Luận văn thạc sĩ Kỹ thuật hóa học: Simulation and evaluation of LPG dispersion into the environment in transient state via computational fluid dynamics (CFD)

VIETNAM NATIONAL UNIVERSITY HO CHI MINH CITY HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY LE NGOC MINH KHOI SIMULATION AND EVALUATION OF LPG DISPERSION INTO THE ENVIRONMENT IN TRANSIENT S

INTRODUCTION

Description of the problem

Liquefied Petroleum Gas (LPG), a clean and efficient energy source, is widely utilized as both a chemical feedstock and fuel in transportation, commercial, and residential sectors LPG has become a popular choice in civilian applications specifically, and in the industry at large [1] However, the highly flammable and explosive nature of LPG poses significant risks in terms of storage and transportation

In the event of an accident, such as a leak, LPG storage tanks can lose pressure, leading to vigorous boiling of the liquefied LPG due to the disruption of the initial gas-liquid equilibrium state inside the ruptured tank [2] When accidents involving LPG occur, such as flash fires, vapor cloud explosions (VCEs), and boiling liquid expanding vapor explosions (BLEVEs), the consequences can lead to severe damage to human health and life, while also posing environmental challenges [3] Liquefied Petroleum Gas (LPG) primarily consists of propane, butane, propylene, butylene, and isobutane This mixture of hydrocarbon gases is referred to as LPG and is commonly used as a fuel due to its highly flammable nature Propane (C3H8) and butane (C4H10) are the two main active chemical compounds in LPG When using LPG, strict attention must be paid to potential hazards These hazards include pool fires caused by unintended ignition of LPG The heat flux from a pool fire can pose greater risks to individuals and cause serious damage to humans, machinery, storage tanks, and equipment simultaneously Standards for managing minimum separation distances from LPG facilities to exposure levels and limits on thermal radiation to humans or structures have been developed and implemented to minimize these hazards [4]

Therefore, understanding the dispersion characteristics of liquefied gas will provide crucial data for risk management throughout the production, transportation, and utilization processes Additionally, it serves as quantifiable evidence with high reliability that regulatory agencies can rely on to accurately modify existing requirements Liquefied gas is heavier than air and can accumulate on the ground Typically, odorants are added to the gas to detect leaks and reduce explosion risks

LPG is non-toxic but highly flammable, so handling LPG must be done with extreme caution, and all equipment used for storage and transportation must adhere to high safety standards, be maintained, and undergo regular inspections LPG businesses are typically required to comply with government safety regulations, and corresponding enforcement efforts help prevent fire and explosion accidents [5]

Figure 1.1 Route of LPG from production to the end-consumer [5]

LPG is produced through natural gas processing or oil refining processes and transported to end-users through a complex supply chain consisting of multiple stages Figure 1.1 illustrates the typical routes of LPG from production to the final end consumers However, the use of LPG in processing industries poses significant safety risks Potential hazards associated with leaks include fire and explosion due to the highly combustible nature of LPG Additionally, environmental concerns are also a negative impact of liquefied petroleum gas when leakage risks occur, causing significant environmental damage as well as harm to humans [5]

Property damage consequences can also occur due to LPG fire and explosion accidents, affecting facilities, equipment, and infrastructure in the surrounding areas Furthermore, business operations of the plant may be temporarily disrupted, posing significant financial losses and delays in the production process However, due to its efficiency and economy, Computational Fluid Dynamics (CFD) technology has

3 rapidly developed to provide safe models for fire and explosion prevention when accidents involving LPG gas occur in recent years

Figure 1.2: Location of LPG station on site plan

In this study, the LPG station (LPG storage facility) must be planned, constructed, and installed within the manufacturing plant in compliance with current regulations, suitable for the use and storage of liquefied petroleum gas According to Vietnamese Standards TCVN 6486:2008 and TCVN 7441:2004 regarding the installation location of storage tanks, as well as requirements for ventilation and fire and explosion prevention, the LPG station of the plant must be installed outdoors, outside of buildings, ensuring to minimize the possibility of creating LPG/air mixtures above the Lower Flammability Limit (LFL) by applying ventilation measures [6, 7] However, if the tank is placed on the ground and has a fireproof wall, Decree 25/2019/ND-CP [8] regulates a safe distance of 15 m for liquefied gas tanks with capacity from 7.6 to 114 m 3 For a project involving an LPG storage tank with a capacity of approximately 33 m 3 , with a distance of only 1.2 m from the area to the factory boundary (Figure 1.2), this distance does not meet the minimum requirements for fire prevention and fighting approval, making it infeasible to construct and implement the plant with safety assurance and compliance with firefighting

4 standards Therefore, it is necessary to investigate the use of CFD modeling to assess operations, safety, and fire and explosion potential This can enable the evaluation of the entire process of handling small-scale gas leaks into the factory environment using CFD methods

In addition, a leak diameter of 3 mm (puncture) and varying dispersion wind speeds based on the Pasquill-Gifford atmospheric dispersion model are utilized to examine the behavior of LPG leaks under different circumstances The simulation problem is executed in a transient state, where results are generated over a duration of 1000 seconds This approach allows for a comprehensive analysis of the dispersion patterns and potential consequences of LPG leaks within the factory environment.

The objective of this work

The objective of this work is to utilize Computational Fluid Dynamics (CFD) to analyze the performance of natural dispersion systems in outdoor factory environments This study considers various research conditions, including emission rate, emission duration (in transient state), wind speeds, the distribution of LPG volume fraction, and the average LPG concentration within the computational domain Besides, ANSYS FLUENT simulation software is employed to conduct this research.

The scope of this work

This project involves the construction of a manufacturing plant specializing in producing blinds made from plastic, aluminum, and fabric, with a production capacity of over five million units annually During the production process, heat is required from LPG gas to shape fabric, ignite burners, etc., to operate the factory's technological processes, making the installation of LPG stations extremely important for manufacturing plants However, it is essential to ensure that these stations comply with the relevant Vietnamese standards and regulations

Figure 1.3: FU VIETNAM’s location on satellite map (Access date: 01/05/2024)

In this study, an LPG storage tank (≈ 33 m3) is located in a separate area within the FU VIETNAM CO., LTD manufacturing plant, Lot 21-10, 8B Street, Protrade International Technology Park (PITP), An Tay Commune, Ben Cat Town, Binh Duong Province, Vietnam, as detailed in Figure 1.3 The leak size remains constant with the assumption that there is a 3 mm diameter leak puncture considered in this study, subject to the influence of different wind dispersion velocities (from 1-10 m/s) based on various stability classes according to Pasquill-Gifford during daytime and nighttime conditions (from A to F) Factors such as temperature variation, pressure differentials, and humidity are all disregarded in this study Additionally, the LPG gas in this study is assumed to be an incompressible gas, and only the gas emission into the environment is considered

The actual dimensions of the LPG station are 21 m (length) × 5.9 m (width) ×

4 m (height), with a simplified model designed with a computational domain set at

25 m (length) × 20 m (width) × 20 m (height) for calculation and simulation with the assumption that there is no external ignition source within its range Figures 1.4 and 1.5 provide more detailed information about the LPG station, including all tank measurements, which are expressed in millimeters (mm) All leak results are analyzed under the same conditions as previously mentioned

Figure 1.4: 15 tons LPG supply station area

Figure 1.5: Front view of the LPG tank

Scientific and practical significance

The comprehensive investigation into the behavior of liquefied petroleum gas (LPG) leaks within the context of outdoor factory environments, utilizing Computational Fluid Dynamics (CFD) simulations, presents significant scientific and practical implications By analyzing various parameters such as wind dispersion velocities, leak size, and environmental conditions, while considering the transient state, the study provides valuable insights by predicting to the dynamic nature of LPG dispersion and the potential risks associated with storage and handling

Moreover, the consideration of real-world scenarios, such as the dimensions of the LPG station and the impact of different stability classes on wind dispersion, enhances the practical applicability of the findings The adherence to Vietnamese standards and regulations ensures the relevance of the study to industrial practices in the region Ultimately, the study contributes to the advancement of safety measures and risk management strategies in LPG facilities, thereby promoting the safety and well-being of workers and the surrounding environment

THE OVERVIEW OF THE DISPERSION SIMULATION

Dispersion models

The dispersion pattern is influenced by various factors, such as the characteristics of the emission source and atmospheric conditions These factors include parameters like the height of the release point, release velocity, wind speed, wind direction, temperature, and other factors Gaussian dispersion modeling is employed to assume a Gaussian distribution for concentration both horizontally and vertically, aiming to depict the movement process of emitted substances from an accident location into the surrounding areas through the air Specifically, after emission, the released substance will mix with the air and disperse in the form of a plume or puff depending on the specific characteristics of the emitted substance At the emission point, the substance concentration reaches its maximum level and gradually decreases as it disperses with the surrounding air at a greater distance Due to uneven mixing and dispersion of the air containing the emitted substance, the concentration decreases significantly with distance in the direction of the wind The height of the emission point significantly affects the concentration at ground level

As the emission height increases, the substance concentration at ground level decreases because the plume must disperse over a greater vertical distance [9]

Figure 2.1: Example in which heavy gas clouds dispersed [10]

Figure 2.1 illustrates the release of dense gas with initial acceleration and dilution, along with buoyancy dominance and ambient turbulence dominance The figure depicts various stages in the dispersion process of dense gas clouds, providing valuable insights into the dispersion dynamics in the atmosphere These stages include [10]:

• Initial acceleration and dilution phase

These phases represent different stages that dense gas clouds undergo as they move away from the source It's important to note that in specific emission scenarios, some of these phases may not occur, but the overall sequence of phases remains consistent Dense gas clouds undergo changes in thermodynamic and chemical properties as they mix with the surrounding air, including processes such as phase change heat exchange, air entrainment, and chemical reactions Advanced computational models simulate these thermodynamic and chemical processes separately, along with cloud mixing and transport calculations [10]

To replicate the dispersion of LPG within the space of the manufacturing plant in case of any leakage from the tank, in this study, the emission source will be simulated in a transient-state along with wind applied for the conducted research These investigations utilize airflow model with a uniformly leaking velocity, continuously emitted from the source from 0 – 1000 seconds This scenario allows for the application of a Gaussian plume model The airflow originates from a continuously stable source at a height Hr above the ground level, and the wind moves in the x-direction with a constant velocity u [9], with the concentration calculated according to the formula:

, (2.1) where C (x, y, z) is concentration (kg/s), x is downwind distance (m), y is transverse wind direction distance (m), z is height from the ground(m), Q is release rate of leakage (kg/s), u is average wind speed (m/s), Hr is effective discharge height (m), 𝜎 𝑦 , 𝜎 𝑧 are horizontal and vertical dispersion parameters, respectively

In the experiment, temperature difference is neglect so another case will be taken in to considered when the effective discharge height (Hr) is zero, which means the leakage at ground level [5] This formula is also used to validate the simulation results:

To simulate the LPG dispersion in the project of production factory as the storage rupture condition, the plume model is applied to the tests The leakage rate of the release source in this study is continuous, uniform, and operates in a steady condition Besides, the rate is calculated using the following formula [11]:

𝛾 + 1) (𝛾+1)(𝛾−1) , (2.3) where (Qm)choked is the discharge mass flow, Co is the discharge coefficient, A is the area of the discharge, P is the absolute upstream pressure, γ is the heat capacity ratio for the gas, gc is the gravitational constant, M is the molecular weight of the gas,

Rg is the ideal gas constant, and T is the absolute temperature of the discharge [9] For a fixed upstream pressure and temperature, a limiting condition known as choked flow occurs when the mass flow does not increase with a further decrease in the downstream pressure environment Choked release flow rate refers to a situation

11 where the fluid being released through an opening or orifice is traveling at the speed of sound This is also known as critical flow and occurs when the pressure drop across the opening is such that the fluid velocity reaches sonic velocity, which limits any further increase in the flow rate In this state, any decrease in the downstream pressure will not increase the mass flow rate of fluid beyond the choked rate It is an important parameter to consider in various industries, including oil and gas, chemical, and aerospace, where the accurate prediction of gas flow rates is critical [11].

Overview of the CFD simulation

Computational Fluid Dynamics (CFD) constitutes a specialized domain within fluid mechanics, dedicated to the meticulous examination of fluid flow dynamics, heat transfer phenomena, mass transport phenomena, and the associated chemical reactions Employing numerical techniques and structured data representations, CFD endeavors to solve sets of governing partial differential equations governing fluid flow These methodologies encompass finite volume methods (FVM), finite element methods (FEM), and finite differences methods (FDM), among others, tailored to address mathematical equations controlling momentum conservation, mass conservation, energy conservation, species conservation, and the influence of external forces on bodies immersed in fluid flows [12] CFD has emerged as an indispensable tool across a diverse array of industries, including but not limited to aerospace engineering, automotive engineering, naval architecture, power generation, turbomachinery design, electrical engineering, electronic engineering, and chemical process engineering By facilitating detailed analyses and simulations, CFD contributes substantially to optimizing engineering processes and designs, thus potentially reducing both time and costs compared to traditional experimental approaches and data acquisition methodologies

CFD is widely applied in research and solving technical issues across various industries, including aerodynamics in aerospace, hydrodynamics in maritime, power plants, structural engineering, electrical and electronic engineering, and chemical

12 engineering Additionally, CFD aids in saving time in process optimization and design, reducing costs compared to experimentation and data collection [12]

Introduced by Ansys Inc in 2006, ANSYS FLUENT is a computational fluid dynamics (CFD) software designed for general-purpose fluid flow, heat transfer, mass transfer, chemical reactions, etc FLUENT offers a modern, user-friendly interface that streamlines the CFD process from setting up simulations to exporting results in a single-window workflow FLUENT is renowned for its advanced physical modeling capabilities, including turbulent flow, single and multiphase flows, combustion processes, battery modeling, fluid-structure interaction, and more Additionally, FLUENT is known for its high-performance computing capabilities, enabling efficient simulation of large models on multiple CPU or GPU processors The CFD packages provide accurate solvers for various flow regimes engineers encounter regularly, from Newtonian to non-Newtonian fluids, single-phase to multiphase, and from subsonic to supersonic flows Each solver for transient simulations ensures high stability, having been rigorously tested, validated, and optimized Both accuracy and speed are achieved through efficient solvers, making FLUENT a trusted tool in both academia and industry [13]

2.2.3 Working principle of ANSYS CFD

The Finite Volume Method (FVM) utilized in ANSYS Fluent is the core principle of this solver This method is applied to solve partial differential equations numerically The domain of the problem is discretized into a finite set of control volumes Within this set, the conservation equations are solved numerically using the finite volume method, which involves the discretization of the governing equations using the divergence theorem [14]

In the Finite Volume Method (FVM), some components of the conservation equations are converted into fluxes across the control volume faces and evaluated at the finite volume faces Because the inflow into a particular control volume is exactly equal to the outflow from the adjacent volume, FVM is an extremely conservative

13 method This fundamental conservation property of FVM makes it a preferred method in Computational Fluid Dynamics (CFD) Another important characteristic of FVM is its ability to be formulated in physical space on unstructured polygonal meshes The use of unstructured grids allows for accurate representation of complex flow behaviors and flow physics, facilitating the analysis of engineering problems in the real world with diverse and irregular geometries The adaptive nature of unstructured meshes provides a robust framework for solving the multidimensional and complex nature of fluid flow and heat transfer phenomena Finally, in FVM, implementing various boundary conditions is straightforward and non-intrusive, as the unknown variables are evaluated at the centroids of the control volume elements, not at their boundaries [15] Figure 2.2 is an example of the cell in computational domain used in CFD [16]

Figure 2.2: Computational domain defined for a circular pipe [16]

2.2.4 An overview of CFD simulation

Computational Fluid Dynamics (CFD) is a computer-based tool that utilizes numerical methods and algorithms to solve the governing equations derived from the mathematics of fluid mechanics to simulate fluid flow CFD simulations are widely used across various engineering disciplines, including aerospace, automotive, chemical, civil, mechanical, and environmental engineering, among others

The basic steps involved in conducting a CFD simulation are as follows:

1 Problem Definition: The first step in performing a CFD simulation is defining the physical problem, which includes identifying the geometry of the system, boundary conditions, fluid properties, and the type of analysis required

2 Mathematical Modeling: Once the problem is defined, mathematical models are constructed by applying conservation laws such as the equations of mass, momentum, and energy, which govern the behavior of fluid flow

3 Discretization: In this step, the governing equations are discretized into a set of algebraic equations using numerical methods such as finite difference method (FDM), finite volume method (FVM), or finite element method (FEM)

4 Solver: The discretized algebraic equations are then solved numerically using iterative methods until convergence is achieved This step requires high computational performance and extensive computing resources

5 Post-Processing: After obtaining the solution, the post-processing phase is carried out to visualize and analyze the results This involves extracting data from the simulation results, creating plots, and evaluating the quality and accuracy of the simulation

CFD simulations offer several advantages over traditional experimental methods They are cost-effective, provide detailed information about flow fields, and allow for rapid design iterations However, they also have limitations such as uncertainty in the models, numerical errors, and high computational time and cost Nevertheless, CFD simulations are a powerful tool that aids engineers and scientists in understanding complex fluid flows and optimizing the designs of various engineering systems related to fluid flow, such as aircraft, automobiles, pipelines, and wind turbines, among others Figure 2.3 provides a concise overview of the CFD simulation process [17]

Figure 2.3: Procedure for CFD simulation [17]

Application of ANSYS FLUENT software in CFD

2.3.1 Application of some modules in ANSYS FLUENT software

2.3.1.1 Investigate the wind speed dispersion and NH 3 emission rate when there are obstacles

In Siddiqui and colleagues' study (2012) on ammonia leakage and dispersion, it was demonstrated that increasing wind speed could significantly enhance the dispersion rate of ammonia The utilization of ANSYS FLUENT played a pivotal role in investigating both wind velocity (Figure 2.4) and emission flow rate (Figure 2.5) [9]

Figure 2.4: The mole fraction of NH 3 [9]

Figure 2.5: The velocity magnitude (m/s) of wind field [9]

The results of computational modeling using ANSYS FLUENT have been compared with experimental data to assess their accuracy in predicting dispersion behavior The study demonstrates that both the Standard k-ε and RNG k-ε models employed in ANSYS FLUENT yield results closely aligned with experimental outcomes, thereby attesting to the capability of ANSYS FLUENT in simulating ammonia dispersion and providing accurate predictions within a specific concentration range Notably, flow velocity and wind speed are fundamental parameters significantly influencing the dispersion characteristics of ammonia It has been observed that the dispersion of ammonia gas tends to increase significantly, and the concentration of ammonia is directly proportional to the emission rate [9]

2.3.1.2 Investigate the effect of liquid LPG combustion flow in the presence of wind diffusion with different fire ignition paths [4]

The application of ANSYS Fluent in the work referenced by Yi et al (2020) specifically focuses on studying the effects of LPG in pool fires ANSYS Fluent 2019 R2 was utilized to numerically solve the governing equations under appropriate boundary conditions It also highlights that the CFD model developed within ANSYS Fluent was employed to investigate the structure of large LPG pool fires under static and windy conditions in Figure 2.6

Figure 2.6: Temperature range (K) of LPG liquid fire diameter (D = 10.4 m) with different wind speeds [4]

Specifically, Fluent is used to measure the burning speed, flame height and tilt of large LPG tank fires using different methods with experimental data The goal is to propose new correlations of flame height and tilt with greater accuracy, especially for large-scale LPG tank fires In summary, the use of ANSYS Fluent in this work allows the evaluation and comparison of different fire rate models, as well as the study of how the height and slope of the fire are affected by the parameters different numbers such as flare diameter and wind speed

2.3.1.3 Investigate the leakage of hydrogen gas over time with different velocities depending on the emission pressure

The accident scenario involves an external leak described as a malfunction of the dispenser hose during refueling, causing high-speed release of hydro gas The hydro gas is simulated to escape through the dispenser hose at a temperature of 40°C The nozzle has a diameter of 30 mm and is set to release gas at a pressure of 30 MPa, allowing rapid discharge of hydro gas into the atmosphere The continuous gas leakage lasts for 5 and 10 seconds, and the dispersion distance of the hydro gas leak with the assumption of the activated shut-off valve is shown in Figures 2.7 and 2.8 [18]

Figure 2.7: Leakage length (left) and dispersion distance 10 s (right) after leak considered at 30 MPa pressure [18]

Figure 2.8: Dissipation distance after 5 seconds (left) and 10 seconds (right) after leakage at pressures of 50 MPa and 70 MPa [18]

The article applies the software ANSYS Fluent for implementation Specifically, Fluent plays a crucial role in simulating the dispersion of hydro gas through species transport modeling The software supports the prediction of temperature, mass, and velocity of hydro gas through energy equations, contributing to the comprehensive analysis of hydro leakage and dispersion scenarios Furthermore, this software also serves as a fundamental tool in utilizing the Realizable k-epsilon (k-ε) turbulence model to simulate gas diffusion, which is a commonly used turbulent model for complex flow conditions in computational fluid dynamics ANSYS Fluent plays a pivotal role in enabling detailed simulation and analysis related to hydro safety in the study [18]

2.3.2 Current research on CFD simulation of emission accidents

2.3.2.1 Application of CFD simulation of liquefied petroleum gas (LPG) diffusion behavior in the case of an LPG tank truck accident in China

Recognizing the increasingly crucial demand for the development of liquefied petroleum gas (LPG) transportation worldwide, Lyu et al (2023) have developed models to better understand the dispersion behavior of vapor clouds resulting from the spontaneous release of LPG in urban areas to propose safety measures to prevent fire and explosion hazards The models also address challenges and concerns

20 regarding safety related to the risk of serious fire and explosion accidents, with a particular focus on road transportation The risks associated with mishandling and improper transportation of energy and chemicals, as well as the importance of providing energy, sustainable urban development, and the safety of people and the environment, are emphasized in this study [3]

The research highlights the impact of various factors such as droplets in the LPG release source, terrain features, and major obstacles on the dispersion behavior and spread distance of vapor clouds (see Figure 2.9) Propane concentrations are provided at different time points (specifically, the first 20 seconds after the leakage accident) for three different cases of propane concentration The results offer a clear insight into the dispersion behavior and concentration levels of propane after release, providing detailed understanding for analysis in various scenarios

Figure 2.9: Schematic of propane concentration with LFL concentration at different times (first 20 seconds after emission) for three cases [3]

Furthermore, Lyu and colleagues emphasize the risks associated with mishandling and improper transportation of energy and chemicals, as well as the importance of providing safe energy and chemical supply for the sustainable development of cities in general and the safety of people and the environment in

22 particular The study also provides detailed information on the dispersion process of vapor clouds, outlining the specific stages of leakage and dispersion, while describing differences in cloud dispersion characteristics based on droplet mass, terrain features, and obstacle arrangements [3].

2.3.2.2 Numerical simulation of the dispersion behavior of hydrogen gas when a leak occurs in a garage in the presence of obstructions

The widespread use of hydrogen requires appropriate mitigation or prevention technologies to minimize the flammability hazard, specifically 4 to 75% by volume can lead to high fire and explosion risks Li et al (2019) investigated the further understanding of hydrogen dispersion in garage environments, specifically focusing on the influence of crossbeam height and the effectiveness of ventilation systems in minimize fire and explosion risks

The computational fluid dynamics (CFD) model developed for the study used a first-order scheme to build the transient formulation and second-order scheme order scheme) for spatial discretization Numerical results are compared with experimental data to confirm the CFD model's accuracy, ensuring the feasibility when investigated At the same time, the study's findings demonstrated that natural ventilation effectively reduces hydrogen accumulation but cannot completely eliminate the risk of explosion Furthermore, the study emphasizes the impact of beam height on hydrogen stratification, emphasizing the Figure into gradual transition layers and the influence of vortices The ventilation system was found to be effective in minimizing hydrogen buildup by facilitating hydrogen discharge and air inflow by adding two symmetrical ventilation holes This effectively reduces the hydrogen mole fraction in the garage, thanks to the graphs showing the evolution of the hydrogen mole fraction in [19]

However, it is important that although ventilation reduces hydrogen build-up, it cannot completely eliminate the risk of hydrogen explosion, leading to some limitations in terms of fire safety

Figure 2.10: The hydrogen distribution for different crossbeam heights at t = 400s

Figure 2.10 provides insight into the spatial distribution of hydrogen mole fraction at a particular time (t = 400 s) for different crossbeam heights (Hc) on the x

= 0 plane It shows the change in hydrogen concentration in the garage space influenced by different cross beam heights The contours show the hydrogen concentration decreasing from the top of the garage to the bottom, showing the impact of cross beam height on hydrogen accumulation and dispersion Additionally, the thickness of the flammable zone and the blocking effect of the cross beam on hydrogen dispersion are depicted visually, providing valuable insights into the distribution behavior of hydrogen in confined spaces

2.3.2.3 Survey of dangers caused by leakage and diffusion in liquefied natural gas (LNG) of ships during operations at sea

Recognizing the increasing demand for marine transportation using liquefied natural gas (LNG) due to global warming and carbon reduction targets by 2050, there is a growing need for reliable emergency response measures for LNG carrier incidents at sea Emergency ship-to-ship (STS) transfers in the event of LNG leakage and dispersion in coastal waters pose significant challenges for safety management for vessels and coastal shipping routes The emergency transfer of cargo from one ship to another will be investigated by Bellegoni and colleagues (2021) using computational fluid dynamics (CFD) modeling of LNG spill dispersion during STS operations Subsequently, in accident scenarios with varying wind speeds and directions, the characteristics of LNG dispersion and the extent of the vapor cloud in the working transition area will be examined and simulated with the model shown in Figure 2.11 [20]

Figure 2.11: Computational model of LNG leakage and diffusion [20]

Figure 2.12: LNG diffusion process when leakage occurs during STS process [20]

MATHEMATICAL MODELING

Conservation equations

The mathematical terms are presented by the governing equations of fluid flow based on these following conservation law of physics:

The mass balance of the fluid elements is presented by [14]:

Rate of increase of mass in fluid element = Net rate of flow of mass into fluid element (1)

Where rate of increase of mass in fluid element is

𝜕𝑡 𝛿𝑥𝛿𝑦𝛿𝑧 (3.1) and the second term of the equation is:

The equation (1) is rearranged by divided by element volume 𝛿𝑥𝛿𝑦𝛿𝑧, then shown in form of:

𝜕𝑡 + 𝑑𝑖𝑣(𝜌𝑢) = 0 (3.4) The 3D continuity equation is used in a compressible fluid, and the equation is nonlinear The equation will change if the fluid is incompressible (i.e., a liquid) and density 𝜌 is constant:

The Newton's second law, on which the momentum equation is based, states that [14]:

Rate of fluid particle momentum increase = Total force acting on fluid particle (2)

All dimensions of the coordinate per unit volume of a fluid particle's left-side source term are provided by: 𝜌 𝐷𝑢

𝐷𝑡 The forces consist of surface forces and body forces Surface forces and body forces are included in the force Centrifugal, Coriolis, and electromagnetic forces make up body forces while pressure, viscous, and gravity forces comprise surface forces The equations below for the x, y, and z (momentum per unit volume) components, respectively, serve as illustrations of equation (2):

( ) xx xy zx xy yy zy xz yz zz x y z x y z x y z

- S Mx , S My , S Mz : body forces

- p is pressure or normal stress and τ is the viscous stress

The energy conservation equation is followed by the first law of the thermodynamic, which is displayed by [14]:

Rate of increase of energy of fluid particle = Net rate of heat added to fluid particle

+ Net rate of work done on fluid particle (3)

- The rate of heat addition to the fluid particle due to heat conduction:

- The total rate of work done on the fluid particle by surface stresses:

- The total energy equation, with SE is the source of energy

- For the total enthalpy equation:

( ) div( u) p div u + div( grad ) xx yx zx xy yy zy xz yz zz h h u u u h k T t x y z v v v w w w x y z x y z S

2(𝑢 2 + 𝑣 2 + 𝑤 2 ); h and h 0 are respectively specific enthalpy and specific total enthalpy

Viscous Model

ANSYS FLUENT provides various turbulence models for noise prediction, such as the Reynolds-averaged Navier-Stokes (RANS) models including the laminar flow, Spalart-Allmaras, k-epsilon (k-ε), k-omega (k-ω), Large Eddy Simulation (LES), and others These turbulence models aim to determine flow parameters including turbulent flow characteristics and noise generation The k-ε turbulence model includes four types: standard k-ε, RNG k-ε, realizable k-ε, and SST k-ε These models share similar equation structures, including transport equations for turbulent kinetic energy (k) and turbulence dissipation rate (ε) However, there are three significant differences among these turbulence models

Firstly, they differ in the method used to compute the turbulent viscosity of the fluid Secondly, there are variations in the Prandtl constant within the turbulence models, which are adjusted based on the turbulence of k and ε Lastly, there are some differences in the terms generated and eliminated in the ε equation among the models [17]

The k-ε turbulence models serve different purposes, each tailored to specific flow characteristics:

- The Standard k-ε model is commonly applied to various turbulent flows but has limitations in capturing anisotropic turbulence

- The RNG k-ε model is capable of predicting complex flows with separation and recirculation

- The Realizable k-ε model aims to improve accuracy by adjusting constants, particularly beneficial for flows with strong curvature or swirling motion

- The SST k-ε model combines aspects of both the k-ε and k-ω models to provide accurate predictions for diverse flows, transitioning based on flow conditions to enhance predictions

The choice of turbulence model depends on specific characteristics of the simulated flow, such as turbulence intensity, flow separation, and boundary layer properties [17]

In this study, the Realizable k-ε turbulence model was chosen because the model of turbulence closure is widely used to provide accurate and efficient predictions of turbulent flows It is particularly suitable for dispersion modeling as it accounts for the anisotropy of turbulence in a way that other models cannot, making it beneficial for complex flows with strong curvature or swirling motion In dispersion modeling, the Realizable k-ε model can be used to simulate turbulent flows with large gradients in concentration or temperature, accompanied by fluid streams with a density higher than air The model considers the mixing caused by turbulence occurring in these streams and accurately predicts the dispersion velocity Specifically, the Realizable k-ε model can include additional terms in its equations to account for the effects of anisotropy on the turbulent transport of mass and heat This makes it more accurate than other turbulence models, such as the Standard k-ε model, in predicting the behavior of turbulent flows in complex flow patterns and computed fluid density in the domain [3, 18]

The Realizable k-ε model is a mathematical model used in computational fluid dynamics (CFD) to simulate turbulent fluid flow It represents an improvement over the widely used Standard k-ε model in CFD simulations The aim of this model is to enhance the accuracy of simulations by addressing some shortcomings of the Standard k-ε model One of the main issues with the k-ε model is that it does not account for the inherently anisotropic nature of turbulence, which can lead to inaccurate results when simulating complex flows [17]

The Realizable k-ε model employs a more advanced set of equations to model turbulence It includes additional terms that represent the effects of anisotropy and also introduces a new variable called the "turbulent kinetic energy dissipation rate" to better predict the behavior of turbulence This allows for more accurate simulations

31 of turbulent flows, especially in cases where the flow exhibits high levels of anisotropy or complexity Additionally, the Realizable k-ε model is designed to provide more accurate equations and functions for complex flow phenomena such as turbulent diffusion, thus improving predictive capabilities in simulating flows of fluids with densities higher than air It is a useful tool for capturing and understanding the effects of density and gravity in complex flow simulations

The performance of the Realizable k-ε model surpasses that of the Standard k- ε model in the formulation of turbulent viscosity Furthermore, it originates from a modified transport equation for dissipation rate, stemming from an accurate equation for the transport of the square mean turbulent fluctuation Similar to the RNG k-ε model, there are significant improvements for flows exhibiting vortices or rotational motion Current research studies have demonstrated the superior performance of the Realizable k-ε model compared to other turbulence models [17]

Same like the other k-ε models the eddy viscosity is calculated by:

(3.14) However, in this model Cμ is not constant, determined by:

+ 𝛺 𝑖𝑗 : mean rate-of-rotation tensor, 𝜔 𝑘 : angular velocity

+ A0 and As: constants which are determined by:

- Turbulent production in the k-ε model:

+ Gi: component of gravitational vector

To sum up, the realizable k-epsilon turbulence model is a more advanced and accurate model than the standard k-epsilon model, and it is widely used in CFD simulations today Realizable k- ε turbulence model is a reliable and widely used tool for diffusion modeling, providing accurate and efficient predictions of turbulent flows in a variety of applications However, like all turbulence models, it has limitations and uncertainties, and should be used with care.

Species Model

The conservation equation for species is solve by predicting each local mass fraction, Yi, within the solution of a convection-diffusion of each species The general form of the conservation equation is presented by [17]:

+ Ri: net rate of species i generation by chemical reaction

+ Si: rate of creation by addition from the dispersed phase

With n fluid phase species existing in the system, n-1 species will be calculated by this form of equation Because of the consistence of the total mass fraction, N th , the selected specie, is defined as 1 minus the total mass fraction of n-1 species To archive the high accuracy, N th should be the one has the largest overall mass fraction

In turbulent flow, the mass diffusion is determined by:

+ 𝑆𝑐 𝑡 : turbulent Schmidt number calculated by 𝑆𝑐 𝑡 = 𝜇 𝑡

Solver

Ansys solver governing integral equations for the conservation of mass, momentum, energy, turbulence, and chemical species are computed numerically using Fluent The pressure-based solver and the density-based solver are the two possible solvers in Fluent While density-based solvers are used for high-speed compressible flows, pressure-based solvers are used for a variety of flow regimes, from low-speed to high-speed incompressible flows [21] Figure 3.1 summarizes the solver in ANSYS FLUENT

Figure 3.1: Solvers in ANSYS FLUENT

The pressure-based solver is typically used for simulating fluid flow problems

It uses the Navier-Stokes equations to solve for the velocity and pressure fields in the domain In ANSYS, the pressure-based solver is available in FLUENT, which is the computational fluid dynamics (CFD) solver FLUENT supports both steady-state and transient simulations using the pressure-based solver Some advantages of using the pressure-based solver in ANSYS FLUENT include its ability to handle complex

35 geometries, its accuracy in solving turbulent flows, and its robustness in handling a wide range of flow conditions [21]

However, it is important to ensure that the mesh is well-constructed and sufficiently refined to obtain accurate results Coupled algorithm is the most accurate but computationally expensive algorithm among all the above algorithms The coupled algorithm solves for the velocity and pressure fields simultaneously using a fully coupled linear system of equations It requires a highly accurate initial guess for the solution and is primarily useful for simulating highly non-linear flows

CFD SETUP AND SIMULATION

Geometry

The three-dimensional computational domain is built into many software applications, including SpaceClaim, DesignModeler, Autodesk Inventor, SolidWorks, etc The geometry in this study is generated using SpaceClaim The model shows an LPG tank located in the station where the natural wind is dispersed from the left to the right side It must be assured that this model allows for the definition of all the boundary conditions and problem domains The dimensions as well as the geometry of this work is illustrated in the Table 4.1, Figures 4.1 & 4.2 below:

Table 4.1: Basic dimension of the geometry

Figure 4.1: The geometry of the LPG station – (A) Isometric view – (B) Top view –

(C) Front view – (D) LPG tank– (E) Isometric view with domain

Figure 4.2: The geometric model of the LPG station

The gas is slowly dispersed into the atmosphere as LPG is heavier than air Therefore, the height of the calculated domain is up to 20 meters assuming no ignition source is above its height Besides, the length and width of the domain are fixed at 25 and 20 meters, respectively, for further simulation and calculation.

Geometric Discretization

One of the crucial procedures that determines how accurate the problem is meshing Poor results may lead to low mesh quality With this method, the goal is to produce a fine mesh that is sufficiently smooth and accurately captures all the crucial aspects of the geometry As long as the mesh quality (Figure 4.3) is met, meshes can be drawn using a variety of CAD, GAMBIT, ANSYS ICEM, or ANSYS Meshing programs Meshes are typically separated into two types: Structured mesh and unstructured mesh Some kinds of meshes are described in Figs 4.4 & 4.5 below, along with a simulation meshing example [22-25]

Figure 4.5: Example of a structured triangular mesh and an unstructured triangular mesh

The 3-dimensional geometry is used for this work, and the simulation and calculation are performed using a poly-hexcore volume mesh Poly-hexcore is a type of meshing technique that creates a mesh consisting of hexagonal elements It is a form of polyhedral meshing, which is a method of dividing space into small, finite cells or elements for use in numerical simulations In poly-hexcore meshing, the cells are formed by dividing the space into a series of hexagonal prisms The vertices of each prism are shared with adjacent prisms, resulting in a highly connected structure that can be used to model complex geometries with irregular shapes [26] Poly- hexcore meshes are preferred for simulation and numerical computation because they combine the flexibility of unstructured grids with the accuracy of structured grids This makes them ideal for applications in fluid and solid dynamics, flow simulations, structural computations, and thermodynamics

This type of meshing is particularly useful in applications where accuracy and efficiency are important, such as computational fluid dynamics (CFD) simulations

By using hexagonal elements, which have fewer corners than other shapes like triangles or quadrilaterals, poly-hexcore meshes reduce the number of nodes needed to represent a given geometry, and thus require less computational resources to solve

Additionally, the structure of the poly-hexcore mesh allows for better alignment with flow directions, resulting in more accurate simulations The Figure 4.6 below shows images of both polyhedral and poly-hexcore in Fluent Meshing

Figure 4.6: Illustration of a pipe with (A) – Polyhedral volume mesh and (B) –

Named Selection

In ANSYS FLUENT, the "domain naming" feature allows assigning a unique name to specific points, edges, surfaces, or a particular region within the simulation domain Naming these entities facilitates the solver process when applying boundary conditions, defining material properties, or setting up specific regions for monitoring or exporting results For example, when processing results, users can filter the display to show only the selected domains that have been named previously Naming selections enable creating parts that can be transferred to ANSYS Setup, ANSYS Post-Processing, or used to generate some features

Any combination of 3D entities can be chosen, including key points, edges, or surfaces By creating "domain names," different parts of the simulated domain can be

41 easily referenced and managed The domain selections depicted in Figure 4.7 below are identified through a domain naming process:

Figure 4.7: Named selection of the LPG storage tank (A) Wind inlet – (B) Wind outlet – (C) Release source

CFD simulation setup

In this study, despite many primary components contained in the LPG storage tank, only investigated the presence of propane and butane Propane is a hydrocarbon with the chemical formula C3H8 It is one of the primary components of liquefied petroleum gas Butane is an organic compound with the chemical formula C4H10 This hydrocarbon is one of the other flammable compounds used in LPG

Table 4.2: Properties of liquefied petroleum gas

The flammability of the hydrocarbons is shown in Table 1.2, note that those factors are under STP conditions and derived by Appendix B Flammability Data [7]:

Table 4.3: Flammability of LPG’s compositions

Heat of combustion (kJ/mol)

Flammability limit vol.% fuel in air

Lower flammability limit (LFL) denotes the minimum concentration of a gas or vapor in air needed to ignite in the presence of an ignition source like heat or flame Many safety experts equate LFL with the lower explosive level (LEL) Gas mixtures are deemed "too lean" to ignite when their concentration falls below the LFL, posing minimal risk In terms of health and safety, concentrations at the LEL are considered Immediately Dangerous to Life or Health (IDLH), despite not having stricter exposure limits On the other hand, the upper flammability limit (UFL) indicates the maximum concentration of a gas or vapor in air capable of igniting under similar conditions Concentrations surpassing UFL are regarded as "too rich" to ignite Operating above the UFL is typically avoided for safety reasons, as the introduction of air can bring the mixture into the combustibility range

The boundary temperature is maintained at 30°C The mass flow rate for four distinct scenarios is derived and computed using Equation (2.3) In the model system, two primary components of LPG, namely C3H8 and C4H10, are utilized for species transport Their respective mass fractions are set at 0.5 each, exclusively for the LPG source leakage This allocation is justified by the prominence of these components within LPG Detailed LPG leakage conditions, encompassing diameters, release flow rates with different wind velocities, are presented in Table 4.2

Table 4.4: The parameters of LPG model with the variation of wind velocity

In the event of a leakage scenario, varying wind speeds are considered These speeds range from Type A to Type F, with respective values of 1, 3, 5, and 10 (m/s), corresponding to the Pasquill-Gifford dispersion model in Table 4.5 [11] For each different wind speed scenario, the leakage duration is simulated from 0 to 1000 seconds, with extracted results Subsequently, the leakage rate as well as the dispersion behavior of LPG gas within the LPG station of the factory will be investigated

Table 4.5: Atmospheric Stability Classes for Use with the Pasquill-Gifford

Thin overcast or > 4/8 low cloud

The computational domain for CFD simulation is constructed by selecting boundary domains, including downstream inlet vents, open spaces (top, exits, and sides), bottom, and source For different scenarios, the size of the simulation space is fixed at 25 m (L) × 20 m (W) × 20 m (H) Expanding the volume of the space brings several benefits such as significantly increasing processing space, yielding more accurate computational results when emitting gas Moreover, increasing the volume promotes better airflow and ventilation in the domain, minimizing the risk of confined spaces and suboptimal dispersion of emitted airflow Thus, a more accurate assessment of LPG emission behavior in the station during leakage accidents can be made

However, this volume increase presents an unfavorable aspect of increasing computational costs When expanding the computational domain, the number of grid cells, nodes, and faces must be increased during the mesh partitioning process for the model, leading to significantly longer computational times compared to using a smaller volume Additionally, other parameters, including turbulence intensity (%), turbulent viscosity ratio, direction, temperature (°C), etc., are set at default values Figure 4.8 provides an overview of all relevant variables of boundary conditions

Figure 4.8: Schematic diagram of computational LPG model

In employing the transient formulation within Fluent, a first-order implicit method was utilized, incorporating variation time step sizes of 0.001, 0.005, 0.1, 0.2, and 0.5 seconds For the spatial discretization of density, momentum, energy, and turbulent quantities (k and ε), a second-order upwind scheme was selected, while a second-order scheme was employed for the pressure field to enhance precision The

46 pressure-velocity coupling algorithm was executed using the pressure implicit with splitting of operators (PISO) method, recommended for transient computations Default settings for the blend between neighboring corrections and deviation were adopted for the PISO algorithm Furthermore, convergence of computational results at each time step was deemed achieved upon meeting one of the following criteria:

1 The scaled residual sum is less than 1e-3

2 At a particular time step, the ratio of the remaining quantity to the initial quantity is less than 0.05

Furthermore, to simplify the calculations, several technical assumptions are made:

1 LPG leaks at a constant rate

2 Phase changes and droplet deposition during leakage are ignored

3 The wall is both isothermal and adiabatic, with no heat exchange between LPG and air

The detailed mathematical formulations of the problem were provided in the previous chapter The table and figure below summarize the conditions of the model

Table 4.6: Summary of boundary conditions

Species Model Diffusion Energy Source

RESULTS AND DISCUSSION

Selection of the grid for the simulation model

To ensure rigorous logic and optimize the simulation before running in transient state conditions, it is essential to first conduct a mesh independent analysis to select the most suitable grid type [27] Be advised that this grid selection solely will be conducted in the steady state at 3 m/s wind velocity (sample) for the simulation

By that, Table 5.1 illustrates the mesh quality, number of cells, nodes, and faces for three different grid resolutions: coarse, medium, and fine It is noted for Table 5.1 that the mesh quality is considered “Good” when the skewness value is approximately 0.54 and the minimum orthogonal quality is 0.5 for all the grid resolutions This information, combined with the velocity values provided in Table 5.2 at three specific test points (A, B, and C) with given coordinates, can be used to determine the most appropriate grid for the simulation

Table 5.1: Mesh resolutions for the LPG simulation

Table 5.2: The coordinates and velocity values of test points in the sample case

Table 5.3: Statistics for specific points at different grid resolutions

Figure 5.1: Velocity contour of the sample case with the locations of test points

Figure 5.2: Velocity of three test points under different grid resolutions

Coarse Grid Medium Grid Fine Grid Finest Grid

Figure 5.1 & 5.2 examining the velocity values at the three points across the different grid resolutions, while Table 5.3 presents the coefficients of variation for specific points across four mesh cases, ranging from coarse to finest Upon examining the velocity values at the three specified points (A, B, and C) across the different grid resolutions, it becomes clear that there is no significant change in the velocity predictions as the grid is refined

At point A, the velocity ranges from 12.41 m/s in the coarse grid to 12.64 m/s in the finest grid, a difference of only 0.23 m/s Similarly, the velocity values at points

B and C show a relatively small variation, with the finest grid only slightly higher than the coarse grid Furthermore, the statistical analysis of the velocity data in Table 5.3 reinforces the similarity in the velocity predictions The mean velocity across point A is 12.41, 12.62, 12.66 and 12.64 m/s for the coarse, medium, fine and finest grid, respectively with point B and C These values suggesting that the finest grid has slightly more consistent velocity predictions, given the lack of significant improvement in the velocity values as the grid is refined, thus the medium grid resolution emerges as the most appropriate choice for the simulation

The medium grid, with 85,865 cells, 226,273 nodes, and 384,019 faces, provides a good balance between computational cost and solution accuracy Compared to the finest grid, which has significantly more cells, nodes, and faces (276,285 cells, 569,939 nodes, and 1,088,225 faces), the medium grid offers a more efficient option without sacrificing the quality of the results The similarity in velocity predictions across the grids suggests that the medium grid is capable of accurately capturing the flow dynamics within the simulation domain

Therefore, based on the data provided, the recommendation is to use the medium grid resolution for the simulation This choice strikes the right balance between computational resources and solution reliability, ensuring that the simulation can be conducted efficiently without compromising the accuracy of the results The coarse grid likely lacks the necessary resolution to fully resolve the flow features,

51 while the fine and finest grids may introduce unnecessary computational overhead without significantly improving the accuracy of the results.

Meshing for the computational simulation

The Fluent Meshing software used by ANSYS Fluent will be utilized for meshing ANSYS Fluent Meshing, a component of the ANSYS Fluent software suite, is a widely used computational fluid dynamics (CFD) tool for simulating fluid flow, heat transfer, and related phenomena Designed to generate high-quality grids, Fluent

Meshing streamlines the process of creating the necessary complex meshes for accurate and reliable simulation results Key features include automatic mesh generation, adaptive mesh refinement, polyhedral meshing, parallel processing capability, boundary layer meshing, compatibility with ANSYS Fluent, flexibility, and user-friendly design By leveraging ANSYS Fluent Meshing, engineers and analysts can efficiently create meshes tailored to their simulation requirements, thereby enhancing the accuracy and reliability of their CFD simulations [28]

From here, poly-hexcore is chosen for the volume meshing technique used in

Ansys Fluent to create high-quality hexahedral grids for complex three-dimensional simulations, especially for 3D models as illustrated in Figure 5.3

Figure 5.3: The poly-hexcore volume mesh of the simulation model

In the context of using Fluent Meshing, skewness and orthogonality are two important measures of mesh quality that can impact the accuracy, convergence, and stability of simulations Skewness assesses the deformation or non-orthogonality of mesh elements Fully regular elements have a skewness value of 0, while highly distorted elements have higher skewness values High skewness values can lead to numerical instability, convergence issues, or inaccurate results in simulations Therefore, it is often recommended to keep the average skewness below 0.9 Orthogonality measures the angle between two faces of mesh elements, with lower orthogonality values potentially resulting in numerical diffusion, mesh noise, or mesh-dependent issues in finite volume method (FVM) simulations Based on the mesh quality in Figure 4.6, both the maximum skewness and minimum orthogonality values meet the requirements before conducting the simulation

Mesh nodes, faces, and cells are crucial to ensure that their density is sufficiently high to achieve the necessary level of detail in simulations when generating the mesh Fluent Meshing provides a more convenient and straightforward

53 approach to FEA simulation, with the number of cells, faces, or nodes from the mesh boundaries displayed in Table 5.3

Table 5.4: Mesh size information of computation domain

In meshing with Fluent, "cells" refer to the individual volumes or elements in a mesh "Faces" refer to the individual surfaces that make up the boundaries of these volumes, and "nodes" refer to the individual points that define the vertices of these elements The number of cells in a mesh will determine its overall resolution, with a higher number of cells usually resulting in a finer and more accurate simulation, but also requiring more computational resources The number of faces and nodes will affect the accuracy of the boundary conditions and the ability to resolve complex geometries in the mesh If the mesh is too sparse, there may be insufficient resolution to capture important flow phenomena, resulting in inaccurate results On the other hand, if the mesh is too fine, it can cause excessive computational time and memory usage, which can significantly increase the simulation time and cost

Therefore, it is essential to strike a good balance between the number of cells, faces, and nodes This balance depends on the specific problem and the required level of accuracy in the simulation Generally, it is recommended to use the minimum number of cells, faces, and nodes that can properly capture the flow features of interest, while minimizing computational cost.

Determining the flammability range of the gas mixture

Using the parameters provided in Table 4.3, determining the lower flammable limit (LFL) for a flammable gas mixture can be accomplished using the Le Chatelier

54 mixing rule The LFL of each individual combustible component in the gas mixture can be calculated using the following equation [3]:

In this equation, 𝑦 𝑖 ′ represents the molar fraction of the combustible component i within the gas mixture, while 𝐿𝐹𝐿 𝑖 denotes the Lower Flammable Limit (LFL) of component i within the gas mixture An analogous methodology is employed to ascertain the Upper Flammable Limit (UFL) of the gas blend Within this investigation, liquefied petroleum gas (LPG), comprising propane and butane, has been the subject of examination Specifically, the lower flammable limits for propane (𝐿𝐹𝐿 𝐶 3 𝐻 8 ) and butane (𝐿𝐹𝐿 𝐶 4 𝐻 10 ) were determined as 2.1% and 1.8% by volume, respectively The computation of the lower flammable limit for the gas mixture is as follows [3]:

From equation 5.2, the results of the lower flammability limit (LFL) and upper flammability limit (UFL) of the LPG gas mixture in the study were 1.96% and 9.04%, respectively.

The simulation computational model in transient-state

The transient-state simulation computational model evaluates a system's behavior over time, capturing dynamic responses and temporal effects that steady- state models do not Figure 5.4 provides a comprehensive visualization of the average LPG volume over time, with time intervals measured in units of 1000s Across all four wind velocities analyzed, a consistent initial pattern emerges, characterized by a progressive increase in the average LPG volume However, this trend was soon interrupted by pronounced fluctuations due to wind diffusion dynamics, resulting in

55 a significant decrease in LPG concentration within the 1s to 100s time range During this period, the LPG in the calculation domain experienced a slight decline, as depicted in the figure

Notably, the stability gradually ensues in the subsequent seconds The volume average of LPG within the enclosure attains steady approximately at 200s into the simulation, marking a crucial milestone in the analysis This value is considered to be in a steady state because the volume average of LPG does not change significantly Specifically, those average values are approximately 1.3e-5, 4.4e-6, 2.6e-6, and 1.5e-

6 for wind velocities of 1 m/s, 3 m/s, 5 m/s, and 10 m/s, respectively A more detailed examination reveals that the expression of the average LPG volume within the enclosure exhibits logical fluctuations over time, characterized by periodic increases, decreases and stability This behavior converges to a nearly constant concentration, signifying the establishment of a stable state

This observation underscores the crucial utility of computational fluid dynamics (CFD) modeling, which not only demonstrates feasibility but also proves to be well-suited for facilitating reasonable predictions and enabling further detailed calculations in this study

Figure 5.4: The volume average of LPG for different wind velocities versus time

LPG behavior during the dispersion with different wind velocities

According to calculations from equation 5.2, the flammability for this simulation ranges from 1.96% to 9.04% by volume of LPG This means that the presence of an ignition or electrostatic source within this range can lead to very serious consequences Therefore, Figures 5.5 and 5.6 will focus on LPG concentrations in the symmetric plane, ranging from 0% to 1.96%, which demonstrates concentrations at wind velocities of 1, 3, 5, and 10 m/s, allowing analysis of the contour when leaks occur at various wind speeds over a period of 1 to

Firstly, at the initial time step of 1 second, the LPG mass fraction contour indicates a concentrated distribution near the source location, depicted by darker shades when t = 1s As one moves away from the source, the concentration decreases, resulting in significantly lighter shades By 5 seconds, LPG begins to disperse more rapidly and extensively as it interacts with the incoming air and rises to higher

57 altitudes, although it remains relatively close to the release point This phenomenon is most noticeable in the case of a 1 m/s wind speed, where the diffusion wind speed is lower than the LPG leakage rate As a result, the gas tends to rise higher compared to the gas profiles observed at higher wind speeds This observation underscores the importance of the regulation TCVN 6486:2008, which mandates that LPG tanks be placed outdoors, rather than on rooftops, balconies, or in basements, to prevent gas accumulation in the event of a leak When the gas rises higher, it is more effectively dispersed by the wind, reducing its concentration and promoting rapid diffusion into the surrounding air This ensures that the LPG concentration becomes extremely low, nearly completely diffused and can not be ignited, thereby mitigating potential hazards associated with gas leaks

As the time increases to 15 and 60 seconds, the LPG will disperse further, with the contours extending in the direction of the wind, corresponding to lighter shades due to the diffusion with the air The contour plots show a clearer dispersion pattern, indicating a significant expansion of the LPG mass fraction, which now covers a larger area than the previous time steps The elongated contour lines reflect the transport of LPG in the prevailing wind direction This expansion represents a larger area affected by gas release and illustrates the complex interaction between wind dynamics and LPG diffusion Higher LPG mass fractions were observed in regions corresponding to wind direction, indicating a significant influence of wind on the widespread dispersion of LPG

Finally, at 200s, the LPG mass fraction contour becomes stable, signifying an extensive distribution of LPG across a considerably larger area in downwind distance, indicating a more even dispersion of the gas The contours exhibit a more uniform spread, indicating a homogenized dispersal of the gas The influence of wind becomes more pronounced as the LPG mass fraction is transported over greater distances, impacting regions distant from the source Additionally, it is observed that the LPG mass fraction is negligible, as depicted in Figure 5.5 & 5.6 At this time (200s), the concentration at the furthest location among all cases is measured below 0.44 vol.%,

58 suggesting that LPG cannot be ignited by any ignition source due to the inability to reach flammable concentrations

In summary, analysis of the LPG mass fraction contour at different time steps and wind velocities shows a clear trend of increasing dispersion with time Notably, in all four cases, the wind speed of 10 m/s demonstrates the earliest stabilization from the onset of the simulated leak, as evidenced by the contour remaining unchanged at t = 5s onwards Similarly, this stabilization is observed at t = 10s for the 5 m/s wind case and between t = 30s and 60s for the 1 m/s and 3 m/s wind cases This observation suggests that higher wind speeds expedite the diffusion process and reduce the time required for stabilization, as a larger volume of leaked air disperses more rapidly Thorough observations and patterns discerned from LPG mass fraction contours offer valuable insights into the spatial and temporal evolution of LPG dispersion, aiding in the comprehension and mitigation of potential risks

Figure 5.5: LPG mass fraction at wind speeds of 1 m/s and 3 m/s on the symmetry plane

Figure 5.6: LPG mass fraction at wind speeds of 5 m/s and 10 m/s on the symmetry plane

Analysis of average LPG concentration in points

To gain a deeper understanding of LPG emission behavior, each wind scenario was configured with a series of measurement points to track changes in emission concentration from the initial release until the model is stabilized The coordinates and names of these points are detailed in Table 5.4 These measurement points were strategically selected based on the LPG concentration contour at t = 200s (shown in Figure 5.7), aiming to survey and analyze the concentration and distribution of LPG at various locations on the same surface, specifically at plane Y = 1.35 m along the model's ZX coordinate axis Additionally, these points were selected for specific analytical purposes: point A is projected to exhibit a concentration range between 1.7 and 2 vol.%, point B is expected to display a concentration range between 0.4 and 0.6 vol.%, and point C is anticipated to demonstrate the lowest concentration, less than 0.2 vol.%

The data captures a detailed examination of LPG concentrations at three distinct points within an enclosure for each wind velocities, measured at 1 m/s, 3 m/s,

5 m/s, and 10 m/s Each wind velocity significantly impacts the dispersion and concentration of LPG, offering critical insights into the gas's behavior under varying conditions, indicated in Figure 5.7

Table 5.5: Coordinates of set of points in center plane at ZX axis

At a wind velocity of 1 m/s, the LPG concentration at all three points (A1, B1, and C1) shows initial fluctuations as the gas begins to disperse Point A1 (1.23, 1.35, 5.92), experiences relatively stable yet higher concentrations of LPG initially due to the lower wind speed, with an average LPG value of 1.2 vol.% as indicated in the chart for this velocity This reduced wind speed results in less effective dispersion, causing initial gas concentrations to range from 1.16 to 1.23 vol.% within the first 20 seconds, nearly reaching the LFL of LPG calculated in the above result (1.96 vol%) Similar trends are observed at Points B1 and C1, where the gas takes longer time to spread out and reaches the points, resulting in sustained higher LPG levels through time with 0.27 and 0.18 (vol.%) at 200s, respectively The lower wind velocity allows ranges of higher gas concentration to form, which can pose significant safety hazards if not adequately ventilated

Increasing the wind velocity to 3 m/s results in a more efficient dispersion of LPG At point A2, the gas concentration witnessed a rapid decrease compared to the

1 m/s scenario Specifically, the concentration decreased from 2% to 1.79% in the first 7 seconds, followed by increased oil and vibration stabilization before reaching stability at 1.86% at 200 seconds indicating more effective dilution Point B2 also benefits from the increased wind speed, with LPG levels decreasing at a quicker rate, thereby reducing the risk of high concentration zones Point C2 shows noticeable improvements, as the higher wind velocity ensures lower LPG levels, enhancing the overall safety within the enclosure This velocity helps in preventing the accumulation of hazardous gas concentrations

At 5 m/s, the LPG dispersion at Point A3 becomes even more efficient The higher wind velocity quickly moves the gas away from the release point, significantly lowering the concentration levels Point B3 exhibits a similar trend, with LPG levels dropping rapidly due to the increased wind speed This scenario helps maintain safer conditions by effectively diluting the gas Point C3 also experiences improved dispersion, with concentrations reducing more swiftly, resulting in a more uniform distribution of LPG within the enclosure Despite the notable reduction in LPG concentration and efficient diffusion, the chart illustrates extended turbulence This leads to a broader fluctuation range in LPG concentration, posing challenges in predicting gas emission flow across all three locations within the initial 50-second period To sum up, higher wind speed at 5 m/s demonstrates a marked improvement in gas dilution and safety

At the highest wind speed of 10 m/s, LPG at point A4 disperses almost immediately, resulting in a very low concentration The chart for this wind speed shows an extremely steep decline in the first few seconds, followed by a gradual increase and stabilization This indicates that when the leak begins, the high wind speed quickly disperses the LPG, causing a significant drop in the measured concentration at point A4 Points B4 and C4 also witnessed the same thing when the concentration at these two locations had a large disturbance at 1.5s and then gradually

64 became more stable before 50s This wind speed shows the most efficient gas dilution, minimizing the risk of high LPG concentrations Point B4 benefits greatly from the 10 m/s wind speed, with LPG levels dropping sharply and maintaining lower averages throughout the measurement period The rapid dispersion at this velocity ensures a safe environment Point C4 sees the most significant reduction in LPG levels, with the high wind velocity ensuring that the gas is quickly spread out and diluted, substantially reducing the potential for hazardous concentrations

The analysis of LPG dispersion at various wind velocities in specific points reveals crucial insights into how airflow affects gas concentrations within an enclosure Increased wind speeds consistently lead to improved dispersion, with concentration values at points B and C, averaging approximately below 0.4 and 0.2 vol.%, respectively, across all four wind speed scenarios at a specific stable time t 200s, lowering the average LPG levels at all cases of measurement This information is vital for designing efficient ventilation systems and ensuring safety in environments where LPG is used or stored By optimizing wind velocities, it is possible to maintain safe and controlled conditions, preventing the buildup of hazardous gas concentrations

Figure 5.7: The average LPG concentration at points within the symmetry plane across four wind velocities

Wind velocity distribution during the dispersion

The impact of wind speed on the dispersion of LPG leakage is a critical issue due to the potential dangers associated with LPG leakage incidents When LPG is released into the air, it can mix with the atmosphere to form a flammable combination that may ignite and explode Therefore, it is essential to comprehend how LPG disperses under different environmental conditions, including varying wind speeds Higher wind speeds facilitate quicker and wider dispersion of pollutants, while lower wind speeds can lead to the accumulation and concentration of pollutants in specific areas, as depicted in Figure 5.8 & 5.9

To be more specific, starting with the time step of 1 second, the velocity profile illustrates the flow patterns associated with each wind velocity By examining the contour plot for each velocity, it is possible to observe the direction and intensity of the air flow at each wind velocity instance Progressing to the time step of 5 seconds, the wind profile updates to reflect the evolving wind patterns The contour plot allows for a comparison between different wind velocities, showcasing any variations in the flow behavior The contour lines will reveal areas where the airflow is more concentrated and regions where it is more dispersed or turbulent with darker shades

Moving further to the time step of 15 seconds, the contour continues to capture changes in the wind patterns The profile’s plot serves as a visual representation of the velocity distribution across the defined area, highlighting any notable variations between different wind velocities This information can provide insights into how the wind interacts with the surrounding environment As progressing to the time steps of

30 seconds and 60 seconds, the contour provides a more comprehensive understanding of the airflow dynamics The contour lines portray the distribution of wind velocities, allowing us to identify areas of high and low speeds, which can be observed how the wind flow changes over time and how different wind velocities impact the overall behavior From this time, the winds start to be stable throughout the cases and the stabilized gradually formed Finally, at the time steps of 200 seconds

70 and 1000 seconds, the wind profile provides a complete depiction of the wind patterns

It can be seen among all cases that the influence of the front-wall on the wind field affects significantly the diffusion of LPG emissions Initially, the wind speed starts at 1, 3, 5 or 10 m/s and then gradually increases to speeds from 4, 8, 12 m/s up to 14-22 m/s, as illustrated in the legends of Figure 5.6 & 5.7, respectively, and has a slight upward direction after crossing the front wall The increased wind field behind the front wall can be explained by the blocking effect of objects on the wind field, amplifying ambient air turbulence, aiming to improve the speed of movement and increase the rate of LPG dilution The wind is then mixed with the emitted LPG source and creates an upward trend in the domain

Furthermore, in scenarios where wind speeds level at 5 m/s and 10 m/s, a segment of the airflow is directed into the void between the tank and the two adjacent walls, creating several vortices This phenomenon arises because high-velocity air disperses alongside the quantity of leaked LPG situated behind the wall, forming a vortex in the empty space Meanwhile, the vortex in front of the wall is a result of obstruction caused by the LPG tank This phenomenon is absent in the wind conditions of 1 m/s and 3 m/s due to the fact that the diffusive wind velocity is lower than the emission velocity of the leak At these lower wind speeds, the diffusion wind velocity remains insufficient to induce vortex formation in the space behind the LPG tank However, the obstruction effect persists, leading to the formation of vertices in front of the tank

In summary, the velocity profile in the provided file offers a comprehensive visualization of wind patterns at different time steps and wind velocities By examining the contour plots, we can gain insights into the direction, intensity, vortices and distribution of airflow, enabling a deeper understanding of the dynamics and behavior of the wind in the given scenario This detailed insight into the velocity behavior aids in understanding the overall airflow dynamics and can be valuable for

71 various applications, such as assessing pollutant dispersion or optimizing ventilation systems

Figure 5.8: Velocity profiles at 1 m/s and 3 m/s on the symmetry plane

Figure 5.9: Velocity profiles at 5 m/s and 10 m/s on the symmetry plane

Insights of velocity contour on LPG dispersion and leakage

To analyze the insights of velocity, three particular planes are examined simultaneously at intervals of 200 seconds These planes are positioned along the YZ axis at three distinct locations: X = -5m (pre-source), X = 1.22m (mid-source), and X

= 10m (post-source) within the boundary and clearly demonstrated in Figure 5.10 Besides, studying the velocity patterns across these planes can provide valuable understanding of LPG dispersion and potential leakage from the room

Starting with the plane at X = -5m, we can examine the velocity behavior in that region through contour plots, which depict the airflow's direction and intensity within the enclosure This insight is pivotal for comprehending potential LPG dispersion patterns and identifying areas prone to leakage Before the source plane (X = -5m), the dynamics of LPG dispersion are heavily influenced by wind velocity This information is crucial for understanding how LPG dispersion may occur and whether there are any areas prone to leakage

At the center plane at X = 1.22m, there are crucial insights into airflow patterns and velocities within the enclosure Contour plots provide visual representations of these dynamics, allowing for the identification of regions with varying velocities Such distinctions are vital in understanding how LPG dispersion and potential leakage may occur Particularly, in the center plane, the dispersion patterns of LPG exhibit notable variations according to wind velocity At lower velocities (1 m/s and

3 m/s), the LPG displays moderate spreading, accompanied by some accumulation in the mid-region of the enclosure with the value not exceeding 6.4 m/s This suggests a relatively slow movement of the gas, potentially leading to localized areas of higher concentration However, at higher velocities (5 m/s and 10 m/s), the LPG demonstrates significant movement through the center plane, resulting in more uniform distribution and effective mixing of over 9.5 m/s This emphasizes the importance of adequate ventilation to facilitate rapid dispersion and prevent hazardous gas concentrations

The velocity distribution at the post-source plane (X = 10m) reveals insights into airflow patterns and velocities in this region At lower wind speeds (1 m/s and 3 m/s), LPG movement increases but remains slower, potentially leading to gas accumulation near edges Conversely, at higher speeds (5 m/s and 10 m/s), LPG swiftly moves towards and out of the boundary, emphasizing the critical role of effective ventilation systems in ensuring quick gas removal Furthermore, both the pre-source plane and post-source plane exhibit the presence of vortices in the LPG station, as noted in result 5.6 Specifically, at wind speeds of 1 m/s and 3 m/s, the absence of significant velocities is attributed to the relatively low wind speed However, at wind speeds of 5 m/s and 10 m/s, noticeable disparities are observed below the LPG tank with an increased velocity between 6.4 – 9.5 m/s within the LPG station, which validates the aforementioned result

In summary, analyzing contour velocity plots within the enclosure provides valuable insights into velocity behavior concerning LPG dispersion and potential leakage, aiding in the identification of areas with varying velocities The pre-source plane reveals that dispersion is heavily influenced by wind velocity, crucial for understanding gas spread and identifying leak-prone areas These findings underscore the importance of considering wind velocity and effective ventilation in managing LPG leaks to prevent hazardous accumulations and ensure rapid gas evacuation across different planes in the enclosure At the mid-source plane, lower wind speeds (1-3 m/s) can lead to moderate LPG spreading and localized accumulation, while higher speeds (5-10 m/s) result in more uniform distribution and effective mixing The post-source plane further emphasizes the importance of adequate ventilation, as lower speeds can cause gas buildup near walls with a slightly increased velocity, while higher speeds facilitate swift LPG removal

Figure 5.10: Contours of velocity distribution at t = 200 s on (a) X = -5 m; (b) X =

CONCLUSIONS AND RECOMMENDATIONS

Conclusions

This research entails developing a three-dimensional model to simulate the behavior of LPG concentrations during a leak incident The model dimensions are derived from factory schematics, and the simulation employs half-symmetry to reduce computational load and processing time The simulations were conducted under transient-state conditions, allowing for dynamic analysis of LPG dispersion in a more realistic point of view Besides, the study provides valuable predictions regarding the impact of wind speed on LPG dispersion patterns and concentration fields with several conclusions below

Firstly, the employment of ANSYS Fluent Meshing software to generate precise computational meshes for Computational Fluid Dynamics (CFD) simulations is a notable endeavor By employing poly-hexcore meshing methodology, this investigation aims to strike a harmonious equilibrium between adaptability and computational efficiency It is noted that the medium grid resolution strikes a balance between precision and computational efficiency, making it the preferred option for the simulation It accurately predicts velocity patterns at key locations while keeping computational demands manageable The resultant mesh comprises 85,865 cells, characterized by skewness and minimum orthogonality metrics are 0.54 and 0.5 respectively, which proposes a favorable range of “Good”, enhancing the fidelity and precision of the computational simulation process

Secondly, the flammability’s result provides the methodology and calculations to determine the flammability range from 1.96 & 9.04 vol.% (LFL and UFL) of the LPG gas mixture, which is an important safety consideration for handling and using such flammable gas mixtures Furthermore, the average LPG volume within the enclosure stabilized around the 200-second mark, exhibiting values of approximately 1.3e-5, 4.4e-6, 2.6e-6, and 1.5e-6 for wind velocities of 1 m/s, 3 m/s, 5 m/s, and 10 m/s, respectively

Next, the analysis of LPG dispersion behavior under different wind velocities provides valuable insights into the dynamics of gas leaks and the importance of proper mitigation strategies Higher wind speeds ranging from 5-10 m/s facilitate quicker and more widespread dispersion of the LPG, while lower wind speeds of 1-3 m/s leads to increased concentration and accumulation The blocking effect of the front wall and rear wall creates turbulence and vortices that further affect the airflow patterns, which means there will be a small amount of LPG is mixed into the space in those vortices, especially at 5 m/s and 10 m/s wind speeds At lower velocities, the diffusion wind is insufficient to induce vortex formation behind the LPG tank

Furthermore, LPG dispersion at various wind velocities in specific points impact on gas concentrations within an enclosure Higher wind speeds consistently enhance dispersion, lowering average LPG levels across all scenarios at a stable time of t 0s, particularly at points B and C, where concentrations average below 0.4 and 0.2 vol.%, respectively This data is crucial for designing effective ventilation systems and ensuring safety in LPG-utilizing or storing environments By optimizing wind velocities, maintaining safe and controlled conditions becomes achievable, averting hazardous gas concentration buildup

Finally, at the steady time, the center plane (X = 1.22 m) reveals the diffusion Moreover, it indicates that this dispersion efficiency progressively enhances with higher wind speeds demonstrated in darker shades in the velocity’s contours (Figure 5.8) Additionally, the surfaces before and after the source display wind vortices within the LPG station's open space, validating the findings of result 5.6, particularly evident at wind speeds of 5 m/s and 10 m/s

To summarize, timely leak detection and resolution are imperative for averting severe incidents These findings indicate that employing CFD simulation is a dependable approach for predicting the dynamics of LPG and wind concentration patterns in open-space factory environments over time Notably, at the 200-second mark, LPG behavior stabilizes, reaching a steady condition, enabling insights and predictions regarding its concentration trends This method aptly anticipates the

79 ramifications of LPG dispersion, underscores safety considerations, and facilitates compliance with National Standards to ensure optimal safety in cases of flammable or explosive gas leaks.

m; (c) X = 10 m

This research entails developing a three-dimensional model to simulate the behavior of LPG concentrations during a leak incident The model dimensions are derived from factory schematics, and the simulation employs half-symmetry to reduce computational load and processing time The simulations were conducted under transient-state conditions, allowing for dynamic analysis of LPG dispersion in a more realistic point of view Besides, the study provides valuable predictions regarding the impact of wind speed on LPG dispersion patterns and concentration fields with several conclusions below

Firstly, the employment of ANSYS Fluent Meshing software to generate precise computational meshes for Computational Fluid Dynamics (CFD) simulations is a notable endeavor By employing poly-hexcore meshing methodology, this investigation aims to strike a harmonious equilibrium between adaptability and computational efficiency It is noted that the medium grid resolution strikes a balance between precision and computational efficiency, making it the preferred option for the simulation It accurately predicts velocity patterns at key locations while keeping computational demands manageable The resultant mesh comprises 85,865 cells, characterized by skewness and minimum orthogonality metrics are 0.54 and 0.5 respectively, which proposes a favorable range of “Good”, enhancing the fidelity and precision of the computational simulation process

Secondly, the flammability’s result provides the methodology and calculations to determine the flammability range from 1.96 & 9.04 vol.% (LFL and UFL) of the LPG gas mixture, which is an important safety consideration for handling and using such flammable gas mixtures Furthermore, the average LPG volume within the enclosure stabilized around the 200-second mark, exhibiting values of approximately 1.3e-5, 4.4e-6, 2.6e-6, and 1.5e-6 for wind velocities of 1 m/s, 3 m/s, 5 m/s, and 10 m/s, respectively

Next, the analysis of LPG dispersion behavior under different wind velocities provides valuable insights into the dynamics of gas leaks and the importance of proper mitigation strategies Higher wind speeds ranging from 5-10 m/s facilitate quicker and more widespread dispersion of the LPG, while lower wind speeds of 1-3 m/s leads to increased concentration and accumulation The blocking effect of the front wall and rear wall creates turbulence and vortices that further affect the airflow patterns, which means there will be a small amount of LPG is mixed into the space in those vortices, especially at 5 m/s and 10 m/s wind speeds At lower velocities, the diffusion wind is insufficient to induce vortex formation behind the LPG tank

Furthermore, LPG dispersion at various wind velocities in specific points impact on gas concentrations within an enclosure Higher wind speeds consistently enhance dispersion, lowering average LPG levels across all scenarios at a stable time of t 0s, particularly at points B and C, where concentrations average below 0.4 and 0.2 vol.%, respectively This data is crucial for designing effective ventilation systems and ensuring safety in LPG-utilizing or storing environments By optimizing wind velocities, maintaining safe and controlled conditions becomes achievable, averting hazardous gas concentration buildup

Finally, at the steady time, the center plane (X = 1.22 m) reveals the diffusion Moreover, it indicates that this dispersion efficiency progressively enhances with higher wind speeds demonstrated in darker shades in the velocity’s contours (Figure 5.8) Additionally, the surfaces before and after the source display wind vortices within the LPG station's open space, validating the findings of result 5.6, particularly evident at wind speeds of 5 m/s and 10 m/s

To summarize, timely leak detection and resolution are imperative for averting severe incidents These findings indicate that employing CFD simulation is a dependable approach for predicting the dynamics of LPG and wind concentration patterns in open-space factory environments over time Notably, at the 200-second mark, LPG behavior stabilizes, reaching a steady condition, enabling insights and predictions regarding its concentration trends This method aptly anticipates the

79 ramifications of LPG dispersion, underscores safety considerations, and facilitates compliance with National Standards to ensure optimal safety in cases of flammable or explosive gas leaks

The research results suggest a number of techniques to improve accuracy as well as simulate most realistically when LPG leaks occur, developed from the steady state model

Firstly, when simulating LPG leakage into the environment, it's essential to account for factors such as temperature exchange and humidity to improve the accuracy and realism of the calculations Temperature differences between the gas in the tank and the surrounding environment affect the rate and behavior of gas leakage, impacting pressure, volume, and density Meanwhile, humidity influences gas behavior by altering its density and diffusion rate, particularly in scenarios where the leaked gas interacts with water vapor Incorporating these factors into simulations involves intricate mathematical models encompassing heat transfer, thermodynamics, fluid dynamics, and chemical reactions

Secondly, expanding the consideration of LPG leak angles beyond the vertical direction is important for comprehensive analysis of leak scenarios While vertical leaks are important, combining different leak angles allows for the assessment of a broader range of situations, providing insights into how gas disperses in different directions and within different conditions Horizontal or angled leaks can result in different dispersion patterns and potential hazards than vertical leaks By including more leak angles in simulations and analysis, researchers and safety professionals can gain a more comprehensive understanding of gas dispersion dynamics, supporting the development of effective mitigation strategies and emergency response plan This broader approach enhances the overall safety assessment and preparation of LPG processing and storage systems

Thirdly, exploring different wind angles is essential for understanding how LPG concentration emission behaviors change While the wind angle is currently fixed in a single direction, varying it allows for a more comprehensive analysis of dispersion patterns and potential exposure risks Different wind angles can significantly influence the trajectory and spread of the gas plume, affecting dispersion rates and the concentration of LPG in the environment By incorporating a range of wind angles into simulations and analyses, researchers can gain valuable insights into how atmospheric conditions impact gas dispersion dynamics, aiding in risk assessment and the development of effective safety measures

Finally, considering cases where the emission remains constant in a period of time while the wind speed varies over time is crucial for simulations that mirror real- world conditions Investigating how LPG behavior evolves within the calculation domain under fluctuating wind speeds under the same leakage condition can yield valuable insights into dispersion patterns, plume behavior, and potential exposure risks By incorporating dynamic wind speed variations into simulations, we can better understand how atmospheric conditions influence gas dispersion dynamics over time, aiding in more accurate risk assessment and the development of effective safety measures for LPG handling and storage systems

K N Le, Y H Duong, T M Le, N T Dang, D T Le, and V T Tran, “Simulation of the LPG Leakage and Dispersion Process to the Factory Environment Using Computational Fluid Dynamics (CFD)”, Chemical Engineering Transactions, vol

Paper Received: 13 June 2023; Revised: 8 September 2023; Accepted: 21 September 2023

Please cite this article as: Le K.N., Duong Y.H., Le T.M., Dang N.T., Le D.T., Tran V.T., 2023, Simulation of the LPG Leakage and Dispersion Process to the Factory Environment Using Computational Fluid Dynamics (CFD), Chemical Engineering Transactions, 106, 499-504 DOI:10.3303/CET23106084

The Italian Association of Chemical Engineering Online at www.cetjournal.it Guest Editors: Jeng Shiun Lim, Nor Alafiza Yunus, Peck Loo Kiew, Hon Huin Chin

Simulation of the LPG Leakage and Dispersion Process to the Factory Environment using Computational Fluid Dynamics

Khoi N.M Le, Yen H.P Duong, Tan M Le, Nguyen T Dang, Duc T Le, Viet T Tran*

Faculty of Chemical Engineering, Ho Chi Minh City University of Technology (HCMUT), 268 Ly Thuong Kiet Street, Ho Chi Minh City, Vietnam

Vietnam National University Ho Chi Minh City (VNU-HCM), Linh Trung ward, Thu Duc District, Ho Chi Minh City, Viet Nam trantanviet@hcmut.edu.vn

Liquefied Petroleum Gas (LPG) is a hydrocarbon gas that exists in liquefied form as an energy-saving, clean, convenient fuel, it has been widely used in civil and industrial applications such as oil fields, gas and oil industry This present work considers the flammable gas release and dispersion safety evaluation caused by the LPG (50 % propane, 50 % n-butane) leakage and diffusion from an accidentally punctured pipe (with the diameters varies from 1-3 mm) during the operation of a 33 m3 LPG tank Process equipment during a puncture can swiftly release hazardous compounds in sufficient quantities to distribute throughout a working and local area in clouds LPG leakage diffusion model is created using the computational fluid dynamics (CFD) method, and the characteristics of LPG dispersion and the various of LPG release rate - dispersed by natural wind in an open area, with wind speeds range of 1.0 – 10.0 m/s corresponding to different atmospheric stability classes from A to F are computed and simulated The result shows that wind speeds have a considerable impact on the spread of LPG from the release site to different distances in the domain The maximum LPG concentration distribution is 101.71 ppm at 1 m/s wind velocity, while the minimum is 3.12 ppm at 10 m/s wind velocity This suggests that the dispersion of LPG in the environment is minimal, accounting for roughly 0.01% of its total volume, which implies that LPG stations are not at significant risk in the event of a small puncture ranging from 1 to 3 mm in diameter In the simulation work, the realizable k-ε model is utilised for the turbulence model This study can provide guidance to authorities to comply with fire protection regulations, firefighting and prevention, and emergency response measures, which can ensure the safety of the factory operating in particular and neighbouring factories in the industrial area in general

Liquefied Petroleum Gas (LPG) mainly composed of propane and other relatively – short hydrocarbon (Sarker et al., 2022) undoubtedly plays a pivotal role in the national industry, primarily due to its significant economic benefits, including its cleanliness, energy efficiency, and its classification as a new type of fuel (Lyu et al., 2022) However, it is imperative to recognize that alongside these advantages, the potential dangers associated with the storage, transportation, and leakage of LPG must be given serious consideration (Li et al., 2019) LPG tanker is easily affected by many factors such as environmental storage, personnel leading to tanker failure, leakage and explosion accidents LPG storage is consequently crucial for studying the effects of LPG leakage and planning for disaster investigations using CFD The consequence of LPG leakage accidents happens quickly and violently in the process industry, leading to experimental investigation still many limits and extremely complex due to the large – size location and large – scale equipment (Baalisampang et al., 2017) In order to address the accidental leakage, researchers utilize mathematical modelling and simulation software as references to predict the potential consequences of LPG

Numerous studies have focused on tank failures and the effects of storage tanks at various levels Wang et al (2020) investigated the fireball characteristics of an LPG tanker after explosion by using FDS software This research indicated that a fuel mass of 10,000 kg resulted in the maximum peak radiation flux of the fireball Yu