Preparation and Substitution Reactions," 1st ed. Re inhold Book Corp., New York, 1968, p. 85. 9) Berqstrom. D. R.. Dow Chemical Company Report, File 311, Midland, Michigan, April 1969. 10) Hegyesi, J. H., Dow Chemical Company Report, 3-4001, Freeport, Texas, Auoust 1972. AuinCivuxu' COFf DOW CONFIDENTIAL INFORMATION DO A 030651 CONFIDENTIAL distribution list LOUISIANA E. L. Daigle D. L. Engibous W. J. Neely CRI (5 ) DOWELL DIVISION C- M. Maddin - 4103 - 2509 - 2306 - 25 0 7 WESTERN - PITTSBURG L. M. Kroposki M'. D. Yeaman SARNIA A. R. McDowell H. W* Quinn INDIANAPOLIS - DOW PHARMACEUTICALS r. R. Fike MIDLAND W. B. Crummett H . J* Dishburger M . V. Koch J. M. Leathers J. w. Linowski E. V. Luoma M* E. Pruitt G- L. Rock w. A. Rogers D* P. Sheetz R- H- Yocum CRI (4) - 574 - 9008 - 574 - 2030 - 1776 - 2020 - 2020 - 2020 - 2020 - 2020 - 2040 - 566 FREEPORT J. B. Buettner E. H. Holt F. D. Martin J. V. Goode *L. Wright V. A. Zeitler GRANVILLE H. Tung OYSTER CREEK L> Hollis - 3-1225 - B- 1225 - A-1210 - B- 1219 - B-1210 - B-1219 SUMMARY" ONLY DO A 030o5.<- donftdfnttal f 'X ? v? i. 1 1. i . .. . i .i a PLANT SAFETY & LOSS PREVENTION 3 C *f* .WX S it CA: 't*4 J ! *=>!.. i' -lS -a r S' -* a; B *. it /ts^ fm p <04, {-*j a '**' *,V K '**' t tt stVIt ai*i o^.ssclc/ rn/- A company team representing ha Operating, Engineering, and R&D Departments was assigned ie job of finding cut what ha; pened, and what to do to prevent a recurrence. R. H. Freeman, and P. fv'cCready, Union Carbide Corp., South Charleston. W. Va. THE FOr.CE or THE EXPLOSION AT Till" TKX VS CITY COM- plex fragmented the lnwi*>- port ion of the column shell; that is, from the skirt-tu-sitell weld to the elevation of the 33rd tray, an overall height c>f -1:1 ft. Approximately 40 trays were torn apart an:! blown free of the column shell. The tray sections and pieces of the column shell were strewn over a wide area surrounding the center .jM; the explosion. Most cl the debris was located within >00 ft. of the refining column. The remaining portion of the ro-ining column, approximately 100 lineal ft. of 00 in. ID shell and the associated trays, fell to the east. Hydrocarbons from the refining column, and from other operating c<iuipmeut which had been rup tured by shrapnel, contributed to the ensuing fires. Fortunately, there were no fatalities or serious in juries as a result of this explosion. However, the con-, trol building, the nearby residential area, the olefins unit itself and, of course, the butadiene refining facili ties, sustained damage of varying severity. A joint-elVort investigation of the explosion has been conducted bv Union Carbide's Operating, Engi neering, and Research and Development Departments. The information gathered in the various phases of the Invest igai ion has beer, successfully correlated, and a plausible explanation of the cause and mcrhniiism of he explosion has been fornuilaleil. In addition, this oup worked with others to define a safe mode of Iteration for all butadiene refining facilities in the corporation. Tin- overall conclusions and recommenda tions are contained in the third article in this series. based on information presented in all three articles.* The Engineering Department has b^en involved both in the reconstruction elfort at Texas C'ity. and in the investigation to determine the cause of the ex plosion. This article describes the Engineering De partment's contributions to the investigation, which include examination of the physical evidence, and cal culation of the process conditions which existed at the time of the explosion. The conclusions derived from this phase of the overall investigation are stated, and they arc reinforced by the results of the other phases of the investigation as reported in the other articles in this series. Investigation of physical evidence Refining column shell-After the explosion, some time was spent in collecting, marking, and identifying the debris. The column shell pieces were separated in a special area for further study. To aid in the identi fication of pieces, and to serve as a framework for the paper reconstruction of the column, a drawing, a. pro jected view of the inside surface, was made of the lower GO ft. of the refining column. The pieces of column shell were physically examined to identity their location on the drawing. Character istics such as shell thickness, clip type ami location, shell weld seams, tray ring welds, downcomer welds, and insulation rings, allowed positive indvntificntioii of all large fragments ami many of the smaller pieces. Once identified, the shell pieces were measured, and *AVr vm f*t* Aw/-'*' >/ /Htf/c ;i. DO A 030653 CONFTDFNTIAL 61 Figure 1. Fracture arid tear diagram. their shapes reproduced u> scale on liic drawing. The result is the fracture and tear diagram, Figure 1. Approximately 00'} of the separated shell pieces were found and located on the drawing. The low recovery of pieces in the in. thick shell ring No. 3, and the relatively small size of the pieces found -sug gest that the most intense force prevailed in this area. The crnsshstchcd areas at the bottom and top of Fig ure 1 represent .-hell plate not fragmented. The bottom section consisted of 8- to 10 ft. of the skirt that had remained attached to the column foundation. The up per crosshatched area represents a portion of the up per 100 ft. of column shell which remained intact. Further examination of the shell fragments con firmed the location of the explosion. The torn edges ''""Njf many, of the fragments exhibited chevron marks, while other edges were smooth and in a few places showed evidence of a slight thinning by drawing. The appearance of the edges of the fragments is an indi cation of the speed and direction of fracture. A rapid crack propagation in the range of 2.000 ft. sec. and above is termed cleavage. Cleavage fractures leave chevron markings or. the edge surface which point back to the origin of the crack. Conversely, .a low rale of crack propagation is termed shear. Shear failures occur at rates less than 500 ft., sec., and may show thin ning of the material adjacent to the edge. Fractures that propagate at rates between 100*; shear and 100f.t> cleavage may exhibit characteristics of both (7). The results of the examination of the fragment edges arc shown in Figure 1. The arrows indicate the direction of shell fracture as evidenced by chevron markings. The letter "S'" refers to fractures that apnoared to have undergone shear failure. The letter "C" idicates chevrons that did not positively show the di rection of tear. The few edges that showed no indi cation of cither the type nr direction of failure had i/COn severely damaged b\ hitting other objects. The fact that the shell lore both up and down front shell ring Xo. 3 centers the maximum intensity of the blast between the 10th and the 18th trays. The high incidence of shear between the 25t.'i and the 33rd trays is explained by the fact that the fracture stopped in this region. Trays and tray rings--During the initial cleanup of the area, many tray sections from the refining column were collected and removed to the salvage area for further examination. The examination -comprised a count by section type, and an evaluation of the direc tion of the explosive force on each section (2). Severe distortion of many of the sections made it impractical to distinguish among nil section types, but many pieces could be identilied with either a side downcorner deck (odd numbered trays) or a center downcorner deck (even numbered trays). In all, 39 trays were accounted for. out of the -13 blown free of the column, with S3 percent recovery of the sections. Tray sections which had constituted the bubbling area were examined in detail to determine whether they ha<l been deformed by an upward or downward force. Of the sections examined. 95 (c appeared to have been subjected to an upward force. This points to an origin of the blast below the bottom tray. Tray support rings had been attached to the inside of the column shell at 12 in. spacing throughout most of the fractured area. Each shell piece was examined to determine 1) if tray rings remained or 2) if miss ing, whether the weld metal remaining showed which way the tray ring broke o(T. The results of the tray ring examination are shown in Figure 2. The wide bands superimposed on the lighter tray ring outlines represent ring segments that remaim-d attached to the shell. All other tray rings were missing, and in many cases the direction of tear could l>e determined. The arrows on Figure 2 show the direction of tear as evidenced by a shear lip produced by drawing of the material. In the lower part of the remaining coin...n section '<>>. " i t' the lr,V.\ > SjJ-an t>rft t'rojll trio ti am; il*`i ;;:.w:.ni. Ynv shaurte;; or siv. -.'n tray . . i J ;,,{ I pr'-bably was by i\ pressure. c that f-uusc-.; t,r:-ya to move up ' hi. column !:Ui ; A *V,n- " vf Wi'.f ffi-i i'ii'.f r> -is muoiniiiy cidum:' :s sm-nvr. r lyure A. 1`. is sijf- Ticant that She- i*ileuiiiis::-; aTteehud to the cb.tll pieces il:0 la:--:: it ay. Th.sm-.l other uviuur.ci* is evaluated ).:li*: ' sirt/a'c. IMmjjUt and manway cmc-The butadiene reii::ir:y plvvia.i*:! with i.\ O, vertical, MlUs*::* ei reillation reltoileiT. h-uied with ip.. sq. in. gauge St.-piir; Vi! !:'1'i .i.vty ro1 .'.lev was sn service ai the lime- < the vvph'uca; : }:^ ;y. ; was isim-Pud by Shlllc-iT VasVM. Thi: i'osfo ic th-' plash.** ripped both rebmieri f ran tie evUimn. 1 >': mli>'r.`b hea-1* on both boilus r-'f-'-Uf* five the stu'K The. cj crating tie-nii-V ttuhe -beet to lute sheet) r.ppcarod to ;* in rr-;-.'li corniithm .xp.-rrolly. \Viu*n the .V oy, -- ' '>-. f.- ,_ .-up*.::red v:?ri; ".Ive'-vu. > r: . .. .e . F - c- . : . wait at the mye.'j w,;s ; a a. jr.iiSraiinsf '.nut the Ve'il'.ti'U: AUm-i :C I>.*: rapO:';: > ai slow cnOugU io Norm;1 yiiviln;;. i'J -j vapor retrri, m.'.v'rvi.: hoad iron: .ie open ti.nr rcte-te ;..*. .eve in !:is;:;rs d. Tlw he.:;! siiii'.vod -itiis tne tiivn s!v:-t lienirv n vv-.-v**;. i iii.nyc- tv= v mutirt.! wilh Lv*:;; ni A-unvi as llsoli^ii *v _ OcC''v j.:;;:ivn c-:' !S:.i!V.v;.y c-.er rr. the I*.ir-s of the eeinier. is r.ad-`ivicenr-e cf rr. overpressarc i.. . - . , * .* .. Caie:i;;itiv!!s ivt:.! i>j vstiir:;:'^ tiw i-resKortif which ivviv nK;::;.*:' i:. *.h-> sy.-f:'.: The Trenirne',:'. :s cc fwin-iT;' v-:."! calruisten to C fverr: ti. Sv;. I'.., C TO li.vvii ii>. -ij. n. sroiV-ij ni *.!'' : V ' .. ndniicv-n y;e;-: :-rsi?>.;rrc o: ii:c i;,;t,t.; -jir The >!!!; ; i..-.ri- rr:;.Mway cover v-'W:; ' . "toe -c>. In. jrt.upc ;it 1-0 ! i. i' ;f'i w visit! iltiVt hat: a Tllllil* mw.v. . ab-:-.: I,C`:v !!>. a;;, in. ynniTt: o: 120 5:. Tho l:i;:h ry :v: ''efin't:;' cs>t':?r*!, v.'t!. i'rtit'nre:} hy ;i h: 1 e v;-:; s.y j.G nti 1^. phyric -i '. ts^rv;:l:--r Msi^Tan'id':* t:;in stAJensent ; 2. rt:r . sin :i:f fv.- ct- na; tildes CCtaCi- sh *>.! ihr- cirri; awry j'r,;" :iiis revi/i;, 2. This area is-c h::;';:cl ^riewi i:y " I then propt'-t:';. of inibsinyy piect:.^.I lv fracture into rbmpamlively sn*a(t swerety sismu^rd piece*. Tin; kc.-.iicoi of :tit* i'Of'r.sc blast, hov/ever, docs n:>t picj'-'-iiit t\r: ovt'i oil iiwth.'itiism. It is theorized that a r-'ceiion c;!;o-;i in the toUr- of the tsp.ay.ti.nB' rc- ix-ilcr. Tii=t iyaciioit tu.ti sniiicit-iil energy to rupture the tishcr.. Miii reviSicd in r* pia;**nrc and temperature aiirvc in i hi. ha.-e of ti:e refitiing column. The pressure wro vjivrico.ily {dow to permit deformation of ihe icftf:;'?r miitwi he.-ui nmi the column manway c*rt*r. The h.wcr treys Wvvo lifted on their rings se- qnenli.iiiv. P.etwmi tray Nos. 10 and 18, tho combiua- inn a{ piv-'orc. tompcsviure. and ciimpt-nitinn led to h leiujir-tioii. The d!i. and physical evidence support the theory, in pnrticUar; 1. The ihcory o plxiro; why tray rings were left at- rached t > j.irces fmm the lower coiumu (sec Fig ure 2:. tin- iov/e.r trays wore out of tfie way, n detonation at tray Nos. 10 to IS would wipe out the tower tray rings by the same- piston action that was evident aiac e the n'->ion a-f-n. In addition, there Visut.o ' cvl.-1 .v.-4.:- ..t.-'. ;c ul >'in:a Ui O t.-lyi u tiofi o! tray sections. 2. The yir-ldiny c-f the reiKsler tubas indicates rela tively Tow si.inr failure. Conversely. The column above wrs :'i.'i:T:;rr i '::>! cio.;'.;;:;e failure with little evi dence ef yiehiir.g. Tiro separate event' are indicated: one slier, the otls'ir fast. S. The yidiiny of the- manway cover and the re tailor mitered head rlange indicates initial pressures above IOC* ib./sq. in. gauge, but because* the shell did . ! i. .... '...A :tss thou 1,0C`* lb. sa. in. gauire prior 10 detonation. Ceiciiir.i'on of proccau conditions Analysis >>1 opersting data--Fuhowij:-!? the explo sion. cp;r:f.dot:: were made available by the Texas l,;ty t'peratiuy t>epartri:e.'il. Those data, which were u &*>! in ir i= peri ion >:*f their,vest': gallon, areas follows: 1. !eufrncteti from tiie process ccmpLiter ter Uciohcr i'.'i i 1 * 2. (ir,:-.:': ,s recovered i;nm the trend recorders for October 22 11. J. 0: 3. inc-ci mentatior:. instrument rending*, and handvaivo r.'.'diions a* 7:2'1 p.m. en October 23 ; I. I'rpb.a! re:':!da:' scUimn system operating condi tions and snaiysei ( J; Thif;e shits indicate that the refining column system Future 3. Stuokr-'l trays thi remnitiin;; column. Figure A. One of tiie ropUutd Uihcs in t!:e reboiler operating at the time oi the explosion. i 030e> RS DO A l&l. C.0^"1 <S3 ; V' . A C'". 1 ) a *.. .* - .,t:' . J : - : : ? i T*" ;-i ; i ... - .. igyrc 5. The miter:?;.! hesd flange hem the epw-stiisg reboiier. Figure 6. Refining column base elevation. operated ncrmidiy nidi! : 1 a.:n. on the day of the ex- 'i r! ;/ *;;* rC ^.1.? * V*'`r- i*'*"*:*'li ;li rrC < * : t .-i., ViVit j'vutr.\ vi tliii w* `ittHiiu . *\ odor to pert'cvw wsairiieftKnei* elsewhere in the unit, "he feeds i<> the fa ;:-] iron ant' the j-er'-ihns column rt- i;;nJ-v;dviii .--if, tin. everv:.A pi cuvet m-A-w vs cn Loth 1 eh.'.!..-.::, were eioiwl, and the rftji-j-.e udvtt from tise ?;*.'. i:;;.' cohio::; v hv:-d-\ai'-ed ob. 7i: standby opm cf !h. rchmiuc colum?; 'Vm: 11 an.. ;-> 7:22 wa- or;;/;;;. A hav.\:\ emenn of overhead product ;v;Uiin:-d ft* flow through Its- m:a<'r valve: ::ts --;v`/;d deplv*ion of the muirriH: in sys-t.ift, h'iv {<.tinredo-;! in:; On' the Steam a/hi r-c r!u>: Hows bur; up me eight jitnirs prior to ;L? ex- piOiir-i:. The standby oner.-.;**.* of the rstiiung cidurm prior u the exniosit.a was intended t-> r*r a Iotal ceilux sitna- , l. th.it is. a AOuby-amie cnr'.hipv.o :.? follows: Xo {ireefiT-s i..!> -.i- c;ti of system 2. Constant a;!::;.' nn-moif'i.-e. vr/,, tempera tures. pve.;*::-o.-. g, 3. forrstnist !:pp;i:c- ;;t 2:1;. pvir.t in the .system, following the *>v.?i:-ie-ni paled irnreC-uin;ei^1 nflOr ces sation 0/ norm.-.-:. tinvous opera lion iihuarbiiation of 'he u.tia. tii-'o-her with a consideration ef tiro system i.fvnrrxtry, -.voeiu define lire operation ;:s !\-!i'A\s.: 1. l.'Kj'eariy-state. eh iUytifbtion of a reSariresy pure imtnrtterio i rarf-.n 2. iierrernliy det'reseii:^ system prevsnros, temper::- tnri-i. or.t! riows, hit; r.nii ff-ri'.err.i-y invreesiey te- !h>,\ ratio d. i.rry'' it; V: -1: *.f .;ao'r;;*; ie `-ho oyster;'.; srr-y a::..! rorio; ri eyi.rr ij-.i'j*.;; ;:!%'? yi'O;;;':r*':;i to onion?! eteity siit.u i t :-:-i-v::h -;' !>f iirt* ni-rs.;lintss e-'ei'iitioj-.s si -5-:* i.f ;t.e ;.<;*!'three r'ir-n. !:.- viti-.Oitr: ?'>. 1-. !*, m,.! rr!i :>'. 11 ftiri liot A iinu- in iio r.itri.ioeU. .hirnnse 6-i they atfoet the dntcon-e of u hr.teh diitiilation. Second. e;'r:;;u:t.:t;.t `d r?rv!<A yin;' r.rr :' -.' Vo's- ii:!t:s yitii i.fs> ciiicieney. Mot.id iiove in he ei-nsii ueieu to represent she steady-state operation prior t<= 11 K.m. A?!:: finally! Iho al-nnrnmi tiiK-rntio-i of the .-yinem. or the- wnstfA-dy-stste with '5;y;:i;aiioii. won;.; have to be sii.oviatcd by- a coiapateriaed 'nanie; and the ip.ieu'.a- *c.- . .... *... n'i.,, tatio:; of what existed in the rcrininp c-.'iein:; system at iho time of the explosion. The. ir.fovrr.aiinn dcvci- -pnd i?. used t-.. help .?Wr-111:;. i'r.e inf.-yimnisii; ;.-i t;:e ex- pivsien an*.] ;o tiefac ty.e.'-ratii'.y; eondiiions to prevent any yevcourre-jce. System holdup--Buriny `he rip1:1 hviirs and 23 min- ;;tes freni 11 a.m. tn 7:23 p.rn., the br.?e ievy] in the rofimnjr eoiumn dtere.r-ed. `i i:c chart on haso iictuio ieve! recovered from ii.-; i.-ond rerortier indicated that ii-.e ieve; had dropped to the elevation of the h;yh side ie.strmner.t tap. norjdc "tV", by 2:30 o.iti. (see Figaro fi), Exiivpidativn of the derresse in ii-.;`*i(! .'eve! pre- dii ted a ievc' in the vidr.lf y of the weld VieTween t;ie shei? arni bott-Mi) head 7:23 p.to. This jast above the eievalinn of iiw riser -bii ii ssippiiei; iio.aid to the east re lev (see Fi;mre hnl heJow tlmt ream'rets *0 tnain'..Viii n.-nnre; eirciibition throngli rho tr.bo-s. The I'Aii.'iibrii-d.overeii derresse in the ainmiiit wf Hspiitl in the eohirnjr base h from 3 !>00 at 11 a.m. to .`>30 gal. at :2*> pc ;i;. The ('p iivPicp' data Indicated cenernHy sieerrasiitit bean ].r':-s:;?e. reiinx and steam Oin-r.g the eight hour? and 23 'oinnU's prior to t::e expiosicn. Therefore, the tray aoi tlawncnmer he-idi:; :; had to be colvain'v-t \>ri->ns Fnics psv.r :> li'-o expi-j-ion. tn1- coost* (On'.oM-s ir *!?vc hoi.iap? w-.-re taken iniw a,-c*>unt iu the fa:..: hcaie-ilatiuns. Io.IiImp calcu- Acre made o r i::;or!y infrrt.ikioy into ;'e- muut !;!: svsiem i-i:a:;>.w-,` and :no .ii'-ss ps-oi.:? comiditm? 1 v:. tSariuit the eipiit hours ;;t:d n?ir- ii*e*. !< ;.'u i f a!i !r:iy aud 'i'lW.icoiiior t.'.hiupr -ie- cri-ai.cd hy d.'.i'.f i.rlTii iia . . 1 DO A 0306 C.ONF TDtrn Figure 7. Passage of time prior to the explosion as a function of the per cent distilled from the system. The reflux receiver holdup also decreased during the period of abnormal operation prior to the explosion. he chart recovered from the trend recorder indicated i hat the level fell from uS4 gal. at 11 a.m. to 140 gal- at 7:23 p.m. During the period of abnormal operation of the sys tem prior to the explosion, a total of 10,"0u lb. of over head product was removed from the system. This is 24.4*7 of the material present originally in the system at II a.m. Figure? comprises the rex nits of all holdup calculations, and indicates the passage of time from 11 a.m. as a function of the percent distilled from the system. This form of presentation is explained in the action of this article on unsteady-state operation. Steady-state operation--A computerized, multicom ponent, tray-by-trav calculation was utilized to simu late steady-state operation of the refining column system (!>). The relative volatilities and tray efficiency utilized v. ere available from laboratory and plant data within Union Carbide. The model derived was tested against the hisiorical operation of the Texas City sys tem under normal conditions, and the model predic tions were found to be in close agreement with actual operation. The steady-state model was "loaded" with a crude ieed composition representative of the operation on the day of the explosion. In addition to processing crude butadiene. the unit had been re-refining some off-specification product. This resulted in a crude feed higher in butadiene content, and lower in residue con tent than normal. The crude butadiene feed composition for October 23 was calculated by completing a component material balance around the system for the ten hours prior to II a.m. Data from the process computer log sheets (4) and analyses of the product and residue tanks (10) were utilized in the calculation. The feed composition, together with other operating data on the refining col umn, then was used in the stcady-staie model to deter mine the process conditions which existed prior to 11 o'clock. The component compositions winch were pre dicted for the reflux receiver, each tray, and the column base were combined with the system holdup calcub- ms to arrive at the system holdup for each component .mediately prior to the period of unsteady-state op eration. Unsteady-state operation--A computerized, finitedifference solution of the differential equations (which describe the multistage, batch distillation of a multi component mixture with stage holdup) was utilized to simulate unsteady-state operation of the refining col umn system (If). The relative volatilities and tray efficiency utilized in this model were the same as those utilized in thcssieady-state model. Because the unsieady-st.ate model applies during the eight hours and 4It minutes prior to the explosion, all the time "n viable operating conditions had to be taken into account. The tray and downcomer holdups, which decreased generally, and the reflux ratio, which in creased generally, were related to time through the use of second-degree polynomials and a least-squares tech nique (12). The reflux receiver holdup was related to time by a linear function. The material initially in the system at 11 a.m. was taken from the steady-state model, and material was "distilled" from the system in accordance with Figure 7. As each increment of prod uct was distilled, the program calculated a new elapsed time. This time was then used to adjust each of the Figure 9. The development of the vinylncetylene profile. DO A 030657 C0NFIDFNTIA1 65 tims variable operating conditions, and the process s related. The inventory in the column base was "nl.-itttt by difference each time, so no correlation ,/this variable was required. The batch distillation calculation was carried out until 21.-1 To of the iqplcrisil originally in the system had been withdrawn as product. At this point, the ma terial left in the system was representative of the pro cess conditions at the lime of the explosion. In Figure 8, final liquid-phase composition profiles are repro duced as predicted by the unsteady-state mode). Mole percentages of Tit-butadiene, vinylacolylene, and the other major components present ate given for the col umn base (K), and for the bottom *15 actual trays. The 1,8-butnciione content of the column base has been stripped down to about 2 mole %, while lire concentra tion rises with increasing tray number from the bot tom. The vinylacetylene content of the column base is 18C7. and a maximum vinylacetyleac concentration of j'J mole has up m in^ \ n.*.t,:' i; ;;j Nos. 10 to 15, caused by the depiction of butadiene and the subsc ent partial separation of vinylacetylene from the bu- Ter.es. Above tray No. 15 the vinylacetylene concentra tion decreases. Damage to the column extended from the br.se section to an elevation corresponding to the 33rd tray; the liquid-phase vinyhicetyicne concentra tion throughout this volume is predicted to have been upwards of -10 mole ;i. The area of maximum frag mentation of the column .shell centered around tray Nos. Id to Ji; the vinylasotyiig-e cuiiccuiration in this volume is predicted to have been in the range of 57- to 50^. A discussion of the results of this study r.aturally enters around the concentrations and quantity of vinyl.-.cetyler.e. because of its inherent thermodynamic instability and because of the tremendous release of energy which occurred. Referring again to Figure 8, the quantity of vinylacetylene present in the liquid phase, between the column base and the 33rd trav, was 4,S0d lb. The development of the vinylacetylene profile is de tailed in Figure 9. For comparison, n curve labeled "normal'' is presented. Such a concentration profile yould exist when tire refining column operates in a Steady-state, continuous manner, with sufficient resi due flow to hold 53 mole % vinylacetylene in the base. The curve labeled 1100 hr. is the result of steady- state, continuous operation on October 23. U has the same shape as the "normal" curve, but it originates at a base concentration of -14T7. At 1100 hr., the col umn was put on standby, and the remaining three curves show the development of the final profile. The curve labeled 1923 hr., reproduced from Figure 8, represents the conditions at the time of the explosion. The curves labeled 1258-. am! 1518 hr. represent in termediate profiles resulting /rent the distillation of 8- and 10of the material in the system, respectively. In summary 1. During the eight hours and 23 minutes preceding the explosion, the Texas City butadiene refining column i was operated in a standby node defined as follows: an unsteady-state, batch distillation of a relatively pure butadiene fraction. 2. a. In this mode of operation, the absence of feed ami residue flows, the depletion of material from the system, and tin* hivli reflux ratio brought alsuit ab normal. liquid-phase concentration profiles within the refining column. b. In particular, vinylacetylene reached concen trations of -18- and 40 mole 'i in the column base and on the 33rd tray, res;tcrlively, and a maximum concen tration of 59 inolc in the vicinity of tray Nos. 10-15. c. The system volume lioiindod by the description in 2.b. above contained -1,800 lb. of liquid vinylacetylene at the time of the explosion. 3. The liquid in the base of the refining column fell to a level above that required to maintain flow to the reboiler, but below that required to maintain natural circulation through the tubes. 4. At this point, a reaction occurred in the reboilcr tubes. Five of the lubes yielded to an interna} pressure and burst. 5. The reaction propagated to the base section of the column, deforming the mitered head on the re boiler and the cover on the column manway, and lifting the lower trio's off the support rings. 6. When the high pressure-high temperature front bcc-aatc p>.-"(I to .-* .'-tnlicienijy biVh concentration of vinylacetylene. in the vicinity of tray Nos. 10 to 15, the liquid vinylacetylene detonated. 7. The column shell was demolished in this area and tore from this point both up and down into relatively large fragments; ultimately, the lower 42 ft. of the vessel shell was destroyed. 8. Additional, widespread damage resulted from the shrapnel and from subsequent fires, caused by the re lease of hydrocarbons from the fallen refining column and from other, punctured vessels. # Literature cited 1. r.on-r-. L, li-ltcr to D. T. Witter* (January 5C, 1*70). 2. yrctmin. ii, II., mcruorannum to J. M, Jcnl:;nt 2i) *i al (December S. W\ G., nd C, C. Neely, tetter to D. T. Waitert* (February 6* 10701* 4. )ir.viiJ*i\v, J. I*., letter to V/. liurntft-ctt ct *} (November 10p 15Mr). 5. Iiuehter, 3. It.# letter to R. Tl, FrerfWiin ct si INovomber 14, 19(59). 6. -- -- -- -r letter to H. II. Frrvmun vt a? (Niivcrr.be* 29, 19G91. 7. t- ---r r , private communication to K. H- Tr^vm-n (Novem ber 7, IMS), 3. HU. B. Y,, private contmunfMi*rjt to R. H. Freemen on the use f Frf>crm No. for tray an<J downcomer hoUiup calculations (November US-2*, 1959). 3. 'VVunc* J- C.. private e>fnmumention to R. H. Fre'man or* the u*e of Prorrnnt No. 47472 for nintnlation >( complex fraclior.&ior (De- embvr 4, 1VJ5). 10. Kei*ter. B. G,, private communication to H. H. Freeman (December n* 11. Pcr*chn. T. F.. and P. L Salisbury. private tQTr.fnt;n;eainf>* to B, K. Freeman on the Uft?ef l'm^ram No. *4790 (Dvecmbtr 1040 to February 1570). 12. "AppffcMioiw Library manual FOLFJT'STA/* p. V-72, TimfvShar- Inf Compniimr Union Carbide Corp*. Tonaw.mda. N.Y, (December 1* 1359). frQ A 030658 conftdfntiai ...--^ ' " R. H. Freeman was responsible for the process calculations performed as part - - i of the investigation. Reed is a senior 1 \ process engineer in the Engineering DeJ partment at Union Carbide's Technical Center, South Charleston. W. Va. He has * 0 -. Bachelor and Master of Science Degrees . * \ . in Chemical Engineering from Mass.v N "* chusctts Institute of Technology and is j a member of ACS, AlChE. Tou Beta Pi -a. t .... --J and Sigma Xi. *- " M. P, McCready was responsible for the . physical investigation o( the incident. \ He is a senior process engineer in the 1 Engineering Department at Union Car- 5 hide's Technical Center. South Charius- : ton, W. Vo. He lias a Ibachelor of Science Degree in Mechanical Engineering from the University of Maine. He is a Itcgis- j tered Engineer >n West Virginia, and a L --*. **.iiJ member of ASME, Along with other findings, investigation of this incident determined that the maximum temperature for butadiene refining systems should be no higher than )0uC, R.G.Keister B.l. Tcsetsky and S.W. Clark Union Carbide Corp., South Charleston, W. Va. part of the investigation into the butadiene unit osior. at Union Caibidc's Texas City plant cn October . 1969, a laboratory investigation into tire stability of rinyhcciyL'ni'-butsdittic mixtures was initiated. Union Carbide's data and the literature indicated that vmylscetylcr.e in the diluents normally present in butadiene refining systems would be stable up to 48 mol %. An operating experience of over 28 years and the lack of documented incidents in tire six operating unit owned by Union Carbide confirmed the validity of these data. ompamon articles by other Union Carbide personnel (l, 2) describe the opera ling conditions at the time of the explosion, the calculations on concentration profile, and the results of the physical investigation, litis investigation, in essence, revealed that the butadiene refining column had leaked butadiene overhead while on total reflux, and generated high concentrations of vinylacetylcne in the column and low inventories in the base. The physical evidence indicated the initiation was in the reboiier and the deflagration propagated to the fourteenth tray where the detonation occurred. The objectives of the experimental work were to support the physical investigation into the cause of the explosion, and to establish the operating conditions that would prevent a leocct.urencc. This article defines a TXimmn hcaliug-niciiiuin temperature, identifies a sihle initiator, and more thoroughly defines the miting concentration level at which vinylacetylcne apois will support a deflagration. The following paragraphs on the individual phases of the experimental work discuss the thermal stability tests and the evaluation of initiators, define the upper limit for heating mediums in butadiene systems, and comment on the positive ignition tests that specify concentration limitations. Thermal stability tests The initial goal was to define the conditions which would lead to an explosive decomposition of a liquid-phase mixture of 1,3-butadienc (BD) and vinylacetylcne (VA). Of particular concern was the temperature at which the mixture became seMhcaiing. A thermal stability test which has had wide acceptance within Union Carbide was chosen to provide the base data. The test vessel, shown in F-l, used has a capacity of 106 cc. of. liquid. Its main body, constructed of 304 stainless steel, is a nominal 1 in. (the same size as the Texas City reboiler tubes) Schedule SO pipe having- an inside diameter of 0.956 in., and an outer diameter of 1.315 in. The bottom is closed with a 304 stainless steel, 3,000 Ib./sq.in. pipe cap. and the top with a Black, Sivalts and Bryson 3.000 Ib./sq.in., Type-IS, 1/3 in. scrcw-typc safety lic.nl containing a 347 stainless steel rupture dinphram rated at approximately 7,500 Ib./sq.in. at a temperature of 72*,F. Measuring appurtenances for determining temperatures DO A 030659 CONFIDENT I Al 67 ami pressures inside the. test vessel consit of two 1/16 in. dh. shielded, grounded-tip. chromclaiumcl thermocouples (time constant 0.1 see. in boiling water) and one lO.OiK) 'b./sq.in. air-cooled pressure transducer which had a esponse of 6.000 to 8,000 cyclcs/scc. One of the thermocouples is axially located approximately 1/2 in. from the bottom of the test vessel foT measuring material temperature, and the sccottd is placed against the wall of the test .vessel approximately 7 in. front the bottom of the vessel for measuring test-vessel skin temperatures. The pressure transducer is attached to the test vessel via a 1/4 in. 01) high pressure lube approximately 11 in. Jong to thermally isolate it from the system. The pressure transducer line has a Inch pressure cross in it to allow insertion of the thermocouple measuring skin temperature of the test vessel. Both the pressure transducer and the thermocouples are conuccicj to a high-speed Uniter Dynograph recurti mg potentiometer capable of a response of 100 j cylcs/sccond/cm. of deflection. Liquid chemical solutions are charged to the test vessel as follows: 1. Tire vessel is evacuated to less than 1/2 nun. mercury and then pressurized to atmospheric pressure with high-purity nitrogen. 2. After repeating the above procedure twice, the system is evacuated to less than 1/2 mm. Jig pressure. 3. The test vessel is cooled with dry ice. and the liquids to be tested are allowed to drain into the evacuated rest vessel. Care is taken to prevent nitrogen from entering. After being tilled to the desired level, the vessel is placed in a furnace consisting of a coiled 3.000 W, Calrod heater contained in an insulated, t gal. can. The temperature of the sample and container is increased at the desired rate by applying a fixed voltage to the heater. Usually 106 V. a. c. will produce a temperature rise of 5 to i0C/min. initially, the temperature rise will lag until the radiant healers reach their normal temperature. In the latter stages of a test, if it spans 200 to 300C, heat loss to the sui roundings. and also the lessening or the temperature difference between the heater and the vessel, may result in a reduction of the heating rate. The voltage is gradually increased dining this period to attempt to hold the heating rate steady. Deviations from the normal time-temperature curve (which is slightly S-shnped) indicate chemical or physical changes such as fusion, reaction, polymerization, or changes in heat capacity taking place within the test vessel. These phenomena are used to recognize the diffcicnccs in the materials tested. Heating rate With test vessel fills of 50%, a series of cxpciimcms was carried out to determine the temperatures and pressures at which various mixtures of BD and VA reacted, or exploded. A test of pure BD. slightly inhibited with niirosophcnyiiiydroxyi-amins (/vBH), carried out at a iaie of temperature increase of 5-to l0C/min. resulted in an exothermic reaction, probably polymerization, beginning at a temperature of 160C, and a pressure of 750 lb./sq.in. gauge. The vessel temperature rose to 285^0 in 16 mitt, due to the reaction, whereas without a reaction, the temperature would have risen only to approximately 200C in this period of time. Further increases in temperature to 360C caused no detectable reaction. No quantitative energy rsieasc data are given inasmuch as the system was not calibrated to yield such information. A vapor pressure type p'ut showed that at a temperature of 170C enough BD had reacted (probably polymerized) so that the rate of pressure rise witli temperature was no VI,tnylacfirlm 171 Fifjun. uernv'l stability test vessel. 68 Figure 2. Thermal stability test of a vinylacctylenc hutadienc mixture. DO A 030&6G CONFIDETNTI Ai. lonp-r maintained. Thus, with pure HD at this heating rate an exothermic reaction is indicated, but no explosion is observed. It is shown later that by increasing the heating m'" both an exotherm temperature and an explosion t etaturc can be generated. j illustration of the type of data obtained with a .lore containing 83 mol % BD in VA is shown in F-2. This material was heated at a rate of 5- to )0C/min. exhibiting an exothermic reaction at a temperature of 15SC. and a pressure of 590 lb./sq.in. gauge. The temperature rose to 2X5C at which point the reaction mixture explosively decomposed, and increased the pressure in the test vessel to failure of a safety diaphragm rated at 7,500 Ib./stj.in. gauge at 72F. From exotherm temperature to explosion temperature required about 10 min. elapsed time. Test results of mixture of BD and VA arc summarized in F-3. All samples of material, except those labeled Figure 3. Thermal stability of vinylacctylcne-butadicne mixtures. specifically as being from the plant residue, were mixed in the laboratory from relatively pure BD, VA. n-pentane (n-P), cis-2-butee, and trans-2-butene. The major portion of the results shown in Figure 3, however, represent mixtures of refined BD and VA. Most samples exhibited an exothermic reaction beginning at a temperature of 135to |45C, with reaction being observed at a slightly lower temperature with increasing VA concentration. Explosions occurred at lower temperatures and pressures he VA composition was increased. J'lte two threshold-temperature curves were eh awn through the nuuiimim points in each set. primarily to ' indicate the potential exotherm and explosion temperatures. In more complex mixtures, like plant residues, the initiation of an exotherm was more difficult to distinguish on the heating curve. The explosion temperatures, however, did occur at predictable values and are included in Figure 3. It is presumed that the presence of more thermodynamically stable compounds than BD masked the inflection in the curve. The influence of heating rate on the data can be shown in a run containing a mixture of 71 mol % BD and 29 mol % VA. This mixture was heated at the slower rate of 0.5* to 1C/ min. exhibiting an exothermic reaction at a temperature of 156- to 158C (similar to the faster heating rate) and a pressure of 480- to 520 !b./sq. in. gauge. Heating rate had a substantial effect on the temperature at which explosions took place, with the temperature of explosion increasing as the heating rate is decreased. It is believed that this phenomenon occurs because substantial amounts of VA and BD are consumed during the initital polymerizations. If too much VA and BD were consumed in the initial polymerization, the critical pressure, temperature, and composition requirements to obtain an explosive decomposition are . never met coincidentally if the rate of rise of temperature is too slow. If the rate of temperature rise was increased 30- to 40C/ min, even BD exploded provided the critical temperature. 200- to 340-'C. and pressure of 1.450- to 1,800 Ib./$q.in. gauge are met coincidentally. This effect of heating rate has been noticed by other experimenters studying the expJosibility of the peroxides of BD ft) Maximum rate of pressure rise for a laboratory prepared sample containing 55 mol % VA in BD was approximately 180,000 Ib./sq.in./sec., while it was 60,000 !b./sq.in./sec. for a similar-prepared mixture representing the calculated composition in the base of the Texas City refining column just prior to the explosion. In samples which exploded, yellowish, black polymers occured. In samples which did not explode, yellowish liquid and tacky, yellow polymers remained. A 50/50 mixture of bd/VA produced polymers in laboratory explosive decompositions which were very similar in nature to the polymers obtained from the Texas City explosion site. Such decompositions produced classic "ladder" polymers with a bicyclic and fused ring structure. Polymers produced via controlled thermal polymerizations in thermal stability tests were distinctly different from those found at Texas City. These were generally low molecular weight species with minimum crosslinking. Contaminants and additives results As was seen in the thermal stability study, any mixture of BD and VA can be considered explosive. When the composition of the mixture is 50 mol 7c VA and 50 mol 7 BD. the composition of material has little effect on cxplosibility at a healing rate of 5- to 10C/ min. For this reason, the above mixture and.heating rate were selected D0 A 030661 COMr IDE NT I AO 69 in an effort to determine the effect of contaminants and additives which mny i>c found in the HD processes. Test results employing a 50/50 mixture with vatious additives ->r contaminants are shown in T-J. j Materials which were considered possible additives in the BD proccs were paratertiarybutylcatechol fl'UC). sodium nitrite, and'nitrosophcnylhydrnxylaminc (NIMH. TBC is generally added to the reltncd BO in concentrations of approximately 100* to 200 pptn to prevent polymerization. Sodium nitiitc solutions (12 wt. % aqueous) or Nl'Il (5 wt. ft aqueous) aie generally added to the distillation columns to prevent the formation of butadiene polyperoxidc. polybutadiene, and "popcorn" polymer formation by scavenging oxygen front the system. Sodium nitrate, copper vinylacctylcnc, oxygen, and residue gatheied fn'.-i the rob.-iV-:- of the explodvi! distillation column were considered as possible contaminants that could sensitize the reactivity of the mixture. The residue solids were analyzed to contain 80ft organic polymer with a sodium nitrite to nitrate ratio of approximately 1:2. From the results shown in Table I, only the addition of a 12ft aqueous solution of sodium nitrate apparently lowered the explosion temperature. All other tested materials considered as additives or contaminants did not alter the results. Literature indicates that the use of over 4% concentrations of aqticous sodium nitrite solutions may be hazardous. The Phillips Chemicals Co. found organic nitriles, which when dry, decomposed explosively when heated above )50C even in the absence of air. These compounds were not found when more dilute nitrite solutions arc utilized and when the pH of the system is maintained above S (4). Phillips maintains aqueous solutions of sodium nitrite of not over 0.5ft. Because they used robot'-r 'team at 2C'7C,a dried polymer which fires at 150C was considered dangerous. The initiation of reactions shown in Table 1 began at Additives er Contaminants Table 1. Effect of additives and contaminants.* Added as wt % Aqueous Solution Weight Added ft of Mixture Conditions at Start ' of Exotherm Temp., C Pressure. Ibysq-in. gauge Hone (control)...................................... ...... -- 141 520 Fara Tertiarybutyl Catechol.................. -- 9 ppm 133 500 70 ppm 143 560 Sodium Nitrite.................................... 0.5 5.0 1421 510 .................................. 12.0 5.0 146 590 Sodium Nitrite.................................... 0.5 5.0 140 520 ................................... 12.0 5.0 143 590 Replicate........................................... 12.0 5.0 140 590 Nitrosephenylhydrazine...................... 5.0 0.1 147 535 Coppervinylacetylene........................... -- 0.1 * 140 490 Oxygen......................................... i.. -- 50 ppm 139 545 Solid Residue Gathered From Calandria Tubes ol Texas City Butadiene Column After Explosion 22 133 74 150 500 350 Temperature at (**) Start of Explosive Decomposition, C 185 175 173 186 175 ISO 166 160 190 186 173 230 210 Base mixture 50/50 mo! ft vinylaceiylene and butadiene. Ruptured diaphiDtjins rated at 0,000- to 7,500 Ib./sq.in. ijiugn at 72F tailed in all eases except test 62 where a partial charge of material was used. Ali samples tested exploded. t In 2 out ol 4 tests no exutherm was ob `J before explosive decomposition occurred. DO .... 066? CONFIDENT! Ai 70 temperatures ranging from 13S- to ISO'^C, and pressures ranging from 480- to 620 lb/sq.in. gauge approximately 50% full. Hxplosions took place at temperatures ranging rrom 160- to 230`*C, and pressures ranging from (>50- to X))b7stj.ii'. gauge. The lowest temperature (|60C) for ,e explosion was in the presence of sodium nitrate. All explosions led ^t> yellowish, black polymers (black probably because of imbedded carbon). In the ease where a smaller than usual charge was used, with just-wet polymer residue from the Texas City BD column calandria, the decomposition ivas contained and the test vessel was full or carbon. The polymers produced in explosive decompositions of mixtures containing additives were similar to those previously observed in explosive decompositions. Pressure Pressures occurring in thermal stability tests just ptior to explosions were in the range of 600- to l ,800 ib./sq.in. ugc, considerably higher than the pressure in the base of -me Texas City refining column at the time of the explosion. To determine the effect of ptessure on test results, the charge to the test vessel was decreased and the __ normal test procedure was followed. To minimize the reaction occurring during the heating, all tests were performed using a rate of temperature increase of 30- to 40t>C/ min. As the charge to the test vessel was decreased, the temperature and pressure required for decomposilon generally increased and decieased respectively. The ratio f maximum explosion pressure to the pressure just prior to the decomposeicu, P;Pj. decreased with a decreasing sample charge. It is thought that this decrease was due to energy losses in the laboratory equipment becoming substantial, compared to the energy released, as the charge to the test vessel was decreased. This ratio should be fairly constant because all of the material tested was probably in a single phase when explosive deccmpositons occurred since critical temperatures were exceeded in essentially all ases. Although BD was extremely difficult to initiate, it ' readily contributed to the explosion while a true diluent such as n-P, a thermodynamically stable compound, did not contribute. Thermal stability test results of laboratory mixtures of BD-VA, and plant BD residues showed that all mixtures will decompose explosively at temperatures ranging from 170- to 400C. All laboratory mixtures exhibited an exothermic reaction, probably a polymerization, starting at a temperature of 135- to. 145C. This polymerization could lead to an explosion if the heat removjl capacity of the system is exceeded, and the necessary temperatures and pressures are realized. With BD alone in the system, heat must he applied at a rate of 30- to 40t'/ min. to cause an explosive decomposition. With 50/50 mixture of BD and VA, a ihcnnal initiation of a reaction is adequate to result in an explosion at predictable temperatures with the polymerization reaction generating the heat. As tire VA concentration is increased in the system, general!'' lower temperatures - and pressures are required for initiation of explosive decompositions. Although BD is considerably more stable than VA, it can contribute energy once an explosion occurs. Tests generally have shown that higher decomposition initiation temperatures arc required as the pressure in the system decreases. Heat soaking tests Although the thermal stability series of tests indicated that all materials present in BD systems can generate an exothermic reaction that can culminate in an explosion, the time at temperature was not considered. Relating these tests to systems that operate at temperatures considerably below those indicated for initiation of an exotherm must include the time variable. Two series of tests were, made to follow the time-teinperature relationship. The tirst to find tire effect of temperature on the stability, and the second to provide the basis for a recommendation of the maximum allowable temperature limit. In these first tests, a 50/50 mixture of BD and VA was maintained at temperature of 100- to I35C for various periods of time. The outside of the test vessel was generally maintained at the designated temperature and tire inside of the lest vessel was allowed to vary and to respond to the chemical reactions taking piece. Tire thermal stability apparatus, previously described, was modified to include a thermocouple strapped to the outside of the vessel approximately 4 in. from tire bottom. This temperature was controlled by using an on-off controller; tire inside temperature and pressure were continuously monitored. Tire vessel was charged approximately 50% full as in the thermal suability tests, A sample was held at approximately I00C for 9.5 hr., cooled to nmhient temperature, and then run through the standard thermal stability test procedure. At 1S0C and 600 lb./sq. in. gauge an explosion occurred which compared well with the thermal stability results of a similar mixture which had not been aged (I40C and 520 lb./sq.in.gauge). A second sample held at 100C for 16 hr. required 240C and 700 lb./sq.in. gauge xo cause an explo sive decomposition in tire stability test. The increase in temperature required for auiodceomposition probably stemmed from substantial amounts of monomer polymerizing during the aging. In a third test, conducted for scvcr( days at 100C, the test vessel was found to be filled with a yellowish, soft, tacky polymer, and no liquid remained. Relating this to the operation of a system, a rcboiler with a maximum of lOO^C steam would only polymerize a 50/50 mixture of vinylacttylene and butadiene, not explode it. In a similar test, a 50/50 mixture maintained at I20C did not explode after one week's aging. Repealing this experiment at I25UC, an exotherm and an explosion occurred after 3.5 hr. At 135C, the explosion occurred after 1.9 hr. Although the test vessel was held at the 00 A 030663 COhtF IDFNTTAt 71 indicated temperatures, tire material inside the vessel increased in temperature and finally exploded. The inside walls of the lest vessel and the thermocouple in the liquid phase were found to he heavily coaled with polymer. In ormal thermal stability tests conducted at a 5- to )0C e of temperature rise per minute, only small quantities ,f polymer were found in the lest .vessel. In previous lasts where pure butaJTcne was decomposed explosively, quantities of polymer found were even smaller and were located primarily on the failed rupture diaphragms. This work suggested that the maximum calandria steam temperature should be in the ranee of 100- to I J0C. To develop a maximum temperature recommendation, a series of 20 tests was made using a 1 in. dia.. 3,000 Jb./sq.in. pipe cap as a test vessel. The cap was scaled with a pipe coupling and a high-pressure valve. Tire cap was charged with 1 cu. in. of liquid (approximately two-thirds full) end placed in er. oven control!.-! ;:t l!0C lor on" week. After cooling the caps and opening them, samples of the residual polymer were examined visually and with a Tiinc-of-Flight Mass Spectrometer to determine whether "an explosion had tal.cn place during the aging. All samples had polymerized in controlled reactions to form a yellow polymer typical of nonviolent reactions. The 20 no-go ' results at ]J0C resulted in a probability of explosion that had a statistical 95% confidence range from 0- to 5%. Based on this wo;!;, a maximum calandiia steam temperature of 105C was recommended and has been implemented in Union Carbide plants. ative-ignition tests Mixtures of BD and VA have been shown to be capable of being decomposed explosively in nil proportions in thermal stability tests. Because of the nature of the tests, pressure control was not achieved during th testing, and the bulk of the tests were carried out at pressures ranging from 700- to 1,500 lb.sq.in. gauge to achieve explosions. These pressures arc far in excess of those occurring in BD fining columns which operate at approximately 50 itf./sq.in. gauge. Although the thermal stability test was useful for determining unsafe temperatures, it was believed to be too severe to be used to predict unsafe compositions. Hence, a series of tests was devised to determine maximum safe VA concentrations as a function of refining column operating pressure. Throughout these tests the liquid and vapor phases were essentially in equilibrium with each other. This series of tests was carried out in a 4 in. dia. (2.410 cc. volume) test vessel shown in F-4. A 4 in. dia. vessel is suitable for predicting safe pressures in acetylene deflagration tests, and hence was deemed suitable for this study based on Union Carbide experience. In addition to the accessories described for the thermal stability test "fscl, the system was equipped with a pressure gauge and ee firing devices. The charging procedure was similar to it described previously. Normally 100 cc. of the mixture were dunged, providing a ratio of vapor to liquid equivalent to that present in the Texas City refining column. The liquid in the test vessel was healed for about 2 hr. until the pressure was lined out at the desired level. A vapor sample was then collected into an evacuated 3 cc. stainless steel sample tube. Tire pressure gauge was then isolated to protect it, and the first igniter was fired. Results were monitored using a Beckman Type R Dynograph Recorder. The first igniter was discharged at the lowest pressure in the series, and if this failed to initiate a vapor-phase deflagration, the pressure was raised 10 lb./sq.in. gauge (by increasing the temperature) to the next, and the second igniter discharged. The ignition sources were either the fusion of Nichrome wire or the discharge of a No. 6 high-iempernture electric blasting cap. The lused-wire igniter consisted of two strands of 30 gauge Nichrome wire in parallel, 0.6 in. long, fused by the discharge of a 17,000, farad condenser charged to J50 V. d.e. This ignition system has been found to produce slightly more than 3.2 joules of energy. The No. 6 blasting cap liberated approximately 2,000 joules and created an instantaneous pressure of approximately 200.000 atmospheres (5) and had a propagation velocity of 6,000- to 7,000/m./sec. Both sources have been used successfully by Union Carbide's Fire Research Group to predict the explosiveness of various materials and to determine safe compositions. Explosions have never occurred in Union Caibiue plants which operated within safe limits predicted using the above ignition sources. Vapor-phase testing was done in preference to liquid-phase testing because it was believed that for propagation to be continuous in a distillation column of the type which exploded in TexasCity.it had to continue through the vapor phase. Secondly, a large number of liquid-phase ignitions had already been completed in terms of thermal stability tests, and conditions for initiation of the liquid had been established. |/2*i3tk M It H(4 Atitmbly Tit* t*qu4 ikii tvttlMr rr/Wr4 l* th# by tit i * towut ttvitn #: * furwt* cw>ttiH tf * C*h*4 tr*ir. Tkt ytril-wiM^ cr# mi e**itmi< s*is Figure 4. Positive-ignition test apparatus. r*i prr*wr* NM*W)tt . 72 DO A 030664 CONFIDENT TAI. The tests were marie in the vapor phase with mixtures of BD and VA, mixtures of BD, VA, ami n-P and plant HD column residue fortified with VA. Pressures were varied ' 'in 65*- to 85 lb./srj.in. absolute in 10 lb./sq.in. -ernents. i est results and conclusions Pure VA exploded from a startin'; pressure or 85 lbVsq.in. absolute, using a No. 6 high temperature electric blasting cap in the vapor phase to yield a final maximum pressure of 4.800 lb./srj.io. absolute in 0.02 see. The vapor-phase deflagration initiated the liquid explosion in approximately 0.01 sec. when the system pressure was 500 lb./sq.in. gauge, and the vapor temperature approximately IO0C. These conditions were in good agreement with the thermal stability test results which indicated that VA explosively decomposed at a pressure of -150 It.'./sq.in. gauge, amt a temperature ut i57v`C. In other words, temperatures and pressures had to be equi- ant to those existing on the upper curve of Figure 3 u-ei'ore a decomposition of the liquid phase would occur. This equivalency reconciled the thermal stability test data with the vapor-phase positive-ignition data, despite the disparity of the operating pressure (60) versus the test decomposition pressure (<500 lb./sq.in. of the op?rz::r? pressure (60 lb./sq.in. absolute) versus the test decomposition pressure (< 500 ib./sq.in. absolute). In laboratory mixture of BD-VA, vapor samples ntaining 53- to 55 mol % VA deflagrated in the vapor ase at 65 lb./sq.in. absolute pressure using the chrome-wire ignition source. A typical time delay to * Figure 5. Typical pnmivc-itiniiion test. develop the pressure and temperature required for decomposition of the liquid is illustrated in F-S. Again they arc in fair agreement with those recorded in the thermal stability tests; an accurate reading is handicapped by the different time responses of tire pressure transducer and the thermocouples. At 75 )b./sq.in. absolute, a vapor-phase concentration of 51% VA in BD resulted in an ignition, while at 85 lbVsq.in. absolute pressure. 41 wt. % VA was unstable. In the above two cases, ignitions were attempted at 65-and/or 65-and 75 lb./sq.in. absolute, respectively, without a vapor-phase deflagration. As in all cases for BD-VA mixtures, a deflagration in the vapor initiated an explosive decomposition of the liquid. Tire vapor composition was measured at the lowest ignition pressure; the actual VA vapor composition may have been somewhat Higher because increased pressure was developed by \upoticing additional liquid. At S5-lb./sq.in. gauge with a vapor composition of 41to 44 mol % VA in BD, there was one explosion in seven tests using a fused wire igniter, none in' two tests using a blasting cap in the vapor, and none in one test using a blasting cap in the liquid. The blasting cap was not a more effective ignition source than the Niehromc wire despite the greater energy released. In one test, a 50% overcharge was used. Although the system was ignited by a blasting cap, propagation was slow and, hence, there was no measurable contribution of pressure developed due to the cap which develops its pressure almost instantaneously and only contributes about 1- to 2 !b./sq.in. detectable with measuring instrumentation. The normai charge (100 cc. volume) yielded an explosion pressure of approximately 1,400 lb./sq.in. absolute, and the 150% overcharge yielded 150% of the pressure of a normal charge, or 2200 !b./sq.in. absolute. Tire final pressure developed in a partially filled system was a function of the quantity of explodable charge, therefore, the vessel was large enough that the wall effects were insignificant. Tire reasons for the delay in the explosion after the blasting cap ignition are not known. In the tricomponent BD-VA-n-P system, 30 mol % n-P replaced part of the BD in the charge liquid. The vapor phase deflagrated with 70 mol % VA (23 mol % n-P) at 65 ibVsq.in. absolute, and with 63 mol % VA (17 mol % n-P) at 75 lb./sq.i. absolute. This latter test was the only test in which the vapor phase did not decompose the liquid. A vapor composition of 60 mol % VA (13 mol % n-P) did not deflagrate even at 85 lb.sq.m. absolute. A sample of residue (kettle liquid) from a Plant BD refining column (36% VA) was found to be stable even at 85 lb./sq.in. Absolute. This material was fort ified with pure VA to produce mixtures up to 61% VA. This composition exploded at 75 lbVsq.in. absolute, white samples contain ing from 55- to 5S?i- VA were unstable at 85 lb./sq.in. absolute, bursiableat 75 Ib./sq.in. absolute. Mixtures con taining ft on* 53- to 54% VA exploded once in eight tests at 85 IbVsq.in. absolute using both ignition sources. n0 A 030665 CONFlorNS-U"1 73 The positive-ignition tests show that with vapor-phase VA-BD mixture. -40 mul % VA will not deflagrate at 85 tb./sq.m. absolute, nor will 50 mol % VA at 65 lh./sq.in. absolute. Adding the components normally piesent In butadiene column residues varies the allowable concentration at 85 lb./sq.in. absolute to 53 mol % VA, and at 65 Ib^sq.m. absolute to 60 mol %. Addin" an inert component such as-n-pcntane to a VA-BD mixture raises the allowable VA concentration to higher levels. These data arc shown in l;-6. In this work, in which the liquid and vapor phases were in equilibrium, the vapor-phase deflagration generated the necessary conditions to initiate the liquid -phase decomposition in all eases except one of the tests using pentane. Based on the above, if plant BD refining distillation columns are treated ns though they were mixtures of 110 and VA, there isa built-in safety factor of about 10% in VA concentration. An additional safety favor r\-y be built in if the column is operated at a pressure less than that for which the safe concentration of VA was ^ developed. It is the authors' opinion that 45- to 50 rnol % VA could be considered safe in a cclmnu operating at maximum prersure of 65 lb./sq.in. absolute, if there me no excursions above 50% VA. The degree to which the VA concentration can be controlled is a factor in establishing the safe opciation limit. Since the data arc available to adjust for opeia;:!-.' pressure and experience is ieq:.:;to determine control variability, a maximum of 40 mo! % VA was chosen for systems operating at 65 lb./sq.in. absolute. The experimental work on the stability of vinyhcetylene-butadicne systems generated the following conclusions: 1. Ail mixtures of vinylacetylene and butadiene are thermally unstable, and will generate exotherms and explosions at predictable temperatures and pressures. 2. Sodium nitrite, conceited to sodium nitrate, can make buladicne-vinylacetylcnc mixtures less stable. 3. The maximum temperature for butadiene refining systems should be no higher than I05C. 4. Concentrations of vinylacetylene up to 40 mol % will not support a vapor-phase deflagration at 65 lb./sq.in. absolute. 5. The temperature and pressure developed from a vapor-phase deflagration is sufficient to initiate an explosive decomposition in the liquid phase. . As a result of the investigation into the explosion in the Tecas City No. 3 butadiene refining column and the subsequent experimental work, the cause can be attributed to the following sequence of events: 1. The butadiene refining still was placed on standby, as .was the usual practice, to perform maintenance on equipment in the system. ' 2. Under near-total reflux conditions, the overhead rroduct valve continuously leaked butadiene from (lie listiliation system. 3. The column fractionated, over S.3S hr., the light V 74 Figure 6. Positive-ignition tert of vinylacetylene mixtures, components overhead and concentrated the heavier components in the lower section. The trays between 10 and 15 from the bottom developed a maximum concentration of 60 mo! % vinylacetylene. 4. The base liquid level fell to a level above that required to circulate the organic liquid, but below that required to maintain all the tubes covered in the natural circulation reboilcr. 5. At this point, a reaction occurred in the reboiler tubes. The exact initiator is not known, but a thermal polymerization sensitized by the presence of sodium nitrate, is a possibility. Since the 15 lbVsq.in. gauge superheated steam has a maximum temperature of 157C, the reboilcr level was low, and the condensate temperature was rising, the conditions for a thermal polymerization were present. 6. Five of the tubes yielded to the internal pressure and temperature and burst. 7. The shock and heat generated by this event decomposed the highly concentrated vinylacetylene vapor in the reboilcr causing a deflagration wave to proceed up the column. 8. The pressure and heat-from the wave over-pressured the reboilcr vapor outlet, the bottom of tlto column, and displaced or tipped, the bottom ten distillation trays. 9. On or about tray number 14, the Temperature and pressure were sufficiently high to decompose the liquid containing about 60% vinylacetylene. JO. The column tore from this point up and down at velocities over 2,000rt./see.'destroying the bottom 42 ft. of the column and causing it to fall. DO A 030066 CONFTDFNTTAi. H.. Tlie release.or the large amount of hydrocarbon from the system surrounding the column ignited. Union Carbide, as a result of this investigation, has implemented the conclusions of ibis report and reunited heir other units to comply with the following .jwimcndaiions: 'r1. Avoid total reflux to prevent the concentration of any component in any*part of the column. 2. As long as there is heat on the rchoilcr maintain a residue Hoe/ from the system at alt times. 3. Do not exceed 40 mol 7c vinylacctylcne in the vapor phase at any point in the system. 4. Provide continuous monitoring of the kettle vapor phase and approximately the sixth tray vapor phase for vinylacctylcne concentration. The column concentration should always be less titan the base concentration. 5. Reduce all heating systems for all butadiene refining columns to a maximum ten'rorature of 105C. 6. Do not use sodium nitrite in the refining system. For polymer and peroxide control, use 10 ppm trosophenyl-hydroxylamine (NPH) in the reflux loop, cased on the reflux flow. Acknowledgement The content of this article represents the combined efforts or a number of people from all levels of Union Carbide. Advice and assistance from Dr. W. B. Howard of (lie Monsanto Co. and Dr. R. A. Schultz of Du Pont is sincerely appreciated. Dr. Paul Johnson, also of Du Pont, provided the vinylacctylcne for the experimental work, Tire cooperation of the industry, in general, was gratifying and it is hoped that this report repays, in part, for their contributions. Literature cited . 1. Jarvis, H.C., Chem. Eng. Prngr., 67 41 (June 1971). 2. Freeman, R.H., and M.P. McCrcady, Chem. Eng. Erogr., 67 4S (June 1971). 3. Alexander, DS.t hid. Eng. Oicm.. SI 733 (June 1959). 4. Campbell. G.G., "Potential Hazard of Using Sodium Nitrite for Inhibiting Popcorn Polymer." Phillips Chemical Co., Report No. 49-58PC (April 23,1958). 5. Rucker, K.G., private communication to the author (October 8,1970). KEISTER, Robert G. Bernard 1. CLARK, Stanley W. D0 A C306&7 COWFlDFNTTrtt. OfYi. --p &' I ;s % 'w:'. 7`**I PLANT SAFETY & LOSS PREVENTION Butadiene Explosion at Texas City--1 The force of V'e ! ex.=s C;t" ?re!oscn sc?':fared l.arjje frc^i^snts to a r^r**'tr ri/,1/ A1 fvvi/ /* Ii and blew one 00 cccticn 3.C00 fi. H. C. Jarvis, Union Carbide Corp,, Texas City, I ex. An explosion* nc'rr.-.r.cn on OrTO-rn 20: 19G9, in the year-old butadiene refining unit ;:t Union Csrhiiix's Texas City plant. Much of the butadiene unit was de stroyed and the adjacent i.2 billion It:., yr. ethylene unit suffered major damage. The explosion. which oc curred in the !ev*iv*c-D. Jitteivd the area within 5iiu ft. v: cue column with shell and tray fragments, and tray valves varying from a few ounces to several hundred pounds in weight. Large '-ell fragments were scattered to a radius of 1,500 ft. ,i cue S00 lb. section landed 2.000 ft. away. Crass res were set at a distance of 750 ft. Damage was not confined to the plant site. Numer ous houses in a residential area starting 750 ft. away were damaged bv flying materials and windows were broken l14 mi. away. FortimatMy. no disabling in juries occurred, am! injuries outside the plant area were confined to minor cuts and abrasions and menial anguish. The butadiene refining unit recovers byproduct butadiene from a crude C| stream originating in the jfwss unit. Similar units have been operated by Union wiirbidc for almost 50 years. The primary separation of butadiene is by absorption. Decently built units, including the one at Texas City, diner from the earlier units in that a new absorbent. dimcfbylncctamide (DMAC). is used. The stripped gases from the satu rated solvent are compressed for recycling to the base of the absorber ns a beat medium and as feed to the refining sertion ot the unit. Light components of the stripped gas are removed in a forecolumn. The high purity butadiene product is .produced as the overhead of the refining column. The heavy components of the feed stream, including vinylacetylenc, arc removed as refining column keltic product. Vinylacoiylene in the kettle product is normally maintained at a concentration "f >m |{5'i by sup plementing the heave ends in the feed stream with `iladicnc. l:!.vpti:.:)blli*,y tests h.tu im.iraled that n>n- ''fraliims of \ iny l.i elylen-: under .*(>'; in the kettle iuct were stable at eoUimn operating conditions. ine RO column years of refining eolonm operation imdrr column rm-.ifi!i,.es similar (, these in the Texas City unit hail tn.-eu annumkited willnmi ineidont. Status prior to the explosion The butadiene refining unit was operating normally, and all other pieces of equipment were in good work ing order when the unit was shut down for repairs to the strirpov make cempre.-^cr. Pcdm-i'Vi: of f-. --.i How to the absorber began about 9:00 a.m. anti ail feed flow was discontinued by 11 :()0 a.m. Circulation of DMAC was continued until all dissolved gases were stripped out and flared. The absorber and stripper were valved in under methane pressure in a standby condition, and the stripper make compressor was purged with inert gas and prepared for maintenance. The forecolumn and refining column were placed on total reflux when the stripper make compressor was removed from service. The columns were shut in with hand vaives in the feed and kettle product lines and by motor valves in the overhead lines. Doth columns continued to operate under total reflux conditions--the normal practice for these columns when the absorption section is shut down for short periods of time. The forecolumn operated smoothly under total re flux at near normal end conditions. The m:iy change in conditions, a reduction in vapor gravity at the ton of the column, was probably caused by an accumula tion of light ends during the total reflux process. Operation of the refining column appeared normal to the operator on duty even though operation was er ratic. This column normally operates under a very high reilux ratio and erratic operation on total reflux was conimmonl Operator attention was repaired in main taining a balance between the liquid in the base of the column available for kettle vapor generation ar.d liq uid in the accumulator available for reflux, Keiliix flow was controlled on manual while steam flow to the calandria was controlled automatically by the liquid level at the base of'the column. This nrlid,', ,i ml the one Unit follow*, ore live of three tn I ides written on the Texas City explosion mill its nflerinnth. Alt three uelides mill he i,nhlisheii xhorlhi in n CKT technienl innnnttl; "[*>$* Prevention," volume S. 4 DO A 0306^8 cNFID8NTIai 57 Although operation of the column appeared normal at the time, examination of the records after Ihe ex plosion showed that the column was slowly losing ma terial through the closed, hut leaking, motor valve in 'the overhead line, Column pressure and pressure drop vtahilized at ?**- UiW.'r of normal showing reasonable* ray loading following a rapid decrease when feed was 'shut otT. Reflux ami steam (low continued to fail slowly throughout the shutdown period, indicating a loss of material from ihe column. The make How meter.showed acdutinuous (low; however, the operator assumed that the meter was otf calibration since the make motor valve was closed and the tracing on the chart was a continuous straigii! line near tnc base of the chart. The column has.-; ievel indicator showed a low level in the base of the column, blit ample kettle vapor was being generated. Loss of butadiene front the system through the leak ing overhead line motor valve resulted in substantial changes in tray composition in the lower section of the column. The concentration of vinylacetylcne in the tray la;:.id in Is.: ci.ixity <.f tin- ten!;: r.wy ..:ly more than doubled '*o c.u er.tirnaied CO mole c'c. Tiie loss of liquid in the base of the column uncovered the calandrir. tubes, allowing the tube well temperature to ' approach the temperature of the heat supply. The in creased vinylaectylcne concentration and high tube wall temperatures set the stage for the explosion which followed. The explosion The explosion in the refinine column occurred some 1J hours .V.l;-:* v . lion in fern uvw am; !> :.- ;;rs after the refining section was placed on total reflux. The explosion occurred without warning other than as covered in the paragraphs above. It was so sudden that none of the pressure and temperature changes which accompanied it were recorded on any of the instru ments in the unit. Two loud noises were produced. The first was probably caused by the disintegration of the column in which the lower -10 ft. was fragmented and scattered around iiie area. The second, which occurred almost immediately thereafter, was probably caused by the ignition of the is rye quantity of gas released from the ruptured linos and equipment. The.extent of physical damage to the butadiene unit is shown in Figure 1. One distillation column was re placed and sever;:' others received major repairs for deformed shells, shrapnel holes, and distorted or miss ing nozzles. A major part of ihe auxiliary equipment was repaired or replaced. The piperack, including pip ing, instrument and electrical cable, and most of the instruments, was replaced. Damage was similar but not as extensive in the adjoining otetins unit, although evaluation and repairs were more complex mu! costly because of the special metals and insulation required for cold temperature service in this unit. Rebuilding costs exceeded >(1 million. Determining the absence of damage was of major importance for economic r.nd safety reasons. The thousands of pieces of (lying metal, some very small, cut holes and gashes in insulation and equipment. Heat effects caused stresses and mc.nl distortion which, in turn, caused mechanical strain in attached equipment. Temperatures over l.lioO K caused grain growth in cold temperature sleel rendering it brittle when cold and unsuitable for reuse. All piping and equipment alferted by the explosion and lire were inspected for damage. Carlton sleel pill ing for warm li-mporatiuv service was considered 58 nielallurgicaljy sound unless visual inspection showed severe scaling or distortion. Stress relief was require*! for cold temperature steels heated above ,`t7-VF. Grain growth was assumed in steels heated over- 1,100'K. The color changes below were used ns indicators of the temperatures reached. The validity of this approach was established by field and laboratory testing. 1. Foam glass cement melts at 280"F 2. Polyurethane insulation turns dark at 250`F, and chars at 300`F 3. Phenolic resin in the fiber glass blankets turns brown at 375`F 1. The insulation weather barrier chars and be comes chalky at 100'F 5. Dimeleote No. 4 turns brown at 1,000*F G. Foam glass melts at 1,800'F Effect on personnel Thirteen people were in the immediate area when the explosion occurred; G in the control building and 7 in the equipment area. Three maintenance men work ing *: ihe romrf-':'-10 ft. away from the refining column were shielded by small items of equipment and thrown to the ground by the blast. One suffered a rup tured ear drum and the other two were uninjured. One operator, approximately 200 ft. away and relatively unprotected, was thrown to the ground but uninjured. The main fireball was over his head, but he crawled out of ihe fire area and immediately took steps that limited the amount of flammable material entering the fire zone. He also activated the emergency relief valves on the distillation eouiriment. His. quick action was in strumental in limiting the extent of the tire. Six mm were in the central control building 200 ft. from the refining column. Although this building took the full brunt of the Mast, no serious injuries occurred to those inside. One man seated in an office downstairs near the wall facing Ihe explosion was shaken up and suffered head bruises, but lie was able to clear the building and the area between the initial blast and ex plosive burning of the released gases. The four men on the control room floor made their way out among falling light fixture; and ceiling parts, and escaped from the area. One of these men suffered a broken heel which was probably cracked when he jumped to the sidewalk from the intermediate level of an outside stairway. The remaining men in the area were far enough away that they were not affected by the blast. The fire was brought under control by unit person nel and the plant fire squad. Employees living in sur rounding areas who were familiar with the unit re sponded immediately to the emergency and assisted personnel on duty in bringing the fire under control. The fire fighting effort was impeded by the terrific noise created by high pressure steam and gas escaping from ruptured linos which made communication al most impossible. Conversation was difficult 800 ft. away from the blast area, and the noise was discom forting even with ear muffs. Fire protection is provided by a water spray system, stationary fire monitors, and hose hydrants. Total fire water pumping capability is 37.GOO gal. min. Six of eight fi in. water spray .system supply headers passing through the blast area were destroyed. These lines sup plied water to portions of the adjacent olefins area, as well as to systems in llv immediate blast area. Loss of those lines was of primary concern in the fire fighting effort. Fire protection to structural steel in the immediaie blast area was lost, and the reduced pressure limited the supply of water to sprinkler systems pro- OO A 030669 CONFTDFNTIAi . the vblumns as the pressure was relieved, A hawirUifUs wndition could lime occurred wiilunrl this anti-vacuum feature since air might easily have been suckr.il in'o the columns. Koine of the shutoff valves were mnrrca.-dble due to the fire and niaintain- !mk a purge on all of this equipment would have been i>sl difficult, tv.'n though nitrogen purge gas was available. Mctal-coveral stecl-framcd buildings are good con struction in areas where explosions con occur. The control and twice building was cl this type. Although the exterior and first lioor elide nica suffered exten sive damage. Die instrument consoles and computer located on the M-cmii: floor escaped major damage. The exterior walls, living ol light metal construction. easily deformed and. Ihun. absorbed energy and minced dam age to the interior of ibe building as well as to the building frame. The interface cabinets, which are con structed of medium weight steel sheet, formed :i shield protecting the instrument consoles. Although exten sive instrument checkout was necessary, the main damag- to viVSi.ib .1 ,.\ku ;!v ;rv':r: !,>"s quantity of dust, dirt, and small ;>ai tide.-. blown inside. The principal damage to the computer occurred when ,,^he shuck wave caused some heads to drag on the drum. A number of hardware problems occurred after the computer was relumed to service. The explosion no doubt contributed to those problems, but sonic of them could well have been problems not yet identified and. corrected prior u> the blast, Intnnnation recovered from the computer gave a complete picture of opera tion of I) ;* ur.it ;ke time f tag explosion, and made possible me subsequent studies vviiich estab lished the cause of Die explosion. One deficiency in building construction could have caused serious personnel injury. The light lixtmos in .he building were supported by the suspended ceiling. The entire ceding id!, leaving ;.!! the light fixtures hanging from, flexible electrical conduit. Surprisingly, none of the six men in the building were injured by the falling fixtures, or by running into them when escaping from the building. A I! light fixtures are now suspended from the building frame. Many of the fire monitors in the area were too far removed from the high steel structures to play water on the upper steel where it was needed. The monitors have been relocated to provide better water coverage, although, in many cases the shutoff valve was retained at the original location for belter access in the event of a fire. Block valves, located outside the area but close enough for operation by unit personnel, are required in all process lines entering the area. All these lines had valves at other locations outside the unit and could theoretically be closed when required. In practice, the severe emergency that occurred in the rest of the plant as a result of this fire and explosion was such that persons in other areas who normally would have closed the valves were not always available or could not be reached. . Problem areas Communications both inside and outside the plant area provided very serious problem.-. Voice communi cation was almost impossible near the blast area be cause of the iiois.; generaleii by escaping gas and steam. Tin; toiepb.-nc system in the immediate area was destroyed and communication with the rest ,,f the plant was limited *i> radio. Tim iliilwuitv in commuuicating with Hie re.-st of the plant was particularly bn* 60 pnrlatil when thir remotely located sin itoff valves in many of the ruptured lines had i be closed. Communication with the outside world was almost impossible. The deluge of telephone rails from t*eople in the surrounding areas completely tied up the tele phone system for miles around, (`alts i*>th into and out of the plant could not he completed, making com munication with personnel needed to deal with the emergency extremely diJlicull. Obviously, public com munication systems have extreme limitations in emer gencies such as this one. Communication with the news media was made diffi cult by the reporter's unfamiliar:ty with technical terms. Emergency press headquarters with direct out side lines was open aL 7 : i0 p.m., 17 min. sifter the ex plosion occurred. Assistant plant managers were avail able for questioning.' Press information booklets were passed out so that all corporation personnel present were properly identified. Questions were answered as factually and accurately as possible. A formal news -release, including names and addresses of employees invoice.! ir. ti.c tire a:v! a picture of ***.. lire, wa< made available at 9:00 p.m. Four more news releases were made as additional information became available. In spite of the special efforts made to accommodate the press and provide accurate information, much that was printed was inaccurate or misleading. Information packets on each operating unit in the plain that will help bridge the gap between the technical and non technical world are being prepared, should their use ever he required again. Three person.'' known to be in th<* area at the time of the explosion couid nor be accounted for several hour? later, although a preliminary count had established that there were no serious injuries. These men escaped from the blast area by scaling the plant fence. Two were taken to a hospital by a passing motorist, while a third walked around the plant and reentered at the front gate. All these men knew the importance of re porting to the plant that they were safe, but those in the hospital were unsuccessful in their attempts to do so. Word of their safety finally reached the plant through a news reporter. Unfortunately, this informa tion was not passed to the blast site and rescue per sonnel were exposed needlessly' to additional danger as they searched for the missing men. The streets leading to the plant site were immedi ately choked with vehicles and people. Some who re membered the Texas City explosion in April. 1947 were rushing to get out of the vicinity but others, much greater in number, were rushing in to watch the fire. The traffic jam seriously impeded those plant personnel needed at the plant site. Fortunately, local jrolice ar rived at the scene in minutes and blocked oft the public highway adjacent to the blast area, greatly reducing the probability of injury to the curious if additional explosions had occurred. # H. C. Jarvis received a B.S. in Ch.E. from / ' \ 1 the University of Texas and joined Union Carbide Corporation in 19-53. He is now head of the Texas City Plant's Man ic. A ' . ` aecnient Services Department v.itli the ' responsibility for the plant Process Com. puler Systems arul the Operations fie---- scorch Croup. He has been involved vntn olefin plants lor yo years and vras tiie ^ Operating Department Head during the design, construction and st3rt-up of the Texas City 1.2 million Ib./yr. elhyienc plant. DO A 000670 CONFIDENT I Ai Butadiene post u:st prhr to ccmptellon of ecttsiriicbon. The erpiosion occurred sft third column iYom \uc itdt. Tftn remains of SmtcciieRe column after lbs rxpicsion. x tccii;:g nr*:;!.* :s-Jj***.:*. IiIt; i1" fire. Stemming * hr: fic-W 'f wafer from the broken Lavs w:,s cnmpVn-atcd by the- ^.a ,\t ' * ' m'5-*'* v.' iiJ>' i l'tiiirf !*: <t ;* !:Ut Ji8*0 ru*i\'i to ease tin* slmUm vub'os. oxpr.slng Shorn ;> uii* '.iifis-nsi! ha/nrti. S-ru.: nose *;e.ere a: ;!'.-d t*'. V: sirnvimr!-':' ud.i ?- !he riiv b. supplement the r.pi i!!; lor sy.-,',*-:..n. The : > fi.;-.: ;ii* : ' *.v-:iC .*, liTf.mrt nr.f j.-n:*r a? (>.. five (.gibing XVnP.Ui. Tho-ai were prip- fir fr -,v- ; r,;* fi'-v a::*: ;'UnO were-kepi in i'-s ; rilii the fire ii iin spite of danii^y? to ifie fir*; pr-fir-;: ire: vm. nrihrlen; 'vn- rrjt cf the ur-ra. Although the fire v.;*s under troi in three hours. '.`ms if bam itself after AT 0, fii'.-. ".as act extinguished bemuse of the unugar .cram!:-, y i-'vyiyii'onf. :::;::y -M ::n- bh-ts .iC-nr ti;y teem* rf :h bins:. tr.e h-. d:rbr.s in -the jii'-w/.s ;:>H! 0.n:h;;,, W0> V>:; !v'h cUlskH: Hu- system. :iie i.n-:-. miVO;;:; t-O-.-.-r V"'li ;>..- iioeiu is. v.r.w./c-.s :-n Soro.: no unr-r,'-;! v;.:;s.rr ;. ".' in- ; thvni>'oiv*-* ;ij 3.: ;e:*'f: ns-.d .n^aiiht'a "t run of fiir:L T;'? Siishl ;!v. M-.iry s-iij-j-is*:*: \ \ t*!-.* f:U p-OS ::! Lhrr yiOc; i;;y :]>>. fj-.io. tiiJ'fti.* *<;t- u<-ieoLfinhoy '.Tsmuniiv effects A ivTsi-h-ijliiu vvhi.-h U- wiih;;'! 750 It. of the rc::nis:;r co-eiv;;: fr-on ilrinjr metfliaud ce:;r:>;.:v:;. Tf;e iomoicc urtcu ;;!?o::r l>ievk< \V;i:e lie-; 1 "!. fiv-rv the* . Over (00 t.Kivr Ss'OO.'Ht'J w-r-v j-aiti. i o: s;:;te-S}`, inj-jriy; m-,-re UniiU'ii !> minor < uU. r.hra- The f hulei: panic reneijor. i\ nearby resh-iertis :a* h -sUhty. The Ain il. 10 57 Tc.^as ( eN)`Ii>sion *.v'iih its <r;cnt iuss of YrX and w.is suit vivid in trniiiy ntomonr-ji, 'i'cie'i.hoti.- > i;:r:;.'ia;!'.Ts '*-cre reivi'.cii *.-}* rise ;r:';;r;;rH>- ;.r-o i'.nii'n -.yee ltehitinns O;vision: for intn-.i-.iny. -,vt\c: c mu-in :i'r, r:, er . tmnrintiinrl tv lho i::.v.:r;:!n ;* : vlii'v in Cs:y. j ii*J vw:; : intiii'nf: 1 ' > on:vi'v uifh f 1 v:.tv|- --i'-yin.-hiv eri jywint. r.nd n:.::tt c:*r- !(>! *.'tv* M til !> the nivn t'v h> i:*. nn'd.in-.- efner- y r<';i.`i:N. 7i:C in.^.ir.inv.i iv.vin.'f* rta.'Ji- i" -.nrit'' t'lniiu-' ;;;i-.1 err*':`i.>> erf irniyj.rnthoji i`n;?nr.`- ji,.ry. . , *, . ('sn'.mtiiiity 'rcinttor.s wnnld have been mev-ii in- 1-rover! if: 1 A:; -: niv: y V - iT=?'i ty L !'a r- rc-rhlifr*. Public Iteioti^ns Divisir v. and infursnee sd- jv-ier?. had Icon set ap inrtne.iiaiely in the .laisiags'd av! :i ih.at be*:: a.-nnya v-cv-r r;:s!:y acveasih'e to ti;e resident,*. j. Tire ;-'.;h!h*. rel-. t'aiii* nter: had havn e-itiK-rirv-r! to make cnter^eia y wil-'/f hnn!; available on the snot. f>. Ah eontpUiiiitani'-! dirt (.ontnet? had b>'eti :nade '.ona::;/-, i*o::v..rat:ov, r-er#or.rvI rather than thmwph f::r iiijnraaee adinrtoi-s. Lessons learned The ::-:vcs.fv ar- atletiuate i:n:r: purge gas sup ply v.v-: jeiaou>tr---vi 'ey this fire. Since 'he rr:a::i pipe- rr.ck in'.erserti-'-!; vc:-.* r-esv the- scene of the blast, a larye number of hyilrocnriitn linns were raptnvei-. The^<: :tae* vv.-mr -a :e JS |p. >0 si te ;:nd, in e cases, wc-ri: eimneet'-t: i;.rg.5 vc-'sei" with nr. shat-af eaivas. Xi-i-.-vvo "ro'ha Avre hr.-.ach: to the area as sas the him i\as under rattroi for ;;se 1st purging cu.uip- ;r.ei:tii-"u runiu'-'Cr. s. rar;;:* g:*s wisp, obtaii'-c;: fr*-nt phmt seuraea as as tetnf urary piping by passing the bias* area rr.al.i ;*e The purge pj was introtiu'-Cii as svon r..s e-juipr;:'. ::! iv;u: iiipiid free and reduced toT te g.is jv-V'-terc'. As ti:e -:t\m!iU:y cf hydro* n.trhvsss the >-..;.rr: diminished, purge gas was in crease-! so that tven'.muiy the fire burning a! the end of open iiivs was eminguishe:!. Thee absence of'sec- u'.dary explosions duri-ig the emergency and eitanup nffe.*! testified ?n ;l >' etiVetivei-ess r.f this approach. Anti -\f1-; e eruerge::: y bluv.Uo.vn valves whit-ii c?e:v at J pr*.-?saro pr.ivide:) an impor- Liiit .-..'(fifty The bl.ivede-vvr, valves on tr-e r.mjer dis! iilatioi: d-.hmr-u v.-ern oprit.-ti i'nit'.eiiiately after the to reue.:' r-ress1!:'!; on e-iuituiient m ar ifie fir*"-, ttii.i-c nmch r.f :,v-s e-pilpiiii-m e..!\!.;iti:t-i! iieiii hy- iiivrt-iirln*n li.piids undor prrssi: e. tin: torm'et-uurc <-l DO A 030671 CONFIDENTIAL 59