Effect of Experimental Conditions on Measuring ... - ACS Publications

May 11, 2010 - The principal application of autoignition temperature (AIT) is to define the ... It is found that ambient humidity does not affect the ...
0 downloads 0 Views 3MB Size
Ind. Eng. Chem. Res. 2010, 49, 5925–5932

5925

Effect of Experimental Conditions on Measuring Autoignition Temperatures of Liquid Chemicals Chan-Cheng Chen*,† and Yen-Cheng Hsieh‡ Department of Safety, Health and EnVironmental Engineering, National Kaohsiung First UniVersity of Science and Technology, 2 Jhuoyue Road, Nanzih, Kaohsiung City 811, Taiwan, Republic of China, and Department of Occupational Safety and Health, China Medical UniVersity, 91 Hsueh-Shih Road, Taichung 40402, Taiwan, Republic of China

The principal application of autoignition temperature (AIT) is to define the maximum acceptable surface temperature in a particular area, usually for electrical classification purpose. However, although AITs are indispensable information for safely handling flammable liquids, the reported AITs of flammable liquids in different data compilations are very much diverse. Sometimes, the difference in separate data compilations might up to more than 300 K. This article aims to explore the quantitative effects of flask material, ambient temperature, and ambient humidity on the accuracy for measuring AIT via the method of ASTM E659. To effectively analyze these factors simultaneously, the L9(34) orthogonal arrays are used to allocate experiments, and experiments are then carefully conducted in a temperature- and humidity-controlled laboratory chamber. It is found that ambient humidity does not affect the measured AIT of ethanol, but both flask material and ambient temperature are significant factors in measuring AIT of ethanol. An experiment of measuring AIT of ethanol conducted with a flask material of quartz is found to result in a higher AIT value than the one conducted with a flask material of borosilicate by 20 °C. A quadratic relation between the measured AIT of ethanol (y) and the ambient temperature (x) is found, and it is also found that a quadratic polynomial of y ) 3.450 × 10-2 x2 - 1.454x + 3.711 × 102 could properly fit this relation with R2 ) 0.9939. According to aforementioned quadratic relation, the ambient temperature at which the lowest AIT of ethanol appears is about 21 °C. 1. Introduction Autoignition temperature (AIT), also referred to as spontaneous ignition temperature (SIT), is defined as the lowest temperature at which a substance will produce hot-flame ignition in air at atmospheric pressure without the aid of an external energy source such as spark or flame. On the basis of the classical thermal theory of ignition, AIT can be regarded as that temperature to which a combustible mixture must be raised so that the rate of heat evolved by the exothermic oxidation reactions of the system will just overbalance the rate at which heat is lost to the surroundings.19 Obviously, the ability of a substance to spontaneously ignite introduces potential safety hazards for all who handle, transport, and store combustible materials. The principal application of AIT to prevent such hazards is to define the maximum acceptable surface temperature in a particular area, usually for electrical classification purposes, for example, article 500.8 of NFPA 70, also known as the National Electric Code, provides that “Class I equipment shall not have any exposed surface that operates at a temperature in excess of the ignition temperature of the specific gas or vapors”.7 Another common application of AIT is to determine in a hazard risk assessment the possible consequences associated with leakage of flammable chemicals. For example, API Publication 581, which is also known as Risk-based inspection base resource document, requires the AIT of a flammable liquid to determine the possible consequences of an explosion or a fire in case a leakage of flammable chemicals occurs.5 Although AITs of flammable liquids are indispensable information to safely handle flammable liquids, the reported * Corresponding author. Tel: 886-7-6011000 ext. 2311. Fax: 8867-6011061. E-mail: [email protected]. † National Kaohsiung First University of Science and Technology. ‡ China Medical University.

AITs are very much diverse in different data compilation. Sometimes, the difference between these reported AITs might up to even 300 K. As it is shown in Table 1, the AIT of benzoyl chloride was reported to range from 358 K to 873 K in five different authoritative sources.2,13,15,16,25 Table 1 also lists some compounds that the difference in the reported AITs is more than 100 K from separate sources. This diversity is mainly attributed to the fact that the experimental result in measuring AIT depends on many experimental factors, including the test procedure and the apparatus, but it is believed that many reported AITs might be conducted in a totally different apparatus or procedure.17 Most methods for measuring AIT of liquid chemicals introduce the sample into the apparatus container which is preheated to a specific temperature and autoignition is evidenced by the sudden appearance of a flame inside the container and by a sharp rise in the temperature of the gas mixture, but the container shape and container size may vary in different test methods.8,17 It was believed that most AIT values reported in the literature were measured by the now-discontinued procedure of ASTM D2155 method, which used a 200-mL flask as the ignition container.17 However, the now-existing ASTM method (i.e., ASTM E659 method), which is proposed to replace ASTM D2155 method, uses a spherical 500-mL flask instead of a 200-mL one. The ASTM E659 method is proposed to replace ASTM D2155 method because of a higher ratio of heat generation to heat removal in the larger flask and the reduction of catalytic wall effects.6 In Europe, the existing method to measure AIT is DIN 51794 method, Determining the ignition temperature of petroleum products, in which a narrow-neck conical flask having a capacity of 200-mL is used as the ignition container.9 Results of these two existing methods are reported to be not comparable and it

10.1021/ie9020649  2010 American Chemical Society Published on Web 05/11/2010

5926

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010

Table 1. Experimental AITs for Selected Compounds in Different Compilationsa Compound Name

The Hazardous Chemical Database25

DIPPR2

Hazardous Chemicals Desk Reference15

SAX’s Dangerous Properties of Industrial Materials16

IPCS INCHEM13

2-butanone 2,4-dimethylphenol hexadecanoic acid piperazidine 1,3-diisopropylbenzene benzoyl chloride methylhexanone 2-methylnitrobenzene 2,4-dihydroxy-2-methylpentane 1-methyl-2-pyrrolidinone 2-heptanone crotonic acid 1,4-benzenedicarboxylic acid 1,3-benzenedicarboxylic acid phenol isobutyl formate

677 753 513 593 722 873 728 693 698 518 805 773 951 973 878 698

788.71 872 650 728.15 349.82 358.15 464.15 578.15 579 619.15 666.15 669.26 769 769 988.15 593.15

988 593

788.7 805.93 -

778 872 593 470.2 693 579 543 666 669 769 988 -

a

The AIT values were reported in K.

is deemed that ASTM E659 method gives a lower AIT because of the larger volume of ignition container in the ASTM E659 method. Because all existing methods of measuring AIT detect the sudden appearance of a flame inside the ignition container by visual inspection, the accuracy for measuring AIT is greatly limited by human capabilities. For example, it has been pointed out that in some cases autoignition actually begins with a nonluminous or barely luminous reaction, which is difficult to detect by visual inspection. Beside the problem of human capability, some researches also revealed that the measurement of AIT was notorious sensitive to vessel cleanliness, injection rate and uniformity of sample dispersion.1,22 All previous effects make the reported AIT of flammable chemicals very much diverse, even the same test method is adopted. The reproducibility, i.e. the difference between measurements reported by two different laboratories, was announced to be 5% of the temperature in degrees Celsius in the ASTM E659 document.6 In fact, some authors reported that the average error in measuring AIT between two different laboratories was around 30 K.14 Because of aforementioned accuracy and precision problems in experimentally measuring AIT, many theoretical methods were proposed to predict the AITs of pure compounds.1,3,4,8,10,14,20-22,24 These predictive methods could be divided into two categories: (1) the quantitative structureproperty relation approach and (2) the structure group contribution approach; the former used selected molecular descriptors to predict AITs, and the latter considered the contributions of molecular groups in a compound to predict AIT. However, all these predictive methods needed a large reliable data set of AITs to construct their models; therefore, the predictive performances of those models are very much limited due to the uncertainty in reported AITs. In fact, a predictive error up to 70 K is a very common case for those predictive methods. To satisfy industry needs for better data to meet new and more stringent requirements, such as environmental and safety regulations, the DIPPR project sponsored by the American Institute of Chemical Engineering (AIChE) has critically reviewed many data to ensure their consistency and soundness; however, AIT values reported in DIPPR database are still flagged as currently unevaluated.2 To our opinion, this situation might be attributed to the following two reasons. First, there are two now-existing standard test methods to measure AIT, i.e. ASTM E659 and DIN 51794 methods, but it is very common that the results obtained from these two methods are different.

Second, the AIT experiment is very laborious and requires very much patience; consequently, there is very limited literature published recently. It is found in our experience that one test to measure AIT for an unknown compound will last 48-60 h and in most instances the operator must be present. Thus, although research on predicting AIT has been very active in recent years, experimental research, to our knowledge, is very rare in recent years.11,12,26 Thus, a systematic experimental study to understand the uncertainty in measuring AIT of flammable liquids is still indispensable. To make the experimental results of AIT be valuable to the process industries, the first question for the researcher is to choose a suitable test method from the existing methods of ASTM E659 and DIN 51794. Because the AIT measured by the ASTM E659 method was cited by NFPA 70, 497, and 921 and because the AIT measured by the ASTM E659 method usually reported a lower value than the one measured by DIN 51794 method, which will result in safer protection strategies for hazard assessment, the ASTM E659 method is adopted in the present study. In the ASTM E659 method, the effect of the volume of the ignition container on the measured AIT has been clearly documented; however, the effects of ambient temperature and ambient humidity on the measured AIT value are mentioned but not clearly defined. This work aims to understand and quantitatively evaluate these effects on the measured AITs of liquid chemicals. This article is organized as follows. First in section 2, the experimental details, including apparatus, chemicals, and procedures are briefly described. The methodology of experimental design is discussed in section 3. Section 4 presents and discusses the experimental results and main findings. Finally, the conclusions reached in this work are explained in section 5. 2. Experimental Section 2.1. Experimental Apparatus and Procedures. Autoignition temperature measurements were made on the K47000 autoignition apparatus manufactured by the Kohler instrument company. This instrument is designed to meet the test requirements described by the test method of ASTM E659-78(2005), Standard Test Method for Autoignition Temperature of Liquid Chemicals.6 This apparatus consists of a commercial 500-mL borosilicate round-bottom, short-necked boiling flask wrapped with aluminum foil and electrically heated furnace with a cylindrical interior shape to maintain uniform temperature.

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010 Table 2. Information of Chemicals Employed in This Study

a

item

content

chemical name manufacturer CAS no. concentrationa densitya formulab structureb molecular weightb boiling pointb flash pointb upper exp. limitb lower exp. limitb autoignition temperaturea autoignition temperatureb

ethanol Merck 64-17-5 99.9 (% w/w) 0.7895 (g/cm3 at 20 °C) C2H5OH HO-CH2-CH3 46 (mass/mol) 78.3 °C 286 K 19% vol. 4.3% vol. 363 °C 423 °C

Data from reference 5. b Data from reference 4.

Furnace temperatures were monitored at the bottom, side, and neck of the flask with three external thermocouples. A fine Chromel-Alumel thermocouple was used for measuring the gas temperature inside the flask. The furnace provides rapid response and (1 °C stability throughout the operating range of 200-1200 °C. A 500-µL hypodermic syringe with a 15 cm stainless needle was used to inject the sample into the flask. A mirror was mounted above the flask at a 45° angle to see into the flask without having to be directly over it. The sample, approximately 50-250 µL, was injected into the uniformly heated flask containing air at a predetermined temperature. After insertion of the sample, the contents of flask were observed in a dark room for 10 min, or until autoignition occurred. Autoignition was evidenced by the sudden appearance of a flame inside the flask and by a sharp rise in the temperature of the gas mixture. When the mixture exhibited flames at the preset temperature, the next sample was tested at a lower temperature. These procedures were repeated until the lowest temperature at which the mixture exhibited flames was obtained. A hot-air gun was used to purge the product gases after a test was completed and before the next test. To avoid interference from the ambient temperature, 10-min elapsed time is considered to allow time for ambient of thermal equilibrium between trials. The quantity of added sample was then systematically varied to determine the lowest temperature at which the hot-flame ignition occurs, and the lowest internal flask temperature at which hot-flame ignition occurred for a specific liquid sample quantity was taken to be the autoignition temperature (AIT) of the chemical in air at atmospheric pressure. To explore the effect of ambient temperature and humidity on the measured AIT, all the experiments were conducted in a laboratory chamber in which the ambient temperature in the chamber could be controlled to be within (1 °C around the set-point and the relative humidity in the chamber could be set with the accuracy of 1%. 2.2. Materials. In this study, ethanol is chosen to be the testing sample. Ethanol is a widely used chemical in process industries, but as shown in Table 2, the reported AIT for ethanol ranges from 363 to 423 °C in different data compilations. Thus, a critical review on the AIT of this commonly used material is necessary. By the way, in our experience a flask made of borosilicate the AIT of which is assigned by the ASTM E659 method is apt to be softened and will change its shape while conducting an experiment with a compound of which is higher than 800 °C. If this situation occurs, the experiment should be stopped to change to a new flask. Thus, in this work we also want to know whether a flask made of borosilicate can be replaced by the one made of quartz which is one of the few

5927

materials that can combine the chemically inert and high temperature properties required. To this purpose, we must first confirm these two flask materials will give the same results for those compounds the AIT values of which are lower than 800 °C. Ethanol is also suitable for this purpose because the reported AIT of ethanol is about half of the softening temperature. The ethanol used in this study is of chemical grade and is directly obtained from commercial companies with a purity of 99 wt %. The entire chemical data (including manufacturer, CAS number, flammability information, purity, and so on) of ethanol are summarized and listed in Table 2. 3. Experimental Design Plan 3.1. Control Factors and Their Levels. As earlier mentioned, the aim of this study is to explore quantitatively the effect of the flask material, the ambient temperature, and the ambient humidity on the measured AITs of liquid chemicals. Thus, there are three control factors in this study: the flask material, the ambient temperature, and the ambient humidity. As we have mentioned that quartz is one of the few materials that can combine the chemically inert and high-temperature properties required, we consider it to replace the flask material of borosilicate which is required by the ASTM E659 method. Thus, we have two levels for the control factor of flask material, that is, borosilicate and quartz. The levels for the control factor of ambient temperature in this study are selected to be of 20, 30, and 40 °C. These levels of ambient temperature are determined according to the temperature control capability of our laboratory chamber. The levels for the control factor of the humidity are chosen to be the absolute humidity of 0.011, 0.012, and 0.013 in kg water/kg air. Although it is a laboratory custom to use the relative humidity instead of the absolute humidity, we adopt the absolute humidity as the unit of the humidity here. As we know, the relative humidity is defined as the percentage ratio of actual partial pressure of the water vapor present to its equilibrium vapor pressure at the ambient temperature, which obviously depends on the ambient temperature. The choice of the absolute humidity could avoid the interaction effect between the ambient temperature and the relative humidity. The levels of the absolute humidity are decided to be the aforementioned values because of the physical constraint of the saturation of vapor. As Figure 1 shows, in the case of 20 °C, a humidity of 0.013 kg water/kg air is about 88% of the relative humidity which is an extremely wet condition; in the case of 40 °C, a humidity of 0.011 kg water/kg air is about only 25% of the relative humidity. Thus, such a choice of control levels in fact covers a wide range of relative humidity, although it seems to be a very limited range in terms of absolute humidity. We plot the experimental levels for the control factors of the ambient humidity and of the ambient temperature in a psychrometric chart (see Figure 1) to clearly elucidate the physical constraint of the saturation of vapor. In that figure the circle points represent the experimental levels chosen in this work and x symbols denote the levels of confirmation tests which will be explained in the Discussion section. Table 3 summarizes the levels for all three control factors considered in this study. 3.2. Experimental Matrix. Orthogonal arrays have been widely utilized by engineers to explore the effects of several control factors simultaneously. Taguchi and Konishi tabulated many standard orthogonal arrays,23 and one of these arrays could be used according to the particular case study at hand. Selection of a viable standard orthogonal array for a case study depends

5928

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010

Figure 1. Experimental conditions: O - orthogonal experiments; × - confirmation experiments. Table 3. Experimental Control Factors and Their Levels factor name

level 1

A: flask material borosilicate B: environmental temperature (°C) 20 C: environmental humidity (kg/kg) 0.011

level 2

level 3

quartz 30 0.012

(borosilicate) 40 0.013

on the degree of freedom required. A candidate array should always meet the requirement that the number of rows must be at least equal to the degrees of freedom required. In this study there are three control factors, two with three levels and one with two levels. Moreover, we do not intend to consider any interaction effects between these control factors, so the required degree of freedom for this study is (3 - 1) × 2 + (2 - 1) × 1 ) 5. So the L9(34) orthogonal array, which is suggested for an experiment of four 3-level factors, becomes our candidate array. However, as there is one 2-level factor in this study, i.e., the flask material, it is imperative to modify the standard L9(34) orthogonal array. Taguchi and Konishi have suggested a technique (i.e., the dummy level technique) for dealing with the case of mixed levels, at the expense of loss part of the orthogonality. This dummy level technique allows us to assign a factor with fewer levels to a column that has more levels in an orthogonal array and this technique is adopted in this work to assign our two-level factor (i.e., the flask material) into a three-level orthogonal array. The details of the dummy level technique and discussions of the corresponding analysis can be found in Phadke’s book18 and the references therein. In dummy level technique the repeated level is called the dummy level which is decided depends on what more information we need. Because the flask material of borosilicate is recommended by the test method of ASTM E659, the dummy level for the control factor of flask material is assigned to be the material of borosilicate to get more experimental information about it. The experimental layout of this study is summarized in Table 4. In Table 4, the nine rows correspond to nine test runs to be conducted. Columns 1, 2, and 3 are assigned to the control factor of flask material

Table 4. Modified L9 Array and Factor Assignment and Testing Conditions column number 2 B

3 C

4

experiment number

1 A

1 2 3 4 5 6 7 8 9

A1 A1 A1 A2 A2 A2 A′1 A′1 A′1

B1 B2 B3 B1 B2 B3 B1 B2 B3

C1 C2 C3 C2 C3 C1 C3 C1 C2

1 2 3 3 1 2 2 3 1

(A), ambient temperature (B), and ambient humidity (C), respectively. Test runs were then conducted at the specified conditions listed in Table 4; for example, test run 3 was conducted at level 1 for flask material, level 3 for ambient temperature, and level 3 for ambient humidity. Similarly, test run 9 was conducted at level 1 for flask material, level 3 for environmental temperature, and level 2 for ambient humidity. The sequence of the test run was randomly determined to avoid the systematical error, and all the test runs were carefully conducted in our temperatureand humidity-controlled laboratory chamber. 4. Results and Discussion A typical time history of the temperature inside the test flask during an experimental run is shown in Figure 2. In this figure exact magnitudes of the temperature are not intended to be necessarily significant as the recorder is set to be of different scaling factors in different temperature ranges. The initial dip on the curve shown in the figure is caused by cooling due to vaporization of the sample. The occurrence of an autoignition was evidenced by the sudden appearance of a flame inside the flask and by a sharp rise in the temperature of the gas mixture. When the mixture exhibited flames at the preset temperature, the next sample

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010

5929

of the same quantity is tested at a lower temperature. These procedures were repeated until the lowest temperature at which the sample of a given quantity exhibited flame was obtained. Such a series of tests was represented by those points on the same vertical line shown in any of the plots in Figure 3. In any plot, a circle is used to represent the flammable case, and an x is used to represent the nonflammable case. Then, different sample quantities are employed until the amount giving the lowest temperature of autoignition is obtained. The autoignition temperature of a compound corresponds to this lowest temperature and is designated as a triangle in the plots. Figure 3 shows experimental results for all the nine test runs described in Table 4. Such experiments are carefully conducted in our laboratory chamber according to the prescribed conditions listed in Table 4, and the results are summarized in Table 5. To construct the analysis-of-variance (ANOVA) table for this experiment design, the total sum of the squares for the experimental matrix is calculated by

ST2 ) Σy2i Figure 2. Time history of a typical autoginition experiment.

(Σyi) 2 N

Figure 3. Autoginition experimental results for the whole experiment matrix: O - flammable, × - nonflammable, 4 - the AIT point.

(1)

5930

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010

Table 5. Testing Conditions, Results, and Data Analysis of the Orthogonal Matrixa

where N is the total number of observations, and the sums of squares (Si2) for the two 3-level factors, i.e., SB2 and SC2 are calculated by (sum of level 1)2 + ) no. of observations in level 1 (sum of level 2)2 + no. of observations in level 2 (sum of level 3)2 (grand total)2 (2) no. of observations in level 3 total no. of observations

S2i

The sum of squares for the 2-level factor, i.e., the flask material (SA2 ), is calculated as (sum of level 1 + sum of level 3)2 + sum of no. of observations in level 1 and level 3 2 2 (sum of level 2) (grand total) (3) no. of observations in level 2 total no. of observations

S2i )

The sum of squares for the error is then estimated by the following equation: s2e ) sT2 - sA2 - s2B - s2C

(4)

The ANOVA table is then constructed and summarized in Table 6. As it was shown by this ANOVA table, the effect of ambient humidity on the AIT is less than that of random error with 95% confidence level, so it was concluded that variation

Table 7. Experimental Conditions for the Confirmation Tests test no. flask material temperature (°C) humidity (kg/kg) AIT (°C) 1 2

borosilicate borosilicate

30 15

0.022 0.0076

356.7 357.0

in ambient humidity does not affect the result of measuring AIT. However, due to the saturation constraint of water vapor, the variation of the control factor of the humidity, which ranges from 0.011 to 0.013 in kg water/kg air, is narrow in our original experimental matrix, so a confirmation test of humidity of 0.022 kg water/kg air at 30 °C with flask material of borosilicate is conducted to reconfirm the effect of ambient humidity on measuring AIT. Experimental conditions and results for this confirmation test are listed and summarized as the first row in Table 7. In the original experiment array there are two test runs (i.e., test runs #2 and #8) conducted at the same temperature and flask material as those of the confirmation test. The average AIT of these two tests is (360.7 + 358.1)/2 ) 359.4 °C, and the reported AIT of the confirmation test is 356.7 °C. As we know that the 100(1 - R)% prediction interval on a single observation from a normal distribution is given by jx t(R)/(2),n-1s√1 +(1)/(n) < x < jx + t(R)/(2),n-1s√1 +(1)/(n), where s is the sample variance, t(R)/(2),n-1 is the t value of the Student’s t distribution with (n - 1) degrees of freedom at 100(1 - R)% confidence level. In our case, the original experimental array has two tests in this condition, so we have n ) 2. Thus, we

Table 6. ANOVA Table for the Control Factors source

sum of squares

degrees of freedom

mean squares

F-value

F(0.05)

F(0.01)

P-value

remarka

flask temperature humidity error

761.80 177.91 9.85 5.10

1 2 2 4

761.80 88.95 4.92 1.27

597.82 69.81 3.86

7.71 6.94 6.94

21.20 18.00 18.00

1.66 × 10-5 7.76 × 10-3 0.117

** **

a

A single asterisk (*) denotes significance with a 95% level of confidence, whereas the appearance of two asterisks (**) denotes significance with a 99% level of confidence.

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010

could obtain jx ) 359.4, s ) 1.8385, and t0.025, 1 ) 12.706. So the prediction interval of the AIT is 359 ( 31.4 °C, and it is obvious that the AIT of the confirmation test (i.e., 356.7 °C) belongs to this interval. So these two values make no difference at 95% confidence level. This behavior could be possibly explained as follows. The AIT is regarded as the temperature to which a combustible mixture must be raised so that the rate of heat evolved by the exothermic oxidation reactions of the system will just overbalance the rate at which heat is lost to the surroundings. However, in the ASTM E659 method, a flask with opening to the ambient environment is used as the ignition container; thus, besides the heat loss to the electrically heated furnace, the heat loss to the ambient environment should also be considered. In the aspect of heat loss, there are three mechanisms that heat could release to the electrical heated furnace and ambient environment: conduction, convection, and radiation. However, only the former two mechanisms are important to a combustion process at low temperature, which is the case of autoignition. Because the rate law of a combustion process is expressed as a power law of the fuel, it is straightforward to know that the rate of the heat generated in a combustion process is not related to the ambient humidity. Furthermore, it is obvious that the conductive heat loss depends on the conductivities of the ignition container (loss to the electrically heated furnace) and of the air (loss to the ambient environment). Because these two conductivities are independent of the ambient humidity, it is clear that heat loss through this mechanism is irrelevant to the ambient humidity. Consider the convective heat loss to the ambient environment. Obviously, heat loss through this mechanism depends on the heat transfer coefficient of the ambient air and temperature difference between the ignition container and the ambient environment. As the ambient humidity has very limited effect on the heat transfer coefficient, the heat loss through the convective mechanism is not affected by the ambient humidity. Thus, experimental results show that the ambient humidity does not affect the measured AIT. It is also shown from Table 6 that flask material was found to be a highly significant factor for measuring AIT because the p value of such a hypothesis test is 1.66 × 10-5. The average AIT for those three runs with flask material of quartz is 380.6 °C, and for the six runs with flask material of borosilicate it is 361.1 °C. This result indicates that an AIT experiment conducted with a flask of quartz will possibly result in the measured AIT higher than that of a flask of borosilicate by 20 °C. Thus, our attempt to replace a flask of borosilicate by the one of quartz should not be further considered. A possible explanation for this observation could be also illustrated through the classic thermal theory of autoignition. It is known that the conductivity of quartz is higher than that of the borosilicate. For example, at 100 °C, the thermal conductivities are reported to be 1.38 W/m · K and 1.21 W/m · K for quartz and borosilicate, respectively. Thus, an ignition container made of quartz will result in more heat loss to the electrically heated furnace than the one made of borosilicate. Therefore, to compensate for more heat loss, the autoignition in an ignition container made of quartz must occur in a higher temperature, because the higher reaction temperature will increase the reaction rate, which enhances the rate of heat generated. The ANOVA table also shows that ambient temperature is a significant factor in experimentally measuring AIT of a compound. The average AITs for those experiments conducted at the ambient temperatures of 20, 30, and 40 °C are found to be 363.1, 366.1, and 373.1 °C, respectively. However, the differ-

5931

Figure 4. Relation between observed AIT and ambient temperature: O average AIT of the experimental matrix; × - AIT of the confirmation test. The regression curve is y ) 3.450 × 10-2 x2 - 1.454x + 3.711 × 102.

ence in AIT due to the change in ambient temperature is not as obvious as that due to the change in flask material. To get more evidence for this argument of ambient temperature, we conducted a test at the ambient temperature of 15 °C, and the second row in Table 7 lists experimental details for this confirmation test. As shown in Table 7, the confirmation test was conducted with flask material of borosilicate, so the test runs conducted with flask material of quartz in the original experimental array should be excluded while assessing the effect of ambient temperature in measuring AIT. After excluding the test runs conducted with flask material of quartz, the trimmed average AITs for those experiments conducted at the ambient temperatures of 20, 30, and 40 °C are found to be 355.8, 359.4, and 368.2 °C, respectively. The trimmed averaged AIT values and the AIT value from this confirmation test are then plotted against the ambient temperature as shown in Figure 4. It is found from Figure 4 that there is a quadratic relation between the reported AIT and the ambient temperature. A quadratic polynomial of y ) 3.450 × 10-2 x2 - 1.454x + 3.711 × 102 could fit these experimental data very well, and the value of R2 is up to 0.9939 in this fitting, where y is the AIT with unit of °C and x is the ambient temperature with unit of °C, too. According to previous quadratic relation, the ambient temperature at which the lowest AIT of ethanol is reported is found to be about 21 °C. The reason why such a quadratic relation between the AIT and ambient temperature holds for ethanol is still not clear to us. However, to make sure this behavior is experimentally repeatable, we have conducted other experiments to be confident about this behavior. New experimental results are shown in Figure 5, in which the ambient temperatures of 25 and 35 °C are newly added experimental conditions. It could be found from Figure 5 that this quadratic behavior is repeated in these new experiments. It should be noted that, although this behavior is repeatable for ethanol, it does not mean that such a quadratic relation holds for other compounds. It needs more observations in future work to make sure that such a behavior holds for compounds other than ethanol.

5932

Ind. Eng. Chem. Res., Vol. 49, No. 12, 2010

Figure 5. New experiments to confirm the quadratic relation between observed AIT and ambient temperature for ethanol: the ambient temperature of 25 and 35 °C are newly added test conditions.

5. Conclusions In this work, we explored effects of flask material, ambient temperature, and ambient humidity on the measured AIT of ethanol using the ASTM E659 method. Experiments are conducted according to a modified L9(34) orthogonal array. It is found that the measured AIT of ethanol conducted with a flask made of quartz will result in a value higher than the one conducted with a flask made of borosilicate by 20 °C. The ambient humidity is found not to be a significant factor in measuring AIT with 95% level of confidence. To be sure of this argument, an extra confirmation test is conducted. Through a hypothesis test of Student’s t with 95% confidence level, this extra test reaches the same conclusion once again. It is also found that the ambient temperature is a significant factor in measuring AIT with 95% confidence level. The measured AIT of ethanol are found to be a quadratic function of the ambient temperature, and a quadratic polynomial of y ) 3.450 × 10-2 x2 - 1.454x + 3.711 × 102 could fit the experimental data with R2 ) 0.9939. According to this quadratic relation, the ambient temperature at which the lowest AIT of ethanol will be measured is found to be about 21 °C. Acknowledgment We thank the National Science Council of the ROC for supporting this study financially under Grants NSC 97-2221E-039-006 and NSC 98-2221-E-327-045-MY3. Literature Cited (1) Affens, W. A.; Johnson, J. E.; Carhart, H. W. Effect of Chemical Structure on Spontaneous Ignition of Hydrocarbons. J. Chem. Eng. Data 1961, 6 (4), 613–619. (2) AIChE, DIPPRO, DIPPR Project 801 Pure Component Data (1996) public version.

(3) Albahri, T. A. Flammability characteristics of pure hydrocarbons. Chem. Eng. Sci. 2003, 58, 3629–3641. (4) Albahri, T. A.; George, R. S. Artificial Neural network Investigation of the Structural Group Contribution Method for Predicting Pure Components Auto Ignition Temperature. Ind. Eng. Chem. Res. 2003, 42, 5708– 5714. (5) API Publication 581: Risk-Based Inspection Base Resource Document; American Petroleum Institute: Englewood, CO, 2000; at www.api.org/ publications. (6) ASTM E659-78: Standard Test Method for Autoignition Temperature of Liquid ChemicalsASTM International: West Conshohocken, PA, 2005; DOI: 10.1520/E0659-78R05. (7) Britton, L. G.; Cashdollar, K. L.; Fenlon, W; Frurip, D.; Going, J; Harrison, B. K.; Niemeier, J.; Ural, E. A. The Role of ASTM E27 Methods in Hazard Assessment Part II: Flammability and Ignitability. Process Saf. Progr. 2005, 24 (1), 12–28. (8) Chen, C. C.; Liaw, H. J.; Kuo, Y. Y. Prediction of Autoignition Temperatures of Organic Compounds by the Structural Group Contribution Approach. J. Hazard. Mater. 2009, 162, 746–762. (9) Determining the Ignition Temperature of Petroleum Products. DINSprachendienst. DIN 51794; 2003. (10) Egolf, L. M.; Jurs, P. C. Estimation of Autoignition Temperatures of Hydrocarbons, Alcohols, and Ester from Molecular Structure. Ind. Eng. Chem. Res. 1992, 31, 1798–1807. (11) Hshieh, F. Y.; Stoltzfus, J. M.; Beeson, H. D. Autoignition Temperature of Selected Polymer at Elevated Oxygen Pressure and Their Heat of Combustion. Fire Mater. 1996, 20, 301–303. (12) Hshieh, F. Y.; Hirsch, D. B.; Williams, J. H. Autoignition Temperature of Trichlorosilanes. Fire Mater. 2002, 26, 289–290. (13) IPCS INCHEM. http://www.inchem.org/pages/icsc.html. (14) Kim, Y. S.; Lee, S. K; Kim, J. H.; Kim, J. S.; No, K. T. Predictions of Autoignitions (AITS) for Hydrocarbons and Compounds Containing Heteroatoms by the Quantitative Structure-Property Relationship. R. Soc. Chem. 2002, 2, 2087–2092. (15) Lewis, R. J. Hazardous Chemicals Desk Reference, 5th ed.; WileyInterscience: New York, 2002. (16) Lewis, R. J. SAX’S Dangerous Properties of Industrial Materials, 11th ed.; John Wiley & Sons: New York, 2004. (17) Mitchel, B. E.; Jurs, P. C. Prediction of Autoignition Temperatures of Organic Compounds from Molecular Structure. J. Chem. Inf. Comput. Sci. 1997, 37, 538–547. (18) Phadke, M. S. Quality Engineering Using Robust Design; Prentice Hall: New Jersey, 1989. (19) Semenov, A. N. N. Some Problems of Chemical Kinetics and ReactiVity; Pergamon Press: Elmsford, NY, 1959; Vol. 2. (20) Susuki, T. Quantitative Structure-Property Relationships for Autoignition Temperatures of organic compounds. Fire Mater. 1994) , 18, 81– 88. (21) Susuki, T.; Ohtaguchi, K.; Koide, K. Correlation and Prediction of Autoignition Temperatures of Hydrocarbons Using Molecular Properties. J. Chem. Eng. Jpn. 1992, 25, 606–608. (22) Swarts, D. E.; Orchin, M. Spontaneous Ignition Temperature of Hydrocarbons. Ind. Eng. Chem. 1957, 49 (3), 432–436. (23) Taguchi, G. Konishi, S. Orthogonal Arrays and Linear Graphs; ASI Press: Dearborn, MI, 1987. (24) Tetteh, J.; Metcalfe, E.; Howells, S. L. Optimisation of radial basis and backpropagation neural networks for modeling auto-ignition temperature by quantitative structure-property relation. Chemom. Intell. Lab. Syst. 1996, 32, 177–191. (25) The Hazardous Chemical Database. http://ull.chemistry.uakron.edu/ erd/index.html. (26) Zalosh, R.; Casey, J. Autoignition Temperature Data and Scaling for Amide Solvents. J. Loss PreV. Process ind. 2009, 22, 1–6.

ReceiVed for reView December 29, 2009 ReVised manuscript receiVed April 5, 2010 Accepted April 14, 2010 IE9020649