TG-DSC and TG-FTIR analysis of heavy fuel oil and vacuum residual oil pyrolysis and combustion: characterization, kinetics, and evolved gas analysis

Abstract

Residual oils, high viscosity and large sulfur content petroleum products from the refining process of crude oil, are receiving increasing interest in pre-combustion carbon capture applications. Gasification is a promising technology to convert such complicated hydrocarbons into syngas. Pyrolysis and combustion are very important stages in the gasification process, and therefore a better understanding of these processes leads to higher efficiency and better development of such applications. In this work, pyrolysis and combustion of heavy fuel oil (HFO) and vacuum residual oil (VRO) were studied in a thermogravimetric analyzer (TGA). The HFO studied in this work is a blend of VRO and diesel, which provides insight into the performance of residual oils/diesel blends. The TGA experiments were conducted using nitrogen and mixtures of oxygen and nitrogen for pyrolysis and combustion studies, respectively, at different heating rates (5–20 °C min⁻¹). The oxygen concentration was varied from 0 to 71.4%_vol. to replicate oxygen concentration in applications ranging from pyrolysis (0% O₂) to combustion (21% O₂) and gasification (high O₂%). The TGA experiments covered a temperature range from ambient to 1200 °C. The results show that pyrolysis is slightly slower than combustion at low temperatures for both oils. However, pyrolysis is significantly faster at high temperatures. The combustion of both oils resulted in minimal residue, while the residue remaining in the pyrolysis is 10–19%. The TGA was coupled with Fourier transform infrared spectroscopy (FTIR) to monitor the evolved volatiles from the pyrolysis and combustion processes. The results show more aromatics evolved from VRO than HFO. Apparent kinetic parameters were calculated using three model-free methods and a model-based method (Coats and Redfern).

Appendices

Appendix A

Elemental analysis

The elemental (CHNS) analysis of the HFO and VRO was performed in a Thermo Flash 2000 (Thermo Fisher Scientific, USA) organic elemental analyzer. In a typical analysis, 2.5–3 mg of sample was used. The combustion tube present in the analyzer was heated to around 1000 °C in the presence of air. To determine carbon, hydrogen, nitrogen, and sulfur content in the sample, the elements were converted to their respective oxides. Gas chromatography/thermal conductivity detector (GC/TCD) was used to separate and detect these oxides evolving from the combustion of the samples. The GC/TCD is pre-calibrated for the determination of carbon, hydrogen, nitrogen, and sulfur content. It can be noted that the nitrogen oxides were reduced to N₂ via Cu-based catalyst before the identification in the GC/TCD. For the determination of oxygen, a different experiment was conducted on the same equipment using a nickel carbon catalyst in the combustion tube. The oxygen present in the evolved vapors is converted to carbon monoxide in the presence of the catalyst. The GC/TCD is pre-calibrated to determine the oxygen based on the amount of carbon monoxide.

SARA methodology

SARA fractionation was achieved by following a modified ASTM D4124 method. The ASTM D4124 is modified by replacing the chromatography with DCVC to allow the separation of larger sample loads, compared with the ASTM method, achieving a fast separation with low solvent consumption and without compromising the resolution between fractions. First, around 10 g of each oil was treated with 400 mL of n-heptane in a reflux setup at 98˚C for three hours. The collected mixture was kept to settle for 24 h, and then solids were separated by filtration (Whatman grade 42) and washed with additional n-heptane until the eluent was clear. The solids were recovered by hot toluene and rota-evaporated to yield asphaltenes, which were quantified gravimetrically. The supernatant, maltenes, was purified by removing the excess solvent and then subjected to dry column vacuum chromatography (DCVC) using a glass column (Chemglass 41 × 305 mm) packed with alumina (Mesh 80–200) as a sorbent to yield individual saturates, aromatics, and resins fractions. All the fractions were quantified gravimetrically once the excess solvent was removed by rota-evaporation. The same eluotropic series of the ASTM method was used. The overall recovery for HFO and VRO was found to be around 98%. Figure 

Fig. 20

SARA fractionation

20 illustrates the SARA procedure and shows a photograph of the fractions collected.

Kinetic analysis

The apparent activation energy changes with the extent of the reaction owing to the multi-step mechanism assumed by the oils during pyrolysis and combustion. The variation of apparent activation energy with the conversion can be given by isoconversional methods. The use of isoconversional models results in the deconvolution of the complex procedure to perform kinetic analysis of these multi-step processes of these feedstock. The utilization of these models assumes that the thermal decomposition of heavy oils occurs as follows:

$$HFO/VRO \to Volatiles\left( g \right) + Residue\left( s \right)$$

(A1)

The conversion of oil can be described by a general equation expressed as:

$$\frac{\text{d}\alpha }{{\text{d}T}} = \frac{A}{\beta }{\text{e}}^{{\frac{{ {\text{- E}}_{{\text{a}}} }}{\text{RT}}}} f\left( \alpha \right)$$

(A2)

T,$\beta$, E_a, R, A, and f(α) represent temperature (in K), heating rate (in °C min⁻¹ or K min⁻¹), apparent activation energy (in kJ mol⁻¹), universal gas constant (in J mol⁻¹ K⁻¹), pre-exponential factor (min⁻¹), and functional dependence of rate on the extent of conversion, respectively. Here α is the extent of conversion normalized with respect to the residual mass of the sample observed in the TGA and is given as follows:

$$\alpha = \frac{{m_{{\text{i}}} - m_{{\text{t}}} }}{{m_{{\text{i}}} - m_{\infty } }}$$

(A3)

m_i, m_t, and m_∞ represent the initial sample mass, mass of the sample at any instant ‘t,’ and steady-state mass (residual mass). Rearranging and integrating Eq. (A2) results in the following expression:

$$g\left( \alpha \right) = \int_{0}^{a} \frac{\text{d}\alpha }{{f\left( \alpha \right)}} = \frac{A}{\beta } \int_{0}^{T} {\text{e}}^{{\frac{{ \text{- E}_{a} }}{\text{RT}}}} dT$$

(A4)

It is assumed for the isoconversional models that the reaction rate for a particular conversion depends only on temperature. The multi-step mechanism involved in the thermal decomposition of organics is identified by the variation of apparent activation energy with conversion obtained from various models. Taking natural logarithm both sides of Eq. (A2) and rearranging gives:

$${\text{ln}}\beta \left( {\frac{\text{d}\alpha }{{\text{d}T}}} \right) = {\text{ln}} \left[ {Af\left( \alpha \right)} \right] - \frac{{E_{\text{a}} }}{RT}$$

(A5)

Equation (A4) is known as Freidman method, which is a differential isoconversional method [67]. Thermogravimetric mass loss data were utilized for evaluation of $\frac{\text{d}\alpha }{\rm{d}T}$. For the same conversion across different heating rates, $\mathrm{ln}\beta \left(\frac{\rm{d}\alpha }{\text{d}T}\right)$ vs 1/T data were plotted. To the plotted co-ordinates, straight lines were fitted, and the slopes of these lines $\left(\frac{-{E}_{\mathrm{a}}}{R}\right)$ were utilized to determine the apparent activation energies for different conversions. Multi-step mechanism involved during pyrolysis and combustion is understood by variation of apparent activation energy with conversion.

It can be noted that the TG analysis results in integral data, and differentiating integral data magnifies the noise. So, there can be errors in the derivative of conversion with respect to temperature. Hence the choice of integral methods is justified over differential methods.

Substituting E_a/RT as x in Eq. (A4), which is also known as temperature integral, the following expression is obtained:

$$g\left( \alpha \right) = \frac{{AE_{a} }}{\beta R} \int_{x}^{\infty } \frac{e}{{x^{2} }}^{ \text{- x}} dx = \frac{{AE_{\text{a}} }}{\beta R} p\left( x \right)$$

(A6)

Different temperature integrals result from the different mechanisms involved during the thermal decomposition of oils. Flynn–Wall–Ozawa (FWO) method assumes the change in the apparent activation energy to be constant during the thermal decomposition and employs Doyle’s approximation [68] in order to have a solution of the temperature integral.

$$\ln \left( {p\left( x \right)} \right) = - 2.315 - 0.4567x {\text{for}}20 \le x \le 60$$

(A7)

Taking natural logarithm both sides and rearranging Eq. (A6), the mathematical form of FWO method is obtained [69]:

$${\text{ln}}\beta = {\text{ln}}\left( {\frac{{AE_{a} }}{Rg\left( \alpha \right)}} \right) - 5.33 - 1.052\frac{{E_{\text{a}} }}{RT}$$

(A8)

At different heating rates, for the same conversion, ${\text{ln}} \beta$ vs. 1/T data were plotted. Different straight lines were plotted for multiple conversions, and $\frac{-1.052{E}_{\text{a}}}{R}$ was used to determine the variation of apparent activation energy with conversion. Applying one more Doyle’s approximation for having the solution of the temperature integral, Eqs. (A9) and (A10) represent the approximation used and the mathematical form of the Kissinger–Akahira–Sunose (KAS) method, respectively [70].

$$\ln \left( {p\left( x \right)} \right) = \frac{e}{{x^{2} }}^{\text{ - x}} {\text{for }}20{ } \le { }x \le { }50{ }$$

(A9)

$$\ln \frac{\beta }{{T^{2} }} = \ln \left( {\frac{AR}{{E_{{\text{a}}} g\left( \alpha \right)}}} \right) - \frac{{E_{{\text{a}}} }}{RT}$$

(A10)

The slope of the straight line $\left(\frac{-{E}_{\mathrm{a}}}{R}\right)$ fitted to the $\mathrm{ln}\frac{\beta }{{T}^{2}}$ vs 1/T obtained for the same conversion at different heating rates is used to determine the apparent activation energy at a particular conversion. Following to this, the approximation proposed by Starinks [48, 71] to Eq. (A10) allows more accurate estimation of apparent activation energy. The mathematical expression of Starink’s method is as follows:

$$\ln \frac{\beta }{{T^{1.92} }} = \ln \left( {\frac{{AR^{0.92} }}{{E_{{\text{a}}}^{0.92} g\left( \alpha \right)}}} \right) - 0.312 - 1.0008\frac{{E_{{\text{a}}} }}{RT}$$

(A11)

The procedure of the determination of apparent activation energies is same as for the previous methods, i.e., slope of the straight lines plotted for multiple conversions was utilized.

For the oxygen-rich case, a model-fitting method was also used along with model-free methods. The apparent kinetics of thermal decomposition of VRO was studied using Coats and Redfern, and the mathematical expression is as follows [48]:

$$\ln \left( {\frac{g\left( \alpha \right)}{{T^{2} }}} \right) = \ln \left( {\frac{AR}{{\beta E_{{\text{a}}} }}\left( {1 - \frac{2RT}{{E_{{\text{a}}} }}} \right)} \right) - \frac{{E_{{\text{a}}} }}{RT}$$

(A12)

It can be noted that this method was not used for the evaluation of apparent activation energies of thermal decomposition of both the oils under different cases which are studied. This model was used for the oxygen-rich case where the model-free methods are inapplicable to determine the apparent activation energies during the oxidation of both the oils. The limitations associated are as follows: First, this method, owing to the assumptions, may not describe the real thermal decomposition process [72]. Second, the method is not capable of describing the reactions with low activation energies associated with low temperatures (< 200 °C) [48, 72]. Due to the simplicity associated with the model for the evaluation of apparent kinetic parameters utilizing single heating rates, it is widely used in recent studies involving TGA of various biomasses under various atmospheres [73,74,75]. In Equation (A12), $g(\alpha )$, functional dependence on conversion, depends on the reaction mechanism assumed by the oil during thermal decomposition.