# Safety and Flammability Analysis for Fuel–Air–Diluent Mixtures Plant: Safety and Flammability Analysis

## Introduction

A series of explosions blamed on propylene leaks from underground pipelines killed 32 people and injured more than 300in the southern Taiwanese city of Kaohsiung in 2014. The explosions resulted in a series of major fires and significantly damaged property and roads. Flammability limits are key characteristics applied to determine the fire and explosion (F&E) dangers of gases and vapors. The lowest and highest concentrations capable of sustaining the propagating flame are the lower (LFL) and upper flammability limits (UFL), respectively (Mannan, 2005; Crowl and Louvar, 2011). The flammability limits of a given fuel depend on several factors, including initial temperature, initial pressure, system scale and configuration, ignition energy, flame stretch, radiation reabsorption, the presence of inert gases, and others (Mannan, 2005; Ju et al„ 2001). In industrial processes, an inert gas is often added to combustible mixtures to prevent the F&E hazard, especially in oxidation processes. Nonetheless, much more flammability limit data is published for pure fuel substances than for fuel/inert gas mixtures. Deriving such data in Taiwan is costly. Therefore, a process to estimate the flammability limits of fuel/inert gas mixtures would provide practical benefits.

## Understanding the Flammability of Inert Gas Mixtures

Kondo et al. (2006a, 2006b, 2007) modified Le Chatelier’s formula to provide an empirical equation for evaluating the flammability boundaries of several fuels mixed with inert gases. The parameters of the empirical equation depend on the tested flammable and inert gases and must be experimentally determined. Theories relating flammability limits to heat loss predict a minimum flame temperature, below which the flame cannot propagate (Spalding, 1957; Buckmaster, 1976; Joulin and Clavin,

1976). When estimating the flammability limits of fuels mixed with inert gases, investigators often assume the adiabatic flame temperature is constant for a particular fuel (Shebeko et al„ 2002; Vidal et ah, 2006; Melhem, 1997; Hansen and Crowl, 2010). When using Melhem’s (1997) method or Hansen and Crowl’s (2010) equations to estimate the flammability boundaries, the adiabatic flame temperatures at the LFL and UFL are required; unfortunately, these temperatures are mostly unknown. A theoretical linear equation was derived describing the flammability boundaries of fuels mixed with inert gases (Chen et ah, 2009a, 2009b). The slope of this linear equation must be determined by regressing experimental data (Chen et ah, 2009a, 2009b). Recently, a model describing the flammability limits of fuel and diluent mixtures by considering thermal radiation loss was validated (Liaw et ah, 2012). The estimated flammability envelopes depend upon the assumed combustion products of the carbon in the fuels, especially at the UFL (Liaw et ah, 2012). Chen et ah (2008) observed carbon monoxide (CO) and carbon dioxide (CO,) in the burned gas at the UFL of pure propane. Ju et ah (2001) and Britton (2002) indicated that the measured flammability limits are different when using different experimental apparatuses. The apparatus used to measure the flammability limits includes vertical glass tubes, spherical glass flasks, and spherical explosion vessels. The vertical glass tube and spherical glass flask are open to the atmosphere after the explosion and are regarded as constant pressure systems. However, the spherical explosion vessel is an intrinsic constant volume system. The methods for the estimation of flammability limits are based on the theory of enthalpy change equaling zero. Thus, they were developed for constant pressure systems. In this work, a model to calculate the flammability envelopes of mixtures containing inert gases for a constant volume system is derived. The combustion products at the LFL and UFL were analyzed to verify the assumption of the model. Methyl formate can be used in a variety of reactions to produce industrial products, such as acetic acid, methyl acetate, and other chemicals (Gerard et al„ 1998). In semiconductor manufacturing, acetone, methanol, and isopropyl alcohol (IPA) are often used in clean room procedures (Liaw and Chiu. 2003). Thus, acetone, methanol, IPA, and methyl formate diluted with either steam or nitrogen (N2) were selected as samples for model validation.

## Experimental Procedures

### Apparatus and Materials

American Chemical Society (ACS) standards (Pharmco Products Inc., USA) were used to verify the purity of IPA (99.8 %).Methanol (99.9 %) and acetone (99.8 %) (Mallinckrodt, USA) met ACS specifications. Methyl formate (97.0 %) was purchased from Alfa Aesar (UK). Water was purified using the Milli-2 Plus filtration purification method from the Millipore Corporation (USA).

### Spherical Explosion Vessel

The flammability measurements were conducted in a 20-lspherical explosion vessel (Adolf Kiihner, Switzerland) standardized to American Society for Testing and

Materials (ASTM) El226-05 (2005). The mixtures tested included acetone + steam, methanol + steam, methyl formate + steam, IPA + steam, IPA + nitrogen, and acetone + nitrogen. Figure Ю. 1 shows the system configuration. The hollow sphere is made of stainless steel. A permanent spark w'ith ignition energy of about 10 J is located in the center of the sphere. The KSEP 332, a unit delivered with the sphere, controls the ignition system and measures the pressure w'ith piezoelectric sensors. Thermal insulation covers the test chamber to reduce energy loss. The measurements were conducted based on American Society for Testing and Materials (ASTM) E681 (1994) and American Society for Testing and Materials (ASTM) E918-83 (2005) standards. A pressure of 1 atm and a temperature of 150°C, w'hich is greater than the normal boiling points of the studied samples, were selected as the initial pressure and temperature settings. Flammability was defined as a pressure increase of 7% or more than the initial pressure in the vessel (ASTM E918-83). The vapor phase composition of the fuel and steam was determined from its mass using a digital scale (EL-410D: sensitivity 0.001 g, Setra Systems, USA), and the sample was stirred for 10 min before being introduced to the vessel. The liquid sample was injected using a syringe. The composition of the nitrogen w'as determined using a partial pressure procedure (pressure sensor sensitivity = 0.01 torr). The test chamber was heated to the initial temperature (150°C), evacuated, and flushed with air three times before the fuel, or fuel + water, were added. Nitrogen w'as then loaded into the chamber followed by the air.

FIGURE 10.1 The basic system configuration of the 20-1. spherical explosion vessel (Reproduced with permission from Liaw et al. 2016, Copyright © Elsevier).

### Fourier Transform Infrared Spectroscopy

The explosion vessel containing the burned gas was cooled to room temperature after each test. Then the combustion products were collected by Flex Foil grab bags, which are strong and flexible with good stability for hydrogen (H2), CO, C02, and methane (CH4) (SKC, 2015). The products were analyzed w'ithin 24 h using Fourier Transform Infrared Spectroscopy (FTIR; FTIR Spectrum 100, Perkin Elmer, USA) at a wave- number resolution of 0.5 cnv1 on the transmission setting. A background spectrum was used to eliminate the environmental signals of the sample spectrum.

### Theory

#### Mathematical Model

The model for estimating the flammability envelope of mixtures containing inert gas in a constant volume system was derived based on the energy balance equation. The estimations of the flammability boundaries are based on the flame temperatures at the LFL and UFL. These two temperatures were assumed to be constant, except at the upper flammability boundary around the LOC. It u'as assumed that the fuel was completely converted into CO, and steam at the LFL while, at the UFL, it was converted to CO,, CO, steam, and H2. When the mole ratio of the fuel, x, is less than xv, where the fuel to oxygen (02) ratio at the UFL equals the stoichiometric ratio, more fuel converts to C02 at the UFL. Additionally, it was assumed that the flame temperature at the UFL was linear w'ith the quantity of CO. Where

The nitrogen in the air is excluded from the inert expression in Equation (10.1). The effect of nitrogen in the air is expressed in other parameters, including PL in Equation (10.2) and Pv in Equation (10.6). The complete derivation of the model is presented in Appendix A. The model is based on the equations for estimating the flame temperatures, values of xL (the value of Л' at LOC) and xL and the lower and upper flammable boundaries.

At the LFL, the constant flame temperature is calculated by Equation (10.2) while, at the UFL, it is calculated by Equation (10.3):

The value of xL, the value of x at the LOC, is obtained:

The lower flammability boundary, L, is calculated:

The ,v(; value is calculated as:

For 1 - x < 1 - xv, the upper flammability boundary, U, is obtained as: For 1 - .v > 1 - л-,-, the upper flammability boundary is calculated as:

Similar equations for a constant pressure system were derived previously (Liaw et al„ 2012). The differences in the equations between the constant pressure and constant volume systems originate from the different forms of the energy balance equation used. The former is presented in enthalpy change and the latter, in internal energy change. Thus, the terms relevant to the gas constant, R, as presented in Equation (10.2-10.8), are not given for the constant pressure system (Liaw et al„ 2012) but the constant volume system.

##### Estimation Procedure

Equations (10.2, 10.3) were used to calculate the flame temperatures at the LFL and UFL, respectively, and x, and xv were obtained using Equations (10.4, 10.6). Then, the LFL was estimated from l-д = 0 to 1-д- = 1-д^ by Equation (10.5), and Equations (10.7, 10.8) were used to estimate the UFL from l-д: = 0 to 1-д = 1 -xL, and from 1-д = 1-д6. to l-д = l-д^ respectively. Figure 10.2 shows the estimation procedure.