Development of Quantitative Hazard Analysis Method for Inherently Safer Chemical Processes

Transcription

1 47 A publication of CHEMICAL ENGINEERING TRANSACTIONS VOL. 3, 3 Guest Editors: Edd De Rademaeker, Bruno Fabiano, Simberto Senni Buratti Copright 3, AIDIC Servii S.r.l., ISBN ; ISSN The Italian Association of Chemical Engineering Online at: Development of Quantitative Haard Analsis Method for Inherentl Safer Chemical Processes Yuto Miuta*, Masaki Nakagawa Safet Engineering and Environmental Integrit Lab. Mitsubishi Chemical Group Science & Technolog Research Center, Kamoshida-cho Aoba-ku, Yokohama 7-85 JAPAN 79593@cc.m-kagaku.co.jp This paper shows a method to anale haard in chemical plants and to practice inherent safet measures for decreasing consequences. Haard analsis method of this work is reconsidered about worst case scenario widel has being used in the world. Calculation results are plotted on a graph which takes fatalities and lethalit distance on each axis, and haard potentials of chemical plants are ranked. Furthermore, this work also shows effective inherent safet measures practices about process equipment that have high haardous potentials.. Introduction Chemical plants have man kinds of chemical haards, and analsis methods based on worst case scenario have been developed b analsis method of API (), U.S. EPA (999) and so on. There are man studies about consequence based haard analsis and implementations, e.g. A. Tugnoli (9) and J. Haddad (). The are good wa to identif haard potential in chemical plants fluentl b using like datasets in table, but those calculation methods are assumed that materials in process equipments are released in normal operation. If operation conditions deviate from normal operation because of an troubles, accidents affecting more significant damage than worst case scenario ma occur. For example, there is a reactor which is controlled temperature as normal condition, and if materials in a reactor are released to the outside, consequence would not be so high. But if heat sstem of a reactor has accident, a reactor would be exploded b reaction runawa and give significant damage. In this case, if release of materials in normal operation is onl assumed as worst case scenario, significant accident ma be missed and haard potential of reactor ma be underestimated.. Definition of worst case scenario In this work, author redefined worst case scenario, because consequence calculated as worst case scenario should be most severe. Author defined abnormal scenario as scenarios generated b deviated from normal operation, and redefined that worst case scenario is constructed with release scenario and abnormal scenario in this work. Table : Plant information for analsis Information of process conditions Information of process equipments Information of materials Other information of plant Temperature ( ) Volume of equipment (m 3 ) Formula of material Flow sheet Pressure (kpa) Internal length (m) Molecule weight (g) Material balance sheet Composition (wt%) Hold up volume (m 3 ) Boiling point ( ) Equipment list Flow rate (kg/hr) design temperature ( ) Flash point ( ) Information of dike Phase of material design pressure (kpa)

2 48 3. Procedure of release scenario Release scenario is defined as worst case scenario widel used as quantitative haard analsis in chemical plant in this work. Author reconsidered that release scenario reported b M. Nakagawa () is able to anale more accuratel and to get same results even if anone calculate. 3. Selection of equipment The calculation method is presented b using ammonia model plant as example shown Figure. At first, analsis object equipments are selected in PFD or PID. For example, pipeline or pumps have small amount of material compared to tanks, condensers or vessels. It is able to assume that material in pipeline or pumps are included in previous and following equipments, and the are omitted from analsis object. However, in the case that pipeline and pumps have a large amount of material, the should be selected as analsis object equipments. In ammonia model plant, analsis object equipments are circled ones. 3. Definition of release scenarios and events Release scenarios from analsis object equipments and occurrence events are shown as Figure. Release scenarios are defined b correlation with process temperature, boiling point, flash point of material and atmospheric temperature. Occurrence events are defined b material phase, flammabilit and toxicit. Release amounts are defined each event and are shown in Figure 3. Desulfuriation 4 kpa 39 Natural gas (CH 4, C 3 H 4, CO, N ) 9 kpa 8 CO Absorption st reformation 8 CO Flasher nd reformation kpa 3 Methanation (CO, CO +H ) High temp. CO conversion Low temp. CO conversion 3 35 kpa 9 4 kpa kpa kpa Vapor 8 4 Drain WT Air 5 H, CO, N kpa Dewater 4 kpa 5 6 Drain WT 7 3 kpa 53 8 Ammonia snthesis reactor Ammonia tank kpa Dewater 4 kpa To boiler 9 To urea plant Figure : Process flow diagram Tb < Tp Gas release Tp > Ti ()VCE, ()Fire Ball, (5)- TGD ()VCE, ()Fire Ball, (3)Flash Fire, (5)- TGD Tb < Ta Liquid and flashing release ()VEC, ()Fire Ball, (3)Flash Fire, (5)- TGD Tf < Tp or Tf < Ta (5)- TGD Liquid release and evaporation Ta : Ambient Temperature Tb : Atmospheric Boiling point Ti :Ignition Point VCE : Vapor Cloud Explosion (4)Pool Fire, (5)- TGD Tp : Process Temperature Tf : Flash Point TGD : Toxic Gas Dispersion Figure : Release scenario and event tpe

3 49 () Vapor Cloud Explosion () Fire Ball (3) Flash Fire (4) Pool Fire (5)- Toxic Gas Dispersion (5)- Toxic Gas Dispersion = the quantit of substance present in the vessel + the feed of substance from the upstream vessel in 3 min. = the quantit of substance present in the vessel + the feed of substance from the upstream vessel in 3 min. = ( the quantit of substance present in the vessel / min. + the feed of substance from the upstream vessel ) * min. = the quantit of substance present in the vessel + the feed of substance from the upstream vessel in min. = ( the quantit of substance present in the vessel / min. + the feed of substance from the upstream vessel ) * min. = ( the quantit of substance present in the vessel / min. + the feed of substance from the upstream vessel ) * min. Figure 3: Release amounts defined b each release event 3.3 Calculation of consequence model Consequences caused b each event are calculated b TRACE software b SAFER Inc. Fatalit numbers are calculated b probit function with reference to TNO (999). Author reconsidered about calculation method of flash fire, because representative calculation methods are that all people will die in flammable gas dispersion area or people are received maximum radiation while flame burning. These methods are possible to overestimate consequence. Therefore, author calculates flash fire as below. Flammable gas dispersion phenomena are calculated b TRACE at first, and change with time of radiation received b an point and geometric factor between flame and arbitrar point during flame moving are calculated b gproms, PSE Inc.(Figure 4 and 5). In Figure 5, radiation intensit is moving with flame front, and fatalities are graduall increasing with burning time. Calculation of geometric factor at an point Y Z x Flame propagation ( -X direction) b: flame width Flame Acceptor (on the ground: =) X: downwind direction a: flame height F, = π + π + π + π a ( a ) ( a ) ( a ) b tan tan tan tan a b b a ( b ) ( b ) ( b ) Figure 4: Calculation of geometric factor

4 5 Time at 5 s after ignition = (population densit*probabilit) da Radiation (W/m ) Move with flame front Downwind distance (m) Time (s) Figure 5: Result examples of radiation intensit and time dependence of fatalities 4. Procedure of abnormal scenario Abnormal scenario is considered that process is deviated from normal operation. Abnormal scenario has scenarios analed b what-if analsis, HAZOP or HAZchart analsis which is developed b M. Nakagawa (4) in Mitsubishi Chemical Corp. An example of abnormal scenario around ammonia tank in ammonia production process plant is shown Figure 6. Abnormal scenario is selected from plotted in high consequence level in risk matrix (Figure 7) which likel cause larger fatalit numbers than release scenario. For example, it assumes that a control valve which control feed rate of cooling medium toward a heat exchanger is failed and feed of cooling medium is stopped. If feed of cooling medium is stopped, internal temperature of ammonia tank graduall rises and at the same time internal pressure rises involved with ammonia vapor pressure. Then, alarms of temperature and pressure are sounded and operators tr to do recover work. But if it is impossible to recover, pressure in ammonia tank rises continuousl. Safet valve is installed on ammonia tank so that pressure is released outside from ammonia tank at set-point below tank design pressure. However, if safet valve is not activated, tank will be ruptured and ammonia is released at higher temperature than normal operation. According to scenario analsis and probabilit calculation performed b HAZchart, probabilit of tank rupture is not high. Because there is enough time to do recover work during pressurerise up to design pressure and safet valve is installed on ammonia tank. But if ammonia tank rupture occur, ammonia is released at higher temperature from tank and it causes flash of larger amount of ammonia, and large amount of ammonia gas will give more sever consequence. In the aspect of analsis of haard potential assumed worst case scenario, it needs to calculate such scenario. Calculation results are described in Section 5. TC High (accident) PC Ammonia tank SV Cooling exchanger TI TCV Consequence Level Medium (severe trouble) Low 3 (trouble) Extremel 4 Low (small trouble) A B C D E Incredible Extremel Low Medium High Low Probabilit Figure 6: Abnormal scenario around ammonia tank Figure 7: Risk Matrix calculated b HAZchart

5 5 5. The presentation of results; release scenario and abnormal scenario Results analed b release scenario and abnormal scenario are plotted on a graph such as Figure 8. Graph takes fatalit number caused b unlikable events such as, fireball, etc on the vertical axis. Graph takes lethalit distance which author defined as unlikable events affect people % death probabilit on the horiontal axis. Furthermore, plant borderline is taken on the horiontal axis, and the graph is divided 3 color parts and ranked b severit of results. If process equipments plotted in red area that unlikable events affect over plant borderline and cause more than fatalit, safet practice should be considered immediatel. In this work, ammonia tank and ammonia snthesis reactor have high haard potential in ammonia plant. Practicing safet measures are described in Section 6. About flash fire in Section 3, representative calculation method gives 9 fatalities, and this work gives 5 fatalities. This work calculates in detail about integration of radiation on arbitrar point that representative method overestimates, and it seems that fatalit number is smaller than result of representative method. Consequence of abnormal scenario in Section 4 got bigger result than release scenario. The reason is that ammonia release temperature and flash gas amount of abnormal scenario are higher. In this wa, abnormal scenario has possibilit that have larger consequence. Furthermore, it is possible to image how large accident would occur b comparing to past accident b plotting on the same graph. Table shows past accident examples. Calculated fatalities and % lethalit distance are calculated b TRACE. Past accident examples are plotted on a graph in Figure 9 added results of ammonia plant. Calculation result of of ammonia released from 5 t tank implies higher haard than Bhopal accident. 6. Inherentl safet practices Though ammonia tank and ammonia snthesis reactor have severe haard potential, it is possible to decrease haard potential b practicing inherentl safet measures. As inherentl safet measures there are four measures, moderate, minimie, substitute and simplif. In case of example of ammonia tank and ammonia snthesis reactor, moderate and minimie are possible to adapt and are described below. 6. Moderate In the case of practicing safet measure moderate to ammonia tank and ammonia snthesis reactor, dilution and cooling are considered effective. In the case of dilution, diluting ammonia liquid with water and changing ammonia liquid to ammonia aqueous are possible to decrease the amount of ammonia flashed when ammonia release from tank and also to decrease vaporiation rate from a dike. In the case of cooling, cooling ammonia tank temperature below boiling point is possible to remove ammonia flash phenomena and to significantl decrease vaporiation rate from a dike. fireball flash fire (representative method) flash fire (this work) toxic gas dispersion.. NH3 tank abnormal scenario fireball flash fire(representative method) flash fire(this work) NH3 tank release scenario NH3 snthesis reactor release scenario Plant borderline. Lethalit distance (m) Lethalit distance (m).. fire ball flash fire(representative method) flash fire(this work) past accident Senegal experiments Flixborough Bophal Mexico Plant borderline. Lethalit distance (m) Lethalit distance (m) Figure 8: Consequence of ammonia plant Figure 9: Comparison to past accident examples Table : Plant information for analsis Places accident occur substances events amount fatalities(actual) fatalities(cal.) lethalit distance t person person m Flixborough cclohexane vapor cloud Bhopal MIC toxic gas dispersion 57 6, 3,43 6,3 Mexico LPG fireball ,476 Senegal ammonia toxic gas dispersion 9 63,5

6 5 fire ball flash fire (this work) toxic gas dispersion +NH3 tank toxic (practiced) *NH3 reactor toxic (practiced) NH3 tank toxic (practiced) NH3 tank fire ball (practiced). Divide to tanks fire ball Chang to pipeline ammonia tank (after practice) (5m length) Change to pipeline (m length) ammonia snthesis reactor (after practice) Dilute to 5% NH3aq Divide to a reactor Cool tank Change to 3 Plant borderline loop reactors. Lethalit Lethalit distance (m).. Divide to tanks Change to pipeline (5m length) fire ball flash fire(this work) ammonia tank(after practice) ammonia fire ball(after practice) Change to pipeline (m length) Plant borderline. Lethalit Lethalit distance distance (m) (m) Figure : Inherentl safer measurement - Figure : Inherentl safer measurement - 6. Minimie In the case of practicing safet measure minimie to ammonia tank and ammonia snthesis reactor, dividing a tank and changing from tank to pipeline are considered effective. Changing to pipeline means that plant does not have ammonia tank and feed product ammonia directl to other process plant or customers. To decrease hold up of ammonia is possible to decrease haard potential and consequence. 6.3 Result of inherentl safet practices Results practiced safet measures to of ammonia tank and ammonia snthesis reactor in release scenario are shown Figure. Results practiced safet measures to of ammonia tank in abnormal scenario and radiation from fireball of ammonia snthesis reactor in release scenario are shown Figure. The show that inherentl safet practices can decrease haard potential and consequence of each accident quantitativel. 7. Conclusion Haard calculation method of this work is able to anale which equipment has how large haard potential in Figure.8, and to express how large accidents would occur b comparing to past accident examples in Figure.9. If analed equipment has high haard potential, it is necessar to consider which safet measure should be done and if it generates other haardous factors. In Figure. and, each plot practiced safet measures looks like effective deceptivel, but it is hard to sa all of safet measures are effective. For example, in case of decreasing ammonia tank temperature b moderate safet measure, it is effective to decrease haard potential in release scenario. On the other hand, in abnormal scenario it is not effective. Because in abnormal scenario ammonia cooling sstem failure is assumed and ammonia tank will be higher temperature, and ammonia will be released at higher temperature. That is, in normal operating condition haard potential is under control, but if process goes abnormal condition haard potential is revealed. Therefore, it is better wa to change tank to pipeline in the aspect of inherentl safet practice. Or the possibilit that an abnormal scenario will occur have to be kept low b further practicing safet measure, though haard potential is not decreased. B using this analsis method, it would be able to practice more effective inherentl safet measures for high haardous chemical plants. References API,, Risk based inspection base resource document st edition, API Publication 58, America A. Tugnoli, G. Landucci, V. Coani, 9, Quantitative inherent safet assessment b Ke performance indicators (KPIs), Chemical Engineering Transactions, 7, J. Haddad, S.A. Salah, M. Mosa,, Major haard risk assessment on ammonia storage at Jordan Phosphate Mines, Compan in Aqaba, Chemical Engineering Transactions, 9, M. Nakagawa, Y. Iiduka, 4, Introduction and application of quantitative process haard analsis- HAZChart, th International Smposium on Loss Prevention, Prague, Cech Republic, pp M. Nakagawa, Y. Miuta,, Development of a worst case scenario, 3 th International Smposium on Loss Prevention, Brugge, Belgium, pp TNO, 999, Guidelines for quantitative risk assessment, CPR 8E purple book st edition, the Netherlands U.S. EPA, 999, Risk management program guidance for offsite consequence analsis, America