Volume 17, Issue 1 (2020)                   ioh 2020, 17(1): 697-711 | Back to browse issues page

XML Persian Abstract Print


Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

AHMADI O, Sarvestani K, Mortazavi S B, Asilian Mahabadi H. Prediction of time to Boilover in crude oil storage tanks using empirical models. ioh 2020; 17 (1) :697-711
URL: http://ioh.iums.ac.ir/article-1-2750-en.html
Department of Occupational Health and Safety Engineering, Faculty of Medical Science, Tarbiat Modares University, Tehran, Iran , mortazav@modares.ac.ir
Abstract:   (2072 Views)
Background and objectives: Storage tanks are used in oil, refining and petrochemical companies to store large volumes of hazardous materials. Tank fires are one of the most common accidents in these industries that cause major losses. The Boilover is one of the most dangerous phenomenon of crude oil storage tanks fires (atmospheric tanks). Many fires in liquid hydrocarbon tanks can lead to Boilover due to various factors such as type of stored fuel. When heavy liquid hydrocarbons such as crude oil, or in other words liquid containing a mixture of hydrocarbons with different boiling temperature ranges, there is potential of Boilover phenomena occurrence. In Boilover phenomenon a distillation process occurs at the fuel surface. The lighter compounds evaporate on the fuel surface and feed the flames, the dense layer of the fuel called hot layer go below the burning liquid surface. If the velocity of the hot layer is greater than the fuel surface regression rate, the heat wave (hot layer) is propagated downstream of the tank. If the tank contains a layer of water or a suspended layer of water and oil emulsion inside the fuel, at a certain point, the heat wave will reach this layer. When the heat wave reaches the water, it will evaporate the water. Suddenly evaporation of water resulting in the ejection of hot fuel from the tank. This phenomenon called Boilover. The occurrence of Boilover results in an increase in flame size and areas exposed to heat radiation. The main and most dangerous consequence of Boilover is the formation of a fireball and ejection of burning materials out of the tank. The heat energy released as a result of the formation of the fireball eliminates the possibility of escape for people that responded to the fire in the area of the accident. It also exacerbated the accident by flaming the contents of adjacent tanks. For example, at the Taco Power plant accident in Venezuela, a fireball after a Boilover killed 150 people and injured 500 others. It also exacerbated the accident by flaming the contents of adjacent tanks. In the Boilover of the crude oil tank at the Czechowice-Dziedzice refinery in Poland, the burning oil was ejected out of the tank up to 250 meters away from the tank. As a result of this Boilover, 30 people were killed and 100 were injured. A massive Boilover occurred at the Amoco refinery tank in Milford Haven, ejected burning oil out of the tank and formed a huge fireball with a height of 900 meters. The Boilover caused firefighters to burn. The availability of detailed information about the time to Boilover plays an important role in the decisions of the accident commander during suppression of storage tank fire. Prediction of time to Boilover is important to alert firefighters to escape the tank fire scene and to obtain information on the amount of fuel remaining in the tank prior to the Boilover. As well as predicting the time to Boilover is an important factor in modeling Boilover's consequences. Empirical models are proposed for the prediction of time to Boilover. These methods are based on experimental study and each of them uses different variables to predict time to Boilover such as radiation received from the flame, the fuel thickness, the average fuel boiling point, the fuel storage temperature, the tank diameter, and the presence of water emulsion inside the tank. The purpose of this study is to predict the time to Boilover in the fire of the crude oil storage tanks using empirical models.
Materials and Methods: In the first step, time to Boilover prediction empirical models were identified from previous studies. For this purpose, appropriate keywords were searched in search engines such as Google Scholar, Scopus, PubMed, and science direct. Experimental studies on the time to Boilover were reviewed to select a study for comparing of the predictions of models. For this purpose the experimental study that has been conducted by Koseki et al. on 1.9 m diameter storage tank containing Sarukawa crude oil was selected. All model equations were transferred to Microsoft excel, version 2016 to reduce the error and speed up the calculations. Time to Boilover was calculated using six different models for the three experiments of Koseki et al. study. The percentage of prediction error of models was calculated. Due to the significant impact of tank dimensions on the time to Boilover, the predictions of the empirical models were compared with the data of the three large-scale Boilover accidents. Boilover accidents were identified by searching the databases such as FACTS, CSB, eMARS, ARIA and previous studies. Three accidents, including Czechowice-Dziedzice refinery accident, Poland, Tacoa Power Plant accident, Venezuela and Amoco refinery accident, Milford Haven, United Kingdom that have comparable data were selected for comparing predictions of time to Boilover of empirical models. The prediction results of the empirical models were compared with accidents data. And finally, given that the time to Boilover is highly dependent on the value of thermal energy reflected from the flame to the fuel surface and there is no way to calculate this parameter value, the time to Boilover for actual accidents for a range of thermal energy reflected from the flame was calculated using Buang, Casal and Michaelis models.
Results: The prediction error of the models presented by Michaelis and Buang for the first experiment was 39.91 and 47.15, for the second experiment was 4.02 and 15.51, and for the third experiment was 8.7 and 17.8, respectively. The equations presented by Tan and Casal estimate time to Boilover more than real data. The Tan model prediction error was more than 100 % and Casal model prediction error was 81%. The Kong and Cia models estimate the time to Boilover much less than the real data. The Cia model has a prediction error of 75% and the Kong model of 77%. The error percentages of the models presented by Michaelis, Buang, Cia, Tan, Kong and Casal for the Czechowice-Dziedzice refinery accident were 3.94, 15.77, 8.57, 6.11, 30.05 and 14.4, for Tacoa accident were 3. 04, 9.76, 44.33, 16.64, 4.48, and 50.72, and for Amoco refinery accident were 11.39, 5.8, 28.52, 54.3, 26.26 and 45.68, respectively. The models presented by Buang and Michaelis models with average error of 6 and 10 % for real accidents and 26 and 17 % for experimental study, more precisely predicted the time to Boilover compared to other models. The results of the Buang and Michaelis empirical models with a reflection fraction between 3.5 and 5% were the exact time of the Boilover accident at the Amoco refinery, while this value was close to 7% for the Casal model. For the Czechowice-Dziedzice refinery accident, the same values of Amoco accident were obtained. In the Tacoa power plant accident, the value of thermal energy reflected from the flame to the fuel surface between 14 and 15 % for Buang and Michaelis models and close to 18% for Casal model, results was same as the real accident data. Given the dependence of the Buang, Michaelis and Casal models on the value of thermal energy reflected from the flame to the fuel surface, applying a range values of thermal energy reflected from the flame to the fuel surface resulted in more accurate results.
Conclusions: When dealing with tank fires, the major challenge for commanders present at the scene of the accident is the prediction of time to Boilover as one of the most hazardous phenomena in tank fire scenarios. Accurate prediction of time to Boilover is critical for developing a tank management strategy for those liquid with Boilover potential. In this study, first the models presented for predicting the time to Boilover were identified by searching in the different sources. The models predictions were compared using the results of an experimental study and the data of three real accidents. The studied models had different percentages of error in predicting the time to the Boilover. Comparison of different time to Boilover prediction models using experimental and real accident data revealed that the models presented by Buang and Michaelis have a lower prediction error than the other time to Boilover prediction models. Given the dependence of the three relationships of Buang, Michaelis and Casal on the thermal energy reflected from the flame to the fuel surface value, they applied by different value of this parameter. The results show that Boilover can be more accurately predicted by applying the Buang and Michaelis method simultaneously and by applying a range of thermal energy reflected from the flame to the fuel surface values. The results of this study can be used to predict time to Boilover and use a safe tactic and strategy to control of tank fire that have Boilover potential. Applying the models to a range of thermal energy reflected from the flame to the fuel surface values ​​also provided more accurate results from time to Boilover, which could be used to develop a relation to predict more accurate time to Boilover in future studies. The results of this study can be used to predict time to Boilover and to adopt more efficiently and safely cooling and suppression strategies and tactics when handling crude oil storage tanks fire. Limitations of the present study include the lack of experimental data on large diameter tanks.
 
Full-Text [PDF 1038 kb]   (828 Downloads)    
Type of Study: Research | Subject: Safety
Received: 2019/03/7 | Accepted: 2020/06/13 | Published: 2020/09/23

Add your comments about this article : Your username or Email:
CAPTCHA

Send email to the article author


Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

© 2024 CC BY-NC 4.0 | Iran Occupational Health

Designed & Developed by : Yektaweb