Disaster Simulation in Chemical Plants Considering Diffusion of Gas and Heat Radiation from Tank Fire

Transcription

1 Disaster Simulation in Chemical Plants Considering Diusion o Gas and Heat Radiation rom Tank Fire Tetsusei KURASHIKI 1 and Masaru ZAKO 2 1 Associate Proessor, Dept. o Management o Industry and Technology, Osaka University, Japan ABSTRACT : 2 Proessor, Graduate School o Eng., Osaka University, Japan kurasiki@mit.eng.osaka-u.ac.jp We have developed a computer simulation program in order to make clear eects o storage s outlow, evaporation, diusion o gas, tank ire and extinguishments in chemical plants. The trigger o disaster is the outlow o the storage subjects in a tank. The occurrence o the outlow is classiied into three modes: little, middle and large leakage. In case o middle leakage mode, we have developed the simulation program or the concentration o gas considering combination o the lash evaporation and the evaporation by temperature o ground. When the tank ire occurs in a plant, the heat radiation rom lame to aected tanks is calculated with Monte Carlo method in order to consider the climate, the location and the shape o tanks. The spread o tank ire is judged by the increase o temperature o the heat aected tank. As the prevention o the disaster progress, the eect o extinguishment with sprinklers is also taken into account. Furthermore, GIS (Geographical Inormation Systems) is very eective, because it is very easy to get the situation immediately o the occurrence o disaster or a plant. The disaster simulation is connected with GIS, and the dangerous zone caused by tank ire can be estimated. When the tank ire occurs in an actual plant model in Kawasaki area in Japan, the dangerous heat aected zone or a human body and the increase o inner temperature o heat aected tanks can be indicated on digital map by using the proposed system. KEYWORDS: Simulation, Chemical plants, Diusion o Gas, Heat Radiation, GIS 1. INTRODUCTION I a big earthquake attacks a chemical plant and tanks are broken, the disaster will extend due to heat radiation rom ire and explosion pressure. Recently, tank ires and explosion phenomena oten occurred at chemical plants in Japan. So, the estimation o the dangerous area caused by disaster is very important or the saety and reliability o chemical plants. However, it is very diicult to estimate the dangerous area caused by the disaster due to complex phenomena, such as ignition, heat radiation rom ire, explosion pressure and so on. I the aected area caused by disaster can be estimated with numerical simulation, it will become a very useul tool or saety and reliability o chemical plants. In the previous study, we had developed the computer simulation program in order to estimate the spread o ire, and proposed a probabilistic estimation approach o the risk based on the developed simulation [1]. In this study, we have developed the computer simulation program in order to make clear the eects o storage s outlow, evaporation, diusion o gas, tank ire and extinguishments. Heat radiation rom lame to aected tanks is calculated with Monte Carlo method in order to consider the climate, the location and the shape o tanks. The spread o tank ire is judged by the increase o the temperature and pressure o the heat aected tank. As the prevention o the disaster progress, the eect o extinguishment with sprinklers is also considered. Furthermore, Geographical Inormation Systems (GIS) is very eective to get the situation immediately o the occurrence o disaster or a plant. The disaster simulation system connects to GIS, and the dangerous zone caused by ire and gas diusion can be indicated in actual geographical area. The detail is described in this paper. 2. CLASSIFICATION OF DISASTER IN CHEMICAL PLANTS The trigger o a tank ire or an explosion is the outlow o the combustibility storage subjects in a tank. Simulating the behavior o outlow and gas diusion is useul or the saety and reliability or chemical plants.

2 The High Pressure Gas Saety Institution o Japan had established the estimation procedure o the gas diusion in a plant based on the guide o KHK-E-007 [2]. In the guide, the estimation methods or gas diusion considering the amount o leakages have been presented. As the case o most severe incident, when the tank are destroyed by the earthquake and all amount o contents in the tank is lowed inside the oil dike, the contents is evaporated by the heat exchange on the surace o the ground instantaneously. And the concentration o gas is calculated as the instantaneous two dimensional area sources. On the other hand, i the amount o the leakage is very small such as an outlow rom a micro crack in a pipe, a phenomenon o lash evaporation, which means the radical vaporization caused by the dierence o pressure and does not occur the heat exchange on the surace o the ground, may appear. And the concentration o gas is calculated as the continuous source. However, a phenomenon o middle amount o leakage, which induces both evaporations, occurs requently. In the KHK guide, the combination eects o both evaporations have not been taken into account. The occurrence o outlow is classiied into three modes: little, middle and large leakage. Figure 1 shows the classiication o the disaster scenario. In case o middle leakage mode, we propose the calculational method o the gas concentration by the combination o lash evaporation and evaporation by temperature o ground. Furthermore, the heat eects when the tank ire occurs are also estimated with Monte Carlo method. Very small amount o leakage Middle amount o leakage Large amount o leakage 3. ESTIMATION OF DANGEROUS AREA CAUSED BY TANK FIRE 3.1 Estimation Method o Heat Radiation by KHK and TNO Flash evaporation Flash evaporation Evaporation on ground Evaporation on ground Gas diusion (continuous point source) Simulation or gas diusion Gas diusion (Instantaneous 2D source) Simulation or tank ire Simulation or tank ire Simulation or tank ire Conventional procedure by KHK Proposed procedure Figure 1 Classiication o disaster scenario A pool ire is one o the conlagration phenomena with a great deal o lames and heat. The shape o the lame can not be determined clearly, however, it is reasonable to treat the shape as cylindrical model. The heat radiation energy rom a cylindrical ire can be evaluated as ollows; 4 E = φ ε σ T = φ R D (3.1) where E is heat radiation energy, ε is emissivity o a lame, σ is Stean Boltzmann s constant, T is the lame temperature, and φ is the coniguration actor indicating heat transer rate rom a lame to a heat aected surace. Based on the equation, the heat radiation can be evaluated by using two parameters φ and R D. Parameter R D means the heat radiant emittance and the value is constant dependent on the kind o uel. The key point is how to estimate the coniguration actor φ. In the KHK guide [2], the estimation procedure o the dangerous zone caused by heat radiation rom tank ire is described. The shape o a tank ire is treated as a cylindrical model, and the estimation o coniguration actor φ is adopted by the Hamilton s equation [2]. However, there are some problem in the guideline, ie., the heat radiation considering the inclination o lame due to the wind had not been estimated because o the diiculty or solving the integral equation o heat radiation. In TNO (The Netherlands Organization o Applied Scientiic Research), the Yellow Book report is published [3]. The report provides the methods or the calculation o physical eects due to the release o hazardous liquids and gases, and the estimation procedure o heat radiation is also described. In the Yellow Book method, an eect o wind on heat radiation is considered, and the method is generally applied to estimate the heat radiation rom tank ire. However, the interruption o heat radiation by oil dikes or other tanks and equipments in plants had not been considered.

3 3.2 Estimation o Heat Radiation based on Monte Carlo Method In order to solve the above diiculty, the coniguration actor φ considering the eect o wind is calculated with Monte Carlo method. The calculational procedure or heat transer rom ire to the aected surace with Monte Carlo method is as ollows. Heat rom ire is radiated by the independent particles, and the radiant behavior o each particle is simulated with random number. Figure 2 shows the geometric coordinate and the starting point o a particle on the surace o lame cylinder. The shape o lame is treated as the shape as cylindrical model. Y Wind 風速 velocity 火災円筒 ξ θ v τ Z' ω Y' Direction o 放射光子の a radiant 射出方向 particle ϕ X' X 受熱面 Heat aected surace Z Figure 2 Geometrical coordinate and a radiant particle. The starting point (x i, y i, z i ) o a particle is given as ollows; x = x i i i c y = y c c z = z + R 1 + R + R 1 1 h h h sinθ cosξ + r v cosθ sinθ sinξ + r v v cosτ sinτ (3.2) where (x c, y c, z c ) means the coordinate o the center o lame cylinder on the ground surace. Symbol h is the height o lame cylinder, r is radius o lame cylinder, ξ is the orientation o wind direction, and τ = (π/2)(2r 2-1). R 1, R 2 are random numbers. Radiant direction o a particle is given by Lambert s cosine law. The directional cosine (α, β, γ) is calculated as ollows; α cosϕ cosτ sinϕ sinω sinτ β = sinϕ cosω γ cosϕ sinτ + sinϕ sinω cosτ (3.3) The distance rom a starting point o a radiant particle to receiving surace is calculated with the directional cosine. The achievement to the aected tank surace o the particle is judged, and the arrived particles are accumulated at each position on the receiving surace. The heat transer rate, i.e., coniguration actor, is obtained rom the number o accumulated particles. 3.3 Numerical Results o Coniguration Factor To investigate the accuracy o coniguration actor calculated with the proposed Monte Carlo method, the numerical results are compared with the conventional method o KHK and TNO. Figure 3 shows the numerical results o the coniguration actors. The horizontal axis means the calculation point which has the distance rom center o tank ire. The coniguration actor is estimated under the wind velocity 0 (m/s). In the Monte Carlo

4 method, the number o generated particles is 50 million. As the results, the proposed Monte Carlo method has good agreements with both methods. The proposed Monte Carlo method is very eective in case that the wind conditions and the interruption o heat radiation by oil dike and other equipments have to be considered. Coniguration actor Monte Carlo method KHK method (Hamilton s Eq.) TNO method (Yellow Book) Distance rom center o lame / Radius o a tank Figure 3 Comparison o coniguration actor 3.4 Simulation method or spread o tank ire The spread o tank ire in a plant is simulated with the developed system. The procedure is as ollows; the data on contents and shape o tanks are inputted at irst. The tanks aected by ire are reerred. Temperature on the surace o the heat aected tank increases by the eect o heat radiation. The calory at the surace is calculated as ollows; Q = S E t (3.4) where S is the sectional area o tank, and t is time. As prevention o disaster progress, the eect o extinguishment with sprinklers was estimated [5]. The calory at the surace o heat aected tank with extinguishment is calculated as ollows; Q = S ( E Π Ev ) t (3.5) where Π is transmission rate o radiation through water curtain o sprinkler, and E v is latent heat o vaporization as ollows. Finally, temperature on inner surace o the wall o the heat aected tank is calculated as ollows; Q Q T = i + T 0 (3.6) A δ ρ κ where A is surace area o tank. δ is average thickness, ρ is density, and κ is speciic heat o tank wall. T 0 is atmospheric temperature, and Q i is calory due to thermal conduction rom ire to contents in heat aected tank through the wall. 3.5 Numerical example o spread o tank ire Based on the Monte Carlo method on the heat radiation, the disaster simulation or chemical plants considering spread o tank ire has been developed in Figure 4. Firstly, the data on uel o tanks, shape o tanks, and initial burning tanks are inputted at irst. And, the tanks aected by initial burning tanks are reerred. Temperature on the surace o the heat aected tank can be calculated by Eq.(3.6). As the most dangerous case, when the

5 temperature exceeds the ignition point o contents, the tank explodes and the occurrence o the sequential disaster is judged. 4. Calculation o heat radiation interrupted by oil dike and tanks 2. Calculation o temperature and pressure in tanks caused by heat radiation 3. Judgment o sequential disaster Heat aected zone or human body (8141 W/m 2 ) 1.Calculation o heat radiation rom ire to tanks Figure 4 Estimation o spread o tank ire by developed simulation 4. ESTIMATION OF SPREAD OF DAMAGES BY SIMULATION WITH GIS 4.1 Geographical Inormation System (GIS) In order to get the situation o the occurrence o disaster or a plant, Geographical Inormation System (GIS) will be very eective. Thereore, the disaster simulation system is connected with GIS. Figure 5 shows the scheme o developed disaster simulation system with GIS. The procedure o calculation is as ollows; the data on tanks entries and the database on material properties or contents o tanks is prepared. By these input data, heat aected zone or a human body caused by tank ire and the damage states o tanks are calculated with the developed system. The results o simulation, which are the heat aected zone, damage states o tanks, etc., are indicated on the GIS. In this study, we have used the GIS program developed by NIED (National research Institute or Earth science and Disaster prevention, Japan). 4.2 Numerical Results o spread o tank ire by simulation with GIS As an application o the simulation on GIS, the dangerous zone caused by tank ire can be estimated in an actual geographical area as shown in Figure 6. The location o plants is a place at Kawasaki in Japan and the area consisted o our parts. Petroleum tanks are accumulated in plant 2 zone and several LPG tanks are located in a zone o Plant 4. When an initial ire occurs at No.3 tank, which the shape is a cylinder and the contents is petroleum, the heat aected zone or a human body is indicated as shown in Figure 7. Wind speed is 4.7m/s and wind direction is shown in the igure. ΔT means the increase o inner temperature o a heat aected tank. From the numerical results, the ire extends inside o plant 2 zone. In case o 30(min.), the dangerous area or human body is only appeared at the surrounding part o initial burning tank in Figure 7(a). On the other hand, the sequential ire appeared at the neighbor tank (contents: petroleum), and the tank ire spreads because the heat radiation rom ire accelerates due to the inclination o the shape o ire by wind in Figure 7(b). Figure 8 shows the numerical result considering an eect o extinguishment with sprinklers. Suppose the total amount o water distributed to tanks is 20,000(l/min.), the occurrence o sequential burning tank can be prevented. By using the developed system, we can estimate an optimum amount o water.

6 th The 14 World Conerence on Earthquake Engineering Database or simulation Input data or simulation material properties or contents (ignition temperature, speciic gravity, coordinate, speciic heat, etc.) heat transer rate obtained by M.C. (considering an eect o wind and location o tanks) vector data or tanks (coordinate, shape, size, kind o contents, etc.) plant1 Disaster simulation or Chemical Plants considering Spread o Tank(dscp_.exe) Fire plant 4 No.3 tank (Petroleum) plant 2 No.25 tank (LPG) Output data Results o simulation (heat aected zone or a human body, damage states o tanks, etc.) plant 3 GIS DiMSIS: Disaster Management Spatial Inormation System Figure 5 Scheme o disaster simulation connected with GIS Figure 6 A plant model on GIS 4.7m/s 4.7m/s (Petroleum) Heat aected zone or a human body ΔT Extinction % % % % 0 20 % Extinction % % % % 0 20 % ΔT (a) Ater 30(min.) (b) Ater 180(min.) Figure 7 Numerical results when an initial ire appeared at a petroleum tank 4.7m/s Initial ire (LPG) ΔT Extinction % % % % 0 20 % Figure 8 Numerical results considering an eect o extinguishments ΔT Extinction % % % % 0 20 % Figure 9 Numerical results when an initial ire appeared at a LPG tank ater 60(min.)

7 As other case o the simulation, the results o dangerous zone when an initial ire occurred at No.25 LPG tank in plant 4 area are shown in Figure 9. The heat aected zone is propagated in plant 4 area. From these results, it is recognized that the sequence o disaster is quite dierent due to the contents o tanks and the location o initial disaster, and the dierence can be estimated by the proposed simulation. 5. ESTIMATION OF DIFFSUION GAS CAUSED BY LEAKAGE 5.1 Calculation o gas diusion in the middle amount o leakage In the case o middle amount o leakage, the storage s outlow, lash evaporation, and evaporation by the environment may occur as shown in Figure 10. The combination eect o both evaporations can be evaluated as ollows [4]. Diusion o gas Leakage Outlow into dike - Flash evaporation - Evaporation on ground surace Figure 10 Outlow o storage subjects, an evaporation and a diusion dl dq dv = ( 1 ) (5.1) dt dt dt where L is the amount o leakage, is coeicient o lash evaporation, Q is the amount o outlow rom a tank, and V is the amount o evaporation by the ground. The emitting amount o gas q i is calculated as ollows. q i dq dv = + S (5.2) dt dt where S is the area o the leakage or the inside area o the dike. The concentration o gas is calculated as ollows. The equation is obtained by the solution o the dierential equation o Fick s low in case o instantaneous two dimensional sources [2]. qi exp( Z / B) C( X, Y, Z, t) = 4U B Y + er M A / 2 X U t + M er A er / 2 Y er M A / 2 X U t M A / 2 (5.3) where C is the concentration o gas, U is the velocity o wind, t is time, X, Y, and Z are coordinates o the calculational point. We take X-axis along the wind direction, Z-axis vertically rom the ground, and Y-axis perpendicularly to them. M is a equivalent diameter o the leakage area. A, and B are coeicients dependent on the climate condition. The key point o the above equations is how to calculate the emitting amount o gas q i. In the conventional method by KHK, q i is all amounts o tank contents in order to estimate the large amount o

8 leakage. On the other hand, the change with time o q i is taken into account by solving Eqs.(5.1),(5.2), and (5.3). The procedure induces that the combination eect o both evaporations in the case o middle amount o leakage can be estimated. 5.2 Numerical example o gas diusion with GIS As the numerical example, we suppose that the shape o damaged tank is sphere, and the content is ammonia. The diameter and capacity are 22.3m, 6000kl, respectively. The tank is connected with the pipeline which height is 1.0m. The crack hole, which diameter is 100mm, is appeared on the pipe, and the diusion o gas is calculated. Figure 11 shows the concentration zone o 700 ppm at height 1.0m and 10m ater 10minutes. From the results, we can estimate the diused zone clearly by using GIS. 6. CONCLUSION Figure 11 Dangerous zone displayed by GIS The computer simulation system to make clear the eects o storage s outlow, evaporation, diusion o gas, tank ire and extinguishments had been developed and the numerical procedures were described. As numerical examples, the gas concentration in case o middle amount o leakage can be calculated by considering o the combination o lash evaporation and evaporation by temperature o ground. When the tank ire occurs in a plant, the heat radiation rom lame to aected tanks is calculated with Monte Carlo method to consider the climate condition, and the spread o ire is judged by the increase o temperature o the heat aected tank. Furthermore, when the tank ire occurs in an actual plant model in Kawasaki area, the dangerous heat aected zone or a human body and the increase o inner temperature o heat aected tanks can be indicated on GIS. From these results, it is recognized that the developed simulation system can be applied the evaluation o the saety and the reliability or plants. ACKNOWLEDGEMENTS This study was supported by the Earthquake Disaster Mitigation Research Center, National Research Institute or Earth Science and Disaster Prevention, JAPAN. We would like to express our thanks or their cooperation. REFERENCES [1] T.Kurashiki, M.Zako, and M.Fumita, A practical estimation method o saety and reliability or chemical plants, Proc. o International Conerence on Structural Saety and Reliability: CD-ROM (2005). [2] The High Pressure Gas Saety Institution o Japan, Saety and Loss Prevention or Large Scale Petrochemical and Related Industries, KHK E , Japan (1974). [3] TNO Committee or the Prevention o Disasters, Methods or the calculation o physical eects resulting rom releases o hazardous materials, CPR-14E (2005). [4] T.Kurashiki, M.Zako, and Y.Yasutake, Disaster simulation or chemical plants (V), J. o the Society o Material Science, Japan, Vol.50 (No.1): (2001).