Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil WAHAB MAQBOOL Master of Engineering in Chemical Engineering Submitted in fulfilment of the requirement for the degree of Doctor of Philosophy School of Chemistry, Physics and Mechanical Engineering Science & Engineering Faculty Queensland University of Technology 2019
160
Embed
Supercritical Carbon Dioxide Extraction and Fractionation ...Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil iii techno-economic study established that increasing
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil
WAHAB MAQBOOL
Master of Engineering in Chemical Engineering
Submitted in fulfilment of the requirement for the degree of
D o c t o r o f P h i l o s o p h y
School of Chemistry, Physics and Mechanical Engineering
Science & Engineering Faculty
Q u e en sl an d U n iv er s i t y of T e ch no lo gy
2019
Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil i
KEYWORDS
Aspen Plus®
Carbon dioxide
Chemical separation
Data regression
Equation of state
Fractionation
Green solvent
Internal rate of return
Net present value
Peng-Robinson
Phase equilibrium
Pilot plant
Process optimization
Process simulation
Process utilities
Renewable chemicals
Supercritical fluid extraction
Techno-economics
Vapour-liquid
ii Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil
ABSTRACT
Bio-oil produced by the thermochemical treatment of lignocellulosic biomass is a
complex liquid mixture of compounds which in its crude state has a relatively low value.
Its typically large aqueous component has further cemented the reputation of bio-oil as a
challenging candidate source of renewable chemicals. In this PhD study a supercritical
fluid extraction (SFE) process using carbon dioxide as a solvent was developed and
investigated as a potentially energy efficient, cost effective alternative to conventional
distillation for the extraction and subsequent fractionation of high value target
compounds from bio-crude. A bio-oil is more commonly known as bio-crude when it is
produced from hydrothermal liquefaction (HTL) process.
To date, SFE has been used commercially for some niche applications such as
decaffeination or the recovery of essential oils and bioactive compounds from plant
derived material. The use of SFE for extraction of bio-oils has been the subject of a limited
number of experimental studies. Although basic vapour-liquid equilibrium (VLE) data
was available prior to the current PhD study for some potential target bio-oil compounds,
no previous attempt had been made to develop and implement the necessary VLE models
required for rigorous process investigation, optimisation and design. There have been no
reports in the literature on the techno-economics of SFE as a means of extracting high
value compounds from bio-oil.
Solubility data for a key (exemplar) target compound was experimentally determined
using both synthetic (no-sampling) and analytic-gravimetric (sampling) solubility cell
methods to appropriately extend the pressure and temperature ranges of data previously
reported in the literature and to develop the phase equilibrium models necessary for
process simulation. The model developed for binary VLE data from the literature
(validated against the original sources of data and bench scale measurements from the
current PhD study) was implemented on the Aspen Plus® process simulation platform
using a Peng-Robinson-Boston-Mathias (PR-BM) property method. The model
predictions for stage-wise pressure reduction fractionation of bio-crude components
using supercritical carbon dioxide as the solvent were successfully validated with a series
of pilot plant trials. A raw bio-crude produced by the HTL of bagasse derived black liquor
was used as feedstock in the SFE pilot plant trials. The pilot plant validation trials also
established the accuracy of utilising multiple binary VLE models to predict the
fractionation of real bio-crude solubilised in a scCO2 column extract stream of known
composition (a key simplifying assumption necessary in the development of the larger
process model).
The validated Aspen Plus® model was subsequently used to undertake a techno-
economic study of SFE using carbon dioxide. Solvent/ bio-oil (S/B) ratio is one of the key
determinants in the economics of any SFE process. Although extraction and fractionation
efficiencies increase with increasing S/B ratios, so too do the associated costs. The
Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil iii
techno-economic study established that increasing the S/B ratio from 6.2 to 12.4 and
20.2, will decrease the corresponding Internal Rate of Return (IRR) from 15% to 12.3%
and 9.5% respectively. The corresponding increase in operational costs were 9.1%
(S/B = 12.4) and 17.4% (S/B = 20.2) relative to that of for the base case (S/B = 6).
The economics of SFE and conventional distillation processes for the recovery of target
compounds from bio-crude, were compared. For a base case plant capacity of 22.8
tonne/hr of biocrude, an IRR value of approximately 15% was achieved for SFE two-stage
(P-1), SFE single stage (P-2) and distillation combined with multistage evaporation (P-4)
scenarios. For the distillation alone scenario (P-3) the IRR value at the base case plant
capacity was -2.1%. To achieve the minimum assumed company hurdle rate (IRR = 10%),
a plant capacity of about 8 tonne/hr of biocrude was needed for both SFE scenarios (P-1,
P-2) and for distillation combined with multistage evaporation (P-4). For distillation
alone (P-3) the needed capacity is huge, at least 820 tonne/hr. Similarly a 20% IRR is
possible for P-1, P-2 and P-4 up to plant capacity of about 50 tonne/hr, while for P-3 the
capacity should be ridiculously higher, more than 5000 tonne/hr.
For the double and single stage SFE scenarios (P-1 and P-2), the IRRs drop to 11.7% and
11.9% respectively with a doubling of the price of imported electricity used. For P-3 and
P-4 distillation processes the corresponding IRR drop will be just to -2.2% and 15.1%
respectively. Similarly upon doubling the steam price, the IRR for P-1, P-2 and P-4 will
decrease to about 13.7%, while the corresponding decrease in IRR will be up to -6.5% for
P-3 process. The IRRs drop from 15% to about 5%, for P-1, P-2 and P-4, upon 25%
decrease in product sale prices, the corresponding increase in IRRs will be up to 39%
when product sale prices increased by 75%. For P-3, the IRR will reach 10% and 20%
with at least 75% and 165% respectively increase in product sale prices.
iv Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil
TABLE OF CONTENTS
Keywords .................................................................................................................................. i
Abstract .................................................................................................................................... ii
table of contents ...................................................................................................................... iv
List of Figures ........................................................................................................................ vii
List of Tables ......................................................................................................................... xii
List of Publications ............................................................................................................... xiv
Statement of Original Authorship ...........................................................................................xv
Acknowledgements ............................................................................................................... xvi
2.4 Composition of bio-oil from thermochemical conversion of biomass .........................14 2.4.1 Monophenols and low molecular weight acid contents of bio-oil ......................15
2.5 Binary system solubility data of bio-oil compounds ....................................................17
2.6 Supercritical CO2 extraction and fractionation of bio-oil .............................................19
2.7 Discussion .....................................................................................................................26 2.7.1 Solubility data .....................................................................................................26 2.7.2 Modelling Binary solubility data ........................................................................32 2.7.3 Use of binary data in preliminary assessment and design of fractionation ........35 2.7.4 Solubility data consistency and accuracy ...........................................................36
3.6 Results and discussion ..................................................................................................59 3.6.1 Solubility data .....................................................................................................59
3.7 Process design and techno-economic evaluation using Aspen Plus® to extract bio-oil from the
4.6 Process design and techno-economic evaluation using Aspen Plus® ..........................90 4.6.1 First separator .....................................................................................................92 4.6.2 Second separator .................................................................................................94 4.6.3 Recycling ............................................................................................................95 4.6.4 Product purification ............................................................................................95
4.7 A techno-economic assessment of process scenarios ...................................................98
4.8 Results and Discussion .................................................................................................99 4.8.1 Sensitivity analysis ...........................................................................................108 4.8.1.1 Capital cost ....................................................................................................108 4.8.1.2 Electricity .......................................................................................................110 4.8.1.3 Steam .............................................................................................................113 4.8.1.4 Product sale price ...........................................................................................115
Figure 1-2 Flow of chapters according to research aims of this work ..........................5
Figure 2-1 General experimental setup of supercritical CO2 extraction system with optional co-solvent addition. T: temperature measurement and control, BPR: back pressure regulator, MV: micrometering valve. .......................20
Figure 2-2 Extract yields and concentration of single ring phenols in extract from supercritical fluid rectification of softwood Kraft lignin microwave-pyrolysis oil for varying solvent to bio-oil ratio. (inherent and experimental random errors were not reported in the original source) [25] ...............................................................................................................21
Figure 2-3 Effect of adsorbent on typical selective enrichment of phenols and acids in scCO2 extraction of corn stalk pyrolysis oil (original pyrolysis oil contained 10.74 % phenols and 28.15% acids). (inherent errors related to extract yields and compositions and experimental random errors were not reported in the original source) [30] ..................................................23
Figure 2-4 Ratio of total benzenoids extracted to total acids extracted as a function of different solvent/bio-oil ratios used in scCO2 extraction of wheat-wood sawdust [29] and wheat-hemlock [26] pyrolysis oils ..............................24
Figure 2-5 Effect of increasing pressure on solubilities of different bio-oil compounds in supercritical carbon dioxide at 333 K temperature. Random or ultimate error were not reported for eugenol in the original source [89]. For vanillin the maximum reported uncertainty of + 16.4% is shown [83]. ..............................................................................................................26
Figure 2-6 Solubility isotherms showing crossover pressure regions for vanillin-CO2 (left) and phenol-CO2 (right) binary systems. The maximum reported uncertainty for vanillin [83] of ±16.4% is shown. ....................................27
Figure 2-7 CO2 densities calculated at 40 oC with PR-EOS [115] and Span and Wagner EOS [111] .......................................................................................28
Figure 2-8 Solubility data (see supporting information, Table 2.8S) plots of different monophenols and acetic acid. CO2 density is calculated here using the Span and Wagner [111] method. The maximum reported uncertainty for vanillin [83] of 16.4% is shown. Random or ultimate error were not reported in the original source for eugenol [89]. .....................29
Figure 2-9 CO2 density variation as a function of temperature and pressure, Left: 3-D surface plot of temperature-pressure and CO2 density, Right: 2-D plane plot of CO2 density curves against pressure axis at different temperatures ...............................................................................................30
Figure 2-10 Effect of CO2 induced acidity (in terms of final solution pH of 3, 3.4 & 4.2 corresponding to initial pH of 3, 5 and 8 respectively) on percent
viii Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil
recoveries of phenol and 2,4,6-trichlorophenol solutes from aqueous matrices at 150 atm pressure during supercritical extraction with pure CO2. Inherent and experimental random errors were not reported in the original source [116]. ..................................................................................31
Figure 2-11 Parity plots of experimental vs predicted solubility of different bio-oil compounds on natural log scale. Dots of one colour correspond to one data source. ..................................................................................................34
Figure 2-12 Solubilities of different monophenols in scCO2 predicted by fitted model at 308 K temperature ......................................................................35
Figure 2-13 Extraction trends of different monophenols with scCO2 from bio-oil mixtures of softwood Kraft lignin [25] and beech wood [23] pyrolysis oils ......................................................................................................................36
Figure 2-14 Parity plots of experimental versus predicted solubilities using data and parameters based on [84] (plot A) and using the same correlation parameters to predict solubility data presented in [86] (plot B) ...........36
Figure 3-1 High-pressure phase equilibrium apparatus used in this study to determine
benzyl alcohol solubility in scCO2. Labels: 1: CO2 cylinder; 2: CO2 pump; 3:
xii Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil
LIST OF TABLES
Table 2.1 Experimental techniques and conditions used in literature studies for measurement of solute solubilities in supercritical carbon dioxide ......19
Table 2.2 Effect of pressure and temperature on extract yield and product concentration in extract during supercritical CO2 extraction of sugarcane bagasse and cashew nut shell pyrolysis oils. (inherent and experimental random errors were not reported in the original source) [27] ..............22
Table 2.3 Yields and acid-phenol contents of extracts obtained at 333.15 K temperature and 150 bar pressure during scCO2 extraction of beech wood pyrolysis oil (inherent and experimental random errors were not reported in the original source) [23] ........................................................25
Table 2.4 Chrastil correlation parameters for the solubility of several bio-oil compounds in supercritical CO2.................................................................33
Table 2.5S Single ring phenolics and low molecular weight carboxylic acid contents in bio-oils ......................................................................................39
Table 2.6S Major chemical compounds in low molecular weight carboxylic acid
fraction of bio-oils ........................................................................................39
Table 2.7S Major chemical compounds in single ring phenolic fraction of bio-oils .40
Table 2.8S Solubility data of single ring phenolics and acetic acid with supercritical
carbon dioxide in binary systems .................................................................40
Table 3.1 Aspen Plus® pure component properties used in modelling of this work .59
Table 3.2 Benzyl alcohol solubility in scCO2 data determined using the visual method
1] Wahab Maqbool, Philip Hobson, Kameron Dunn, William Doherty; Positive sealing material for supercritical carbon dioxide, Supergreen Conference, December 1-3, 2017, Nagoya, Japan.
Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil xv
STATEMENT OF ORIGINAL AUTHORSHIP
The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the best
of my knowledge and belief, the thesis contains no material previously published or
written by another person except where due reference is made.
Signature:
Date: 5th November, 2019
QUT Verified Signature
xvi Supercritical Carbon Dioxide Extraction and Fractionation of Bio-oil
ACKNOWLEDGEMENTS
I thankfully acknowledge Queensland University of Technology (QUT) to assist me with
my Doctor of Philosophy (IF49) candidature with QUT Postgraduate Research Award,
Australia-India Strategic Research Fund Top Up Scholarship and QUT HDR Tuition Fee
Sponsorship.
I would like to thank my supervisory team Philip Hobson, William Doherty and Kameron
Dunn for accepting me as a PhD student under their supervision, and for providing
valuable guidance and practical research expertise in the field of Energy and Process
Engineering.
I am thankful also to Neil Mckenzie, Kameron Dunn and Barry Hume for their generous
experimental and facilities maintenance support.
Lalehvash Moghaddam, Dylan Cronin, Adrian Baker, Wanda Stolz and Daniela Tikel
provided me with a lot of research support while I was working in analytical chemistry
laboratory. I am thankful to all of them.
I am fully aware of the sacrifices made by my parents and my brothers and sisters in
preparing me and allowing me to take on such a lengthy endeavour, and for making me
able to do it.
I will be under the burden of favours made by my Pakistani Fellows here at QUT, Fawad
Shah (mastermind), Aziz Pawar and Imran and Company. These Pakistani fellows helped
me a lot to keep going towards the end of my PhD.
My wife Sehar held me together and going through the final year of my PhD.
Chapter 1: Introduction 1
Chapter 1: Introduction
1.1 Research problem
Thermochemically produced bio-oil from lignocellulosic biomass is a complex
mixture of oxygenated hydrocarbons, pyrolytic lignin and water. Bio-oil is
potentially an abundant renewable source of fuels and high value chemicals [1-
3]. The complex nature of bio-oil presents many technological obstacles in
exploring its potential as a renewable feedstock for chemicals production.
Conventional chemical separation processes such as distillation and solvent
extraction often require high thermal energy inputs [4]; solvents may be
hazardous, expensive and difficult to recover [5]. Although supercritical
extraction has been widely studied with food, pharmaceutical and other niche
production systems it has been used industrially for only a few niche applications
such as the decaffeination of beverages as well as the extraction of essential oils
and bioactive compounds [6, 7].
Interest in the application of SFE to bio-oil fractionation has emerged in recent
years. Previous investigations into the use of SFE as a lower cost, environmentally
friendly alternative to conventional bio-crude separation processes have been
carried out [2, 3, 8-16] although these studies have been limited to bench scale
investigations.
This PhD study was part of larger project entitled integrated technologies for
economically sustainable bio-based energy run under the Australia-India Strategic
Research Fund (AISRF). This larger collaborative research project addressed
major gaps in knowledge and understanding around the production of biofuels
from surplus non-food non-fodder agriculture and forest residues in Australia
and India and was aimed at developing and demonstrating scalable and
sustainable technology platforms for commercial deployment. The PhD study is
a part of Sub-project 4 in the broader AISRF project (see Figure 1-1) and was to
establish the technical and financial feasibility of using SFE as an upgrading and
value adding process for bio-crude by addressing the following research
questions:
How does SFE as a process for the separation of bio-crude compounds compare
to other more conventional separation techniques in terms of the degree of
separation and energy efficiency?
What is the capability of thermodynamic model to accurately describe SFE and
fractionation of bio-crude?
2 Chapter 1: Introduction
What are the prospects of process integration and optimization to make SFE of
renewable chemicals from bio-crude a technically and economically viable option
in a bio-refinery?
1.2 Novelty of this work
SFE has been flagged previously as a potentially low cost, energy efficient process
for the recovery of renewable chemicals from bio-oil. This PhD study provides the
first comprehensive experimental and theoretical analysis to confirm the
extraction conditions, process configuration and financial outcomes required to
establish SFE as a potential technology for recovering high value renewable
chemicals from bio-oil. To this end this study has made notable contributions to
knowledge in related areas including:
Figure 1-1 Schematic showing the relationship of the current PhD
study to the broader AISRF project
Sub-project 1:
Biomass supply economics and logistics
Sub-project 4: Thermochemical
conversion (hydrothermal
liquefaction) of lignin-rich
streams to bio-oil
PhD Study:
Supercritical CO2
extraction and
fractionation of high
value chemicals from
bio-oil
Sub-project 3:
Biochemical
conversion of
cellulose-rich streams
Sub-project 2:
Biomass analysis and deconstruction
Techno-economics of integrated project technologies
Chapter 1: Introduction 3
• The proposed and demonstrated effectiveness of binary VLE data and
models as an accurate and relatively simple means of determining
supercritical fractionation conditions associated with the recovery of
individual chemicals from complex mixtures.
• The experimental measurement and use of existing VLE data to determine
the equations of state models required to accurately simulate the process
of SFE extraction of renewable chemicals from bio-oil.
• The implementation and use of the above EOS models within detailed
process simulation scenarios to determine preferred configurations and
establish optimum process conditions for chemical extraction and
purification using SFE and conventional technologies.
• The first published continuous pilot plant scale demonstration trials of the
recovery of chemicals from bio-oil using supercritical CO2.
• The provision of a detailed techno-economic assessment of SFE for
recovering chemicals from bio-oil and a comparison of the capital and
operating (including energy) costs of an equivalent plant utilising
conventional distillation and multi-stage evaporation technologies.
1.3 Research aims and objectives
This work was aimed at evaluating and optimising the efficacy of SFE utilising
supercritical CO2 for the recovery of chemicals from bio-crude produced from the
hydrothermal liquefaction (HTL) of black liquor (a by-product of sugarcane
bagasse pulping process).
The above aim was achieved so far by defining the objectives:
• Determine the experimental vapour-liquid equilibrium (VLE) data for key
bio-crude compounds (where this data was not available from the
literature);
• Investigate the thermodynamic modelling of a bio-crude mixture as a
series of binary VLE systems;
• Analyse the developed model in a process simulation environment for
validation of the QUT SFE pilot plant results;
• Study and compare the optimized techno-economics of SFE with
conventional distillation as a means of recovering high value compounds
from bio-crude
1.4 Research outcomes
Research outcomes associated with this study include:
4 Chapter 1: Introduction
1. Equation-of-state (EOS) based thermodynamic model development for SFE of
bio-crude with use of only solute-solvent binary interaction parameters
thereby neglecting the solute-solute interactions. The developed model
successfully predicted the fractionation conditions for stage-wise pressure
reduction separation of a multicomponent supercritical extract stream in our
experimental pilot plant SFE trials. The model was able to predict that for our
specific bio-crude system, catechol was the least soluble compound among
acetic acid, phenol, 4-ethylphenol and p-cresol, and could be selectively
separated into first separator when set at a relatively higher pressure.
2. The developed model was used in the construction of SFE process simulations
in Aspen Plus®, in which it was shown through evaluation of techno-
economics that increasing the S/B ratio from 6.2 up to 20.2 will decrease the
IRR from 15% to 9.5%.
3. A comprehensive techno-economic evaluation and comparison was made
between Aspen Plus® simulation scenarios for SFE with and without
fractionation and conventional distillation of bio-crude. Distillation of bio-
crude when combined with multistage evaporation, incurs slightly lower
annualized operating costs than SFE processes, yet the IRR value of about
15% is achieved for both SFE and distillation combined with multistage
evaporation processes. Distillation alone did not prove economical for bio-oil
separation with an IRR value of -2.1%. In terms of IRR, two-stage SFE was
shown to be marginally better (by 0.3%) than a single stage SFE process. For
the double and single stage SFE scenarios (P-1 and P-2), the IRRs reduced to
11.7% and 11.9% respectively with a doubling of the price for imported
electricity used. For the P-3 and P-4 distillation processes doubling imported
electricity costs reduced the IRR’s to -2.2% and 15.1% respectively.
Similarly doubling the steam price, the IRR’s for P-1, P-2 and P-4 will decrease
to about 13.7%, whilst the P-3 process IRR is reduced to -6.5%.
For P-1, P-2 and P-4, the IRRs drop from 15% to about 5% with a 25%
decrease in product sale prices, whilst the corresponding increase in IRRs will
be up to 39% when product sale prices increased by 75%. For P-3, the IRR
will reach 10% and 20% with at least 75% and 165% respectively increase in
product sale prices.
Chapter 1: Introduction 5
1.5 Summary of chapters
Figure 1-2 Flow of chapters according to research aims of this work
6 Chapter 1: Introduction
This chapter (Chapter 1) describes the research problem investigated, reasoning
for and scope of research work investigated. It articulates the aims of this work
and the associated target outcomes.
Chapter 2 is a critical review (review paper) of experimental studies to date on
SFE of compounds from bio-oil (pyrolysis oil, HTL oil) and of the availability of
relevant solubility data on binary systems associated with bio-oil compounds.
Binary data from the literature are correlated by empirical models as a means of
evaluating the quality of and comparing data from different sources. The focus on
VLE data of binary systems at this stage was in anticipation of implementing the
associated models within Aspen Plus® as a process optimisation and design tool
for commercial SFE systems (see Chapter 3 and 4).
Previous experimental SFE studies are critically reviewed to establish at an early
stage, the relative importance of process parameters such as temperature,
pressure, solvent density, pH etc., on SFE of bio-oil. Knowledge gaps are identified
and used to further refine the proposed experimental program associated with
the current study.
Chapter 3 is a submitted research article reporting on the phase equilibrium
experiments undertaken as part of this study to determine the solubility of benzyl
alcohol (as an exemplar of a bio-oil compound) in scCO2. Solubility data points
were determined for this binary system encompassing the full range of
temperatures and pressures relevant to the extraction and fractionation of target
bio-oil compounds. Data was determined using both synthetic (no-sampling) and
analytic-gravimetric (sampling) methods. Data measured by these means were
compared where possible to that reported in the literature.
This chapter describes a validation process by which it was shown that
thermodynamic correlations (developed in this study) based on data from the
literature could be used to accurately predict solubility characteristics measured
at the higher temperatures and pressures measured in the current study. The
purpose of determining experimental solubility data in this work is to compare
VLE data sets from literature, for use in process modeling and simulation.
Chapter 4, also a published research article, reports on the pilot scale trials
undertaken in this study. The trials were used to establish the accuracy of
utilising multiple binary VLE models to predict the fractionation of real bio-crude
solubilised in a scCO2 column extract stream of known composition.
This chapter also summarises the results of implementing binary VLE models in
the Aspen Plus® simulation code as the basis for the design and thermodynamic
optimisation of a practical, commercial scale SFE plant for the extraction and
recovery of target compounds from bio-crude. A financial overlay was developed
to determine the associated capital and operating costs of the plant in Aspen
Plus® based process scenarios. These models enabled a detailed techno-
Chapter 1: Introduction 7
economic comparison to be made between the proposed SFE and distillation
technologies.
A more complete description of the Aspen Plus® model and outputs (not
included in the above publication) is provided in Chapter 5.
Chapter 5 draws together the findings of this PhD study to provide a series of
conclusions regarding the viability of SFE compared with conventional
distillation technology. Recommendations for future work required to advance
SFE technology to the point where its full credentials as an economically and
environmentally sustainable means of extracting high value compounds from
bio-oil, can be realised.
An appendix at the end of this document provides details of the Aspen Plus®
simulation data and flowsheets associated with the summary results provided in
Chapter 4. It includes predicted stream conditions (temperature, pressure, flow
etc.) and a more complete reporting of utility requirements and costs.
matrices [11, 12], rapeseed oil [13] and olive oil [14] etc. Use of modern process
simulation software (Aspen Plus®) is also reported [15-18] in modelling the
phase behaviour of fatty acid/ scCO2 systems.
There are relatively few experimental studies in the literature [19-30] describing
the SFE of compounds from bio-oil. Unlike other SFE studies (i.e. for mixtures
other than bio-oil) detailed EOS modelling associated with bio-oil (including bio-
crude from HTL derived black liquor) is, to the best of our knowledge, entirely
absent from the literature.
This chapter also describes the use in this study of Chrastil type models [31] of
binary mixtures of individual bio-oil compounds and scCO2 in order to provide:
a) a means of comparing and elucidating any discrepancies between
reported data, and
b) an understanding of factors impacting the relative solubility trends of
selected bio-oil compounds in scCO2
2.1 Title: Supercritical carbon dioxide separation of carboxylic acids and phenolics from bio-oil of lignocellulosic origin: understanding bio-oil compositions, compounds solubilities and their fractionation
Wahab Maqbool, Philip Hobson*, Kameron Dunn, William Doherty
Queensland University of Technology (QUT), 2 George St, Gardens Point 4000 Brisbane,
Australia
2.2 Abstract
Bio-oil produced from the thermochemical treatment of lignocellulosic biomass
is increasingly recognised as a potentially abundant source of renewable
10 Chapter 2: Literature Review
chemicals and fuels. Single ring phenolics and low molecular weight carboxylic
acids are significant constituent compound groups found in bio-oil and are
important end product or intermediate commodity chemicals. Fractionation of
bio-oil using supercritical fluids (usually with CO2 as a solvent) is a relatively new
process being investigated worldwide at both laboratory and pilot scales.
Solubility data associated with supercritical carbon dioxide (scCO2) and the many
chemical compounds in the complex bio-oil mixture are required to predict the
extraction behaviour of different bio-oil compounds.
This article starts with a review of the composition of bio-oil in terms of the
phenolic and low molecular weight carboxylic acid fractions which are
potentially of commercial interest. Binary solubility data of major compounds in
these bio-oil fractions with supercritical CO2 are summarized and discussed.
Results from previously reported studies in which scCO2 is used as a solvent to
recover bio-oil fractions are reviewed and collated. Density and temperature
based Chrastil type models are developed using available data for the solubility
in scCO2 of some of the major bio-oil compounds. Finally, extraction of
compounds from the complex bio-oil mixture is discussed in terms of the trends
predicted by the respective individual binary solubility models.
Chapter 2: Literature Review 11
QUT Verified Signature
12 Chapter 2: Literature Review
2.3 Introduction
Thermochemical conversion processes have the potential to provide a highly
effective means of biomass valorisation through the production of a range of high
value fuels and chemicals. Among these technologies, fast pyrolysis and
hydrothermal liquefaction have in recent years attracted significant interest due
to the relative simplicity of the associated processes, high value products and the
potential to target a range of compounds of special interest through judicious
control of process conditions [32-34]. Both technologies produce an intermediate
bio-oil product which is a complex mixture of compounds forming a micro-
emulsion in which holocellulose (cellulose + hemicellulose) decomposition
products are stabilizing the lignin macro-molecules through hydrogen bonding
[35]. Bio-oil typically has a high water and pyrolytic lignin content together with
a number of other chemical classes including acids, sugars, esters, aldehydes,
ketones, phenol and phenol derivatives [36-38]. Bio-oil in its original state has
high acidity (pH 2.0-2.5), high viscosity, is thermally unstable and largely
immiscible with conventional liquid fossil fuels. Fractionation into thermally
stable and concentrated compounds is critical if the full potential of bio-oil as a
source of fuels and chemicals is to be realised [39, 40].
Bio-oil can be fractionated by standard process separation techniques. Liquid-
liquid extraction [41] may require large solvent volume [42] and separation of
the solvent itself from the fractionated products as an additional step.
Conventional distillation methods like steam distillation [43] and fractional
distillation [44] can also be used but they are generally energy intensive
processes and can cause thermal degradation of the products [42]. In a review by
Kim et al., [45] supercritical fluid extraction (SFE) using CO2 as a solvent and a
limited number of other techniques such as switchable hydrophilicity solvents
(SHS) and molecular distillation [46] were endorsed as appropriate means of
fractionating phenolic rich bio-oils. SHS are solvents which show change in
properties (such as polarity) in response to the addition of a trigger component
(usually CO2) in the system [47]. With the use of SHS, extract yields may increase
because of the enhanced dissolving power of the solvent but the recovered
solvent is more contaminated with products [48] when compared with the use of
scCO2 alone as a solvent. Molecular distillation can require the use of excessive
temperatures (up to 130 ⁰C) [49] and has limited scope for tuning selectivity
based on vapour pressure differences of compounds. Supercritical carbon
dioxide fractionation (SCF) by contrast permits a high degree of selectivity
through control of both density and temperature where the temperatures
employed do not cause product degradation.
SCF has been the focus of a number of major collaborative research programs to
explore its potential in fractionation and stabilization of bio-oil [24, 50, 51].
Historically, supercritical extraction and fractionation techniques have been used
Chapter 2: Literature Review 13
in the food and nutraceutical industries for the recovery of plant, animal and food
extracts [52, 53]. Supercritical solvents are favoured in these and other
applications due to their relatively high densities and diffusivities [54]. CO2 is a
commonly used supercritical solvent because of its non-toxicity, non-
flammability, low cost and abundant availability [54, 55]. In addition CO2 has
advantages over other commonly used solvents like ethanol, methanol, acetone,
ethane and propane due to its near ambient critical conditions [29]. Although
solvents such as ethane and propane have lower critical pressures than carbon
dioxide they are highly flammable [56].
Some of the advantages of using SCF for bio-oil separation are the high level of
control of solvent density (and therefore solubility) that can be achieved through
relatively small variations in temperature and pressures, [57] its suitability for
thermally labile natural substances [58] and selective extraction of low polarity
compounds (aldehydes, ketones, phenols etc.) [59]. The scCO2 extraction is not
without its disadvantages. For example: a) it is a weakly polar solvent and
therefore limited to the selective extraction of non-polar to weakly polar
compounds; b) the use of high pressures and densities in this process to enhance
total extract yields may result in poor separation of feed mixture components
and; c) although the extract yield and selectivity associated with scCO2 can be
modified with the use of a polar co-solvent, some of these solvents may be
problematic particularly for pharmaceutical and food applications.
The wide spectrum of chemical compound classes present in bio-oil provides
significant challenges for extraction using scCO2. For this reason, the number of
experimental studies reporting on SFE of actual bio-oil (rather bio-oil synthesised
from model compounds) are relatively few. One of the main challenges in the
extraction of compounds from such complex systems is the non-availability of
appropriate vapour-liquid equilibrium (VLE) data for the design of multistep
fractionation processes. In addition to reviewing available data, this work will lay
down some simple procedures to estimate the extraction behaviour of bio-oil
compounds using simple binary VLE data.
Bio-oil is a complex mixture of compounds. The presence of water in bio-oil
requires special attention and has been challenging in the past for extensive
experimental studies aimed at designing effective fractionation processes for
aqueous mixtures. For future more detailed design purposes, complete phase
equilibrium data including distribution coefficients of components between the
scCO2 and aqueous phases need to be determined. This study explores the use of
a simple methodology in which binary solubility data alone is used to understand
fractionation of the solutes-rich scCO2 phase typical of that produced by a
relatively high temperature and high pressure scCO2 counter flow water
stripping column. Components extracted by the scCO2 water stripping stage will
have minimal water content and therefore the assumption of negligible solute-
solute interactions (in relation to water) may be invoked.
14 Chapter 2: Literature Review
In this paper the most prevalent low molecular weight carboxylic acids and single
ring phenol (monophenol) components typically found in bio-oil will be
identified and the availability and accuracy of the corresponding binary VLE data
summarized. Binary solubility data of these major compounds will be discussed
and modelled to compare their solubility trends. Experimental bio-oil extraction
studies from the literature will be critically reviewed and the reported extraction
behaviour will be discussed in the light of binary VLE data.
2.4 Composition of bio-oil from thermochemical conversion of biomass
Composition data for thermochemical conversion of biomass is abundantly
available in the literature for a wide range of operational conditions, process and
rector designs. Experimental studies have used a range of temperature and
pressure conditions for collection and condensation of and subsequent analysis
of hydrothermal liquefaction and pyrolysis products. The role played by water
both as a reactant and product as well as the way in which bio-oil water content
is reported provides further complications in interpreting reported data. Water
is produced in large quantities during pyrolysis and is considered a part of
pyrolysis oil; in hydrothermal liquefaction water (or a hydrocarbon or a
combination of both) is added to the biomass as a reactant [60-62].
In a study by Doassans-Carrere et al. [36] fast pyrolysis and direct liquefaction of
identical biomass feedstocks (beech sawdust) is compared in terms of bio-oil
compositions. Here, differences in chemical compositions of pyrolysis and
liquefaction oils may be explained by different fraction collection and fraction
designation procedures. In this study, [36] the removal of water from the
liquefaction oil also caused removal of acetic acid, phenol and two other
unidentified compounds. Differences in chemical compositions of the pyrolysis
and liquefaction oil samples may also be explained through the prevalence of
hydrolysis reactions in the liquefaction process in which opening of the
levoglucosan ring structure occurs resulting in the production of sugars. By
contrast levoglucosan was reported as a significant component of pyrolysis oil.
Acetic acid, acetone, furans, phenols, oxalic acid and levoglucosane were largely
present in pyrolysis oil while liquefaction oil contained ketones, phenols
(guaiacol, syringol), furans, levulinic acid and etheric compounds.
Castellvi Barnes et al. [63] also compared the pyrolysis and liquefaction of pine
wood feedstock. Liquefaction studies were carried out with 10 wt % pine wood
using a reaction time and temperature of 30 min and 300 ⁰C respectively in three
solvents: guaiacol (GL), a guaiacol-water (GWL) mixture and water (WL)
separately. Pyrolysis oil was obtained by treating the pine wood at 500 oC for a
reaction time of 20-25 min for solid particles and below 2 seconds for the oil. Gel
permeation chromatography (GPC) was used to isolate solvents and different
fractions based on apparent molecular weight. The apparent molecular weight
Chapter 2: Literature Review 15
distribution through GPC showed a significantly greater proportion of heavy
molecules in liquefaction compared to pyrolysis oils. In terms of deoxygenation,
between 35-45% oxygen is lost in liquefaction while 20% is removed in pyrolysis
compared to the original oxygen contents in the wood. In both types of bio-oil,
carbohydrates and lignin are believed to be contributing to the production of
aromatic and aliphatic compounds the relative proportions of which are
dependent on the type of process and in the case of liquefaction oils, the nature
of the solvent. Generally, in liquefaction, the yield of aromatics (typically 40% to
60%) was greater than the lignin content of untreated wood (25%) which
suggests that carbohydrates are converted to aromatics in bio-oil along with
lignin. The aromatic contents of pyrolysis oil were consistent with those present
in the original untreated wood. Furans, phenols, acetic acid and other aromatic
and aliphatic compounds were usually present in both types of bio-oils. For the
three liquefaction solvents and pyrolysis trials the extent of deoxygenation
appeared to occur in the order of: WL > GWL > GL > pyrolysis. A qualitative
parameter of reaction severity (extent of decrease in residual carbohydrates and
oxygen content both in oil and solid residues) was defined to compare the
liquefaction bio-oils and it was proposed in the order of: WL > GWL > GL. This
indicates that reaction severity in effect increases with increasing water
concentration as it will cause a decrease in both residual carbohydrates and
oxygen content.
Ponomarev et al. [64] reviewed thermochemical methods for biomass conversion
including hydrothermal liquefaction, liquefaction in organic solvents and
pyrolysis. Use of different liquefaction solvents such as low molecular weight
acids, phenols, alcohols or different combinations of these compounds with or
without water were reported as the cause of large differences in bio-oil
composition. Bio-oil composition from fast pyrolysis is strongly dependent upon
operating temperature and residence time and as with liquefaction the resulting
bio-oils contained many chemical classes such as acids, phenols, alcohols and
other lignin and carbohydrate degradation products.
In summary, separation techniques using polarity or any other property related
to intermolecular interactions can be used for fractionation of both pyrolysis and
liquefaction oils owing to significant similarities in their compositions.
2.4.1 Monophenols and low molecular weight acid contents of bio-oil
Phenolics form the largest group of chemical compounds within bio-oil (up to 50 wt%)
[65] and are present in the form of monomeric units (monophenols) and oligomers
(pyrolytic lignin, weight up to ~ 5000 amu) [66]. Monophenols and low molecular
weight carboxylic acids are always present in lignocellulosic derived bio-oils.
16 Chapter 2: Literature Review
Table 2.5S (supporting information) summarizes the phenolic and acid contents of
bio-oils from pyrolysis and liquefaction of different biomass feedstocks. Major
chemical compounds of both bio-oil fractions are listed in supporting information
Table 2.6S and Table 2.7S on wt% dry biomass basis.
Monophenols are of special interest to the chemical industry as intermediates for a
wide range of products such as paints, resins and adhesives.
Table 2.5S (supporting information) shows a collation of bio-oil composition data
reported as wt% of dry biomass (where the appropriate mass balance has been
reported) and area% of the spectra produced by gas chromatography–mass
spectrometry (GC-MS) analysis of the bio-oil to determine monophenols and
carboxylic acid contents. Where data is reported on an area% basis amounts are seen
to vary over a wide range; when reported on dry biomass basis monophenols and acids
yields are in the range of 6-10 wt% each. The yield values calculated in our work using
composition data reported in the literature match those quoted more generally (i.e.
without reference to specific biomass sources) in the literature where yields of both
acids and monophenols are in the range of 5-10 wt% each on dry biomass basis [67,
68].
It is evident from Table 2.6S (supporting information) that acetic acid is the most
abundant of all the low molecular weight acids. Acetic acid derives from the
cellulose component of biomass via the production and subsequent
decomposition of 2-Furancarboxaldehyde and 5-methyl-2-Furancaboxaldehyde
[61]. Acetic and formic acids may also originate from the rupture of lignin
aliphatic chains [69].
Phenolics are formed from the lignin in biomass and it is believed that
degradation of lignin produces mainly 2-methoxyphenol (guaiacol) [61] and
syringol [70] depending upon the nature of the wood (softwood or hardwood)
feedstock. Further decomposition of guaiacols at higher temperatures (> 350 oC)
Chapter 3: Fundamental Experimental Data and Equation of State Model 49
Chapter 3: Fundamental Experimental Data and Equation of State Model
This chapter will determine high-pressure solubility data of an exemplar bio-oil aromatic compound (benzyl alcohol) in scCO2. The data is determined in a variable volume full-view cell by two different techniques of solubility determination for the sake of comparison and validation with literature. The purpose of determining experimental solubility data in this work is to compare VLE data sets from literature, for use in process modeling and simulation. Aspen Plus® is used for modelling the phase behaviour of benzyl alcohol and CO2 binary system with the help of PR-BM property method. This property method will then also be used in Chapter 4 for different other compounds of our bio-crude mixture.
3.1 Title: Comparison of literature data, thermodynamic modelling and simulation of supercritical fluid extraction of benzyl alcohol
Wahab Maqbool, Kameron Dunn*, William Doherty, Neil McKenzie, Philip Hobson
Queensland University of Technology (QUT), 2 George Street, Gardens Point, 4000 Brisbane,
Australia
3.2 Abstract
Benzyl alcohols are important class of aromatic alcohols used in the fine chemical
and pharmaceutical industries which can be found in extracted bio-oils produced
from the thermochemical liquefaction of lignocellulosic biomass.
Equation-of-state (EOS) models can be used to describe the vapour-liquid
equilibrium (VLE) to support supercritical fluid extraction (SFE) studies of key
compounds from bio-oils, but ideally require experimentally determined binary
VLE data at appropriate conditions of temperature and pressure.
In this study, high-pressure binary solubility data is reported for benzyl alcohol
and a supercritical carbon dioxide (scCO2) mixture. Data has been determined
experimentally at temperatures of 313 K, 333 K and 353 K and at pressures up to
284 bar. Data was determined with both continuous flow analytic (sampling) and
static-synthetic (visual) methods, and used to validate and support existing
published solubility data which was subsequently used in process modelling and
simulation.
50 Chapter 3: Fundamental Experimental Data and Equation of State Model
It was shown that the literature VLE data regression on Peng-Robinson-Boston-
Mathias (PR-BM) model was good in predicting benzyl alcohol solubility data
determined both in this study and in previous literature. The regressed model
was incorporated into Aspen Plus® process simulations for the SFE of benzyl
alcohols from an aqueous mixture (representing bio-oil). Techno-economics of
different SFE process scenarios are determined and compared with by
thermocouple; 14: temperature indicator; 15: syringe; 16: two-way valve; 17:
distributor; 18: rupture disc.
Solubility was determined by both static-synthetic (visual) and continuous-flow
analytic (sampling) methods. The purpose of acquiring data via two distinct
solubility determination methods was to provide a means of cross-checking new
data, and to assist in comparing data of this work to data from other sources
potentially using different methods and procedures. A known amount of solute
was injected into the cell for each method via a syringe that was connected to the
bottom flange of the cell with a valve. Liquid CO2 from a gas cylinder (having an
internal dip tube) was fed to a manual capstan pump where it was cooled to 12 oC and subsequently further pressurised and delivered by the capstan pump at
the required pressure. The amount of liquid CO2 injected into or otherwise passed
through the cell to the sample loop outlet, was determined by reading (from a
Vernier scale) the number of calibrated manual turns applied to capstan pump. It
is preferential to use a gas-meter to calculate the amount of gas in an
experimental setup, but due to non-availability of a gas-meter, it was found
through literature data comparison of our data that the amount of CO2 calculation
by counting the number of pump capstan turns was also an accurate technique.
Once the desired temperature inside the view cell was achieved, the chosen
procedure to determine the solubility data was followed.
For the visual method, solute was first completely solubilized by increasing the
pressure within the view cell via a second capstan piston attached directly to the
56 Chapter 3: Fundamental Experimental Data and Equation of State Model
view cell. The pressure was then slowly reduced until the solute started
precipitating out. The corresponding pressure was noted and the solubility at
that point was simply calculated from the already known amounts of solute and
solvent that had entered the view cell. Subsequent solubility points were
determined by adding progressively more solvent and determining the
corresponding (lower) precipitation pressure. In the visual method the pressure
was reduced in small increments of about 0.5- 2.0 bar that was subsequently
followed by a waiting period of approximately 5 seconds between each
increment.
In visual method, the amount of solute injected into the view cell was usually up
to a few grams to minimize the uncertainty in determined solubility. It is
important to note that using small quantities of solute are more likely to
introduce larger uncertainties in the determined solubility data. Also, charging of
solute into the view-cell was followed by pumping some air to completely carry
the solute into the view-cell. The view cell was then purged with gaseous CO2 to
remove air.
Figure 3-2 Configuration of view cell assembly used in this study to measure
solute solubility in scCO2 by continuous flow sampling method.
For the sampling method (Figure 3-2), cell connections were rearranged in such
a way that CO2 was continuously fed in through the bottom flange and left the
sample cell via the sampling loop outlet. In the sampling method of solubility
Chapter 3: Fundamental Experimental Data and Equation of State Model 57
determination, both valves across the sample loop were open to allow the flow of
CO2 out of the view cell. A small sampling vial was attached to the loop outlet
where depressurisation of the solute laden sample occurred and the previously
solubilised solute collected. Each sample was collected over a period of 30
minutes to 1 hour. Mass flow rates of liquid CO2 from the pump into the cell
ranged between 0.15 to 0.32 g.min-1. The amount of solute sample collected in
the experimental trials ranged between 0.15 – 0.40 g. The variation in CO2 mass
flow rates had no effect on determined solubility indicating that saturated
equilibrium conditions were prevalent.
3.5 Thermodynamic modelling
Modelling was performed in Aspen Plus® software using the Peng-Robinson-
Boston-Mathias (PR-BM) property method. The Peng-Robinson EOS is the basis
of PR-BM property method [14]. The Boston-Mathias alpha function and
asymmetric mixing rules [15] are used in conjunction with EOS to enable
modelling of polar, non-ideal chemical systems [16]. Eqs. (1-14) are
mathematical expressions of the PR-BM model with asymmetric mixing rules.
𝑃 =𝑅𝑇
𝑉𝑚−𝑏−
𝑎
𝑉𝑚(𝑉𝑚+𝑏)+𝑏(𝑉𝑚−𝑏) (1)
𝑏 = ∑ 𝑥𝑖𝑏𝑖𝑖 (2)
𝑎 = 𝑎0 + 𝑎1 (3)
𝑎0 = ∑ ∑ 𝑥𝑖𝑥𝑗(𝑎𝑖𝑎𝑗)0.5
(1 − 𝑘𝑖𝑗)𝑗𝑖 (4)
Eq. (4) is the standard quadratic mixing term, where 𝑘𝑖𝑗 has been made
temperature dependent.
𝑘𝑖𝑗 = 𝑘𝑖𝑗(1)
+ 𝑘𝑖𝑗(2)
𝑇 + 𝑘𝑖𝑗(3)
𝑇⁄ (5)
Where 𝑘𝑖𝑗 = 𝑘𝑗𝑖 and superscripts (1), (2) and (3) are numbered terms in Eq. (5)
𝑎1 = ∑ 𝑥𝑖[∑ 𝑥𝑗((𝑎𝑖𝑎𝑗)1 2⁄ 𝑙𝑖,𝑗)1 3⁄𝑛𝑗=1 ]
3𝑛𝑖=1 (6)
Eq. (6) is an additional asymmetric term used to model highly non-linear systems
𝑙𝑖𝑗 = 𝑙𝑖𝑗(1)
+ 𝑙𝑖𝑗(2)
𝑇 + 𝑙𝑖𝑗(3)
𝑇⁄ (7)
Where 𝑙𝑖𝑗 ≠ 𝑙𝑗𝑖 and superscripts (1), (2) and (3) are numbered terms in Eq. (7)
The pure component parameters for PR-EOS are calculated as follows:
𝑎𝑖 = 𝛺𝑎
𝑅2𝑇𝑐𝑖,𝑒𝑥𝑝2
𝑃𝑐𝑖,𝑒𝑥𝑝𝛼𝑖 (8)
𝛺𝑎 =8(5𝜂𝑐+1)
49−37𝜂𝑐≈ 0.45724 (8a)
58 Chapter 3: Fundamental Experimental Data and Equation of State Model
𝑏𝑖 = 𝛺𝑏𝑅𝑇𝑐𝑖,𝑒𝑥𝑝
𝑃𝑐𝑖,𝑒𝑥𝑝 (9)
𝛺𝑏 =𝜂𝑐
𝜂𝑐+3≈ 0.07780 (9a)
𝜂𝑐 = [1 + √4 − 2√23
+ √4 + 2√23
]−1
≈ 0.25308 (9b)
The parameter 𝛼 is a temperature function, and is meant to improve the
correlation of the pure component vapour pressure. In standard PR-EOS, this
parameter is expressed with Eqs. (10-11).
𝛼𝑖(𝑇) = [1 + 𝑚𝑖(1 − 𝑇𝑟𝑖1 2⁄
)]2 (10)
𝑚𝑖 = 0.37464 + 1.54226𝜔𝑖 − 0.26992𝜔𝑖2 (11)
𝛼, defined in Eqs. (10-11) is used when 𝑇𝑟 < 1 (aka subcritical temperature),
otherwise the Aspen BM alpha function (Eqs. (12-14)) is used.
𝛼𝑖(𝑇) = [𝑒𝑥𝑝[𝐶𝑖(1 − 𝑇𝑟𝑖𝑑)]]
2
(12)
𝑑𝑖 = 1 + 𝑚𝑖 2⁄ (13)
𝐶𝑖 = 1 − 1 𝑑𝑖⁄ (14)
Such an α-function like BM does not pass the consistency test recently developed
by Le Guennec et al. [17-19]. In our case we did not notice any special loss of
accuracy when regressing Walther et al. [12] VLE data to PR-BM model in Aspen
Plus® Data Regression system.
Binary interaction parameters (𝑘𝑖𝑗 , 𝑙𝑖𝑗) must be determined from regression of
phase equilibrium data. The optimized values of these binary interaction
parameters were obtained by the maximum-likelihood objective function (Eq.
(15)), defined within the Aspen Plus® data regression system.
𝑄 = ∑ 𝑤𝑛 ∑ [(𝑇𝑒,𝑖−𝑇𝑚,𝑖
𝜎𝑇,𝑖)
2
+ (𝑃𝑒,𝑖−𝑃𝑚,𝑖
𝜎𝑃,𝑖)
2
+ ∑ (𝑥𝑒,𝑖,𝑗−𝑥𝑚,𝑖,𝑗
𝜎𝑥,𝑖,𝑗)
2𝑁𝐶−1𝑗=1 +𝑁𝑃
𝑖=1𝑁𝐷𝐺𝑛=1
∑ (𝑦𝑒,𝑖,𝑗−𝑦𝑚,𝑖,𝑗
𝜎𝑦,𝑖,𝑗)
2𝑁𝐶−1𝑗=1 ] (15)
In phase equilibrium measurements, there can be errors in measurement of
temperature, pressure and in the compositions of both vapour and liquid phases.
During data regression in this study, the standard deviations specified for
measured variables were as follows: 0.1 oC in temperature, 0.1% in pressure
(bar), 0.1% in liquid mole fraction and 1% in vapour mole fraction. The weighting
factor (𝑤𝑛) value of 1 was specified for all involved data groups in our regression
case. The objective function (Eq. (15)) was minimized using Britt-Luecke
algorithm [20].
Pure component properties of critical temperature (Tc), critical pressure (Pc) and
acentric factor (ω) used in the EOS modelling of this work are given in Table 3.1.
Chapter 3: Fundamental Experimental Data and Equation of State Model 59
Table 3.1 Aspen Plus® pure component properties used in modelling of this work
Component Tc (oC) Pc (bar) ω
Carbon dioxide 31.06 73.83 0.2236
Water 373.9 220.6 0.3449
Benzyl alcohol 447 43.74 0.3631
3.6 Results and discussion
3.6.1 Solubility data
Temperature, pressure and solute amount were determined with an uncertainty
of ±0.2K, ±1.5 bar and ±0.0005 g respectively.
Benzyl alcohol solubility was determined by both visual and sampling methods
at temperatures of 313.15K, 333.15K and 353.15K. The pressure range
investigated was 93-284 bar. This study extends the availability of existing
solubility data to above 200 bar. At least duplicate measurements were taken in
each case. Table 3.2 and Table 3.3 report average solubility measurements for
benzyl alcohol in scCO2 determined by visual and sampling methods respectively.
60 Chapter 3: Fundamental Experimental Data and Equation of State Model
Table 3.2 Benzyl alcohol solubility in scCO2 data determined using the visual
method
P
(bar)
CO2 density a (g/L)
Mole fraction
solubility, y x
103
P
(bar)
CO2 density
(g/L)
Mole fraction
solubility, y x
103
313.15 K 313.15 K
284 901 21.4 93.7 562 5.6
214.2 852 18.3 93.2 554 5.0
198.4 838 17.3 333.15 K
169.7 808 16.2 168.6 661 13.8
160.7 796 14.9 163 646 12.3
151.4 782 13.9 155.8 624 11.0
140.8 765 13.0 150 604 10.0
132 747 12.2 147 592 9.7
119 715 11.5 145.8 588 9.8
117.5 710 10.9 142.5 573 8.7
116 706 10.6 138 551 7.7
112.8 694 10.0 131 512 6.3
110.1 684 9.4 126 479 5.2
106 665 8.8 353.15 K
103.2 650 8.2 149 423 7.7
101 636 7.8 146.8 413 7.0
98.7 618 7.1 143.8 400 5.9
97.3 606 6.8 137 370 4.9
94.4 572 6.1 131 344 4.5
a Calculated by Span and Wagner equation of state [21].
Chapter 3: Fundamental Experimental Data and Equation of State Model 61
Table 3.3 Benzyl alcohol solubility in scCO2 data determined using the sampling
method
T (K) P
(bar)
CO2 density
(g/L)
Mole fraction
solubility, y x
103
313.15 100 629 6.9
313.15 135 754 12.0
313.15 200 840 16.8
313.15 250 879 18.8
333.15 200 724 17.4
333.15 280 814 27.8
353.15 280 724 29.8
Reproducibility of the results in both methods was considered acceptable. For the
visual method the precipitation pressure was determined with a maximum
uncertainty of ±2.5 bar; in the sampling method the maximum standard deviation
(see Appendix) between solubility measurement replicates was ±4.8% but
typically ±2%. Ultimate inherent uncertainty in mole fraction solubility resulting
from the propagation of individual system errors was estimated to be within
±1%.
A comparison of solubility data determined in this study at 313.15K showed good
agreement between the two methods (see Figure 3-3) with an average absolute
relative deviation (AARD1) of 5.9% between the two sets of data obtained from
the visual and the sampling method. A possible explanation for the slightly lower
values obtained from the sampling method is that some benzyl alcohol vapours
escape with the vapourised CO2 solvent leaving the collection vial (see Figure
3-2). When CO2 depressurizes in the collection vial, it releases the absorbed
benzyl alcohol in the vial and then CO2 vents out of the vial. As the collection vial
is not sub-cooled, some benzyl alcohol vapours might not condense and hence
escape with the CO2 being vented out.
1 See Appendix for the definition of AARD
62 Chapter 3: Fundamental Experimental Data and Equation of State Model
Figure 3-3 Comparison of benzyl alcohol solubility in scCO2, determined in this
study by visual and sampling methods of solubility determination. Horizontal
error bars represent the average uncertainty in measured precipitation pressure;
vertical error bars are standard deviation in the measured mole fraction
solubility.
Through our work on both solubility experimental methods, it was noted that the
visual method was quicker than the sampling method, and required
comparatively very small amounts of solute for solubility determination. On the
other hand, the sampling method was evidently more accurate in terms of
dividing the experimental error across a larger amount of samples collected over
extended time periods. It is also inherently difficult for the visual method to
determine solubility for different mixture compositions at exactly the same
pressure values, whilst for the sampling method this would not be an issue. Often
during the visual method, it was evident that small condensed or precipitated
droplets were hard to observe upon pressurization-depressurization cycles for
phase equilibrium measurements, thereby compromising the accuracy of data.
In this study, the Walther et al. [12] data was modelled within Aspen Plus® using
our selected model of PR-EOS. From our preliminary modelling works, it was
found that PR-EOS with BM mixing rules was quite good in representing binary
phase equilibria when at least one compound (benzyl alcohol) was of a polar
nature. Regressions were performed on 313.2K, 353.2K and 393.1K isotherms,
over a pressure range of 80.9 to 200.8 bar. The resulting optimized binary
interaction parameter values are given in Table 3.4. The parameters given in
Chapter 3: Fundamental Experimental Data and Equation of State Model 63
Table 3.4 are for only those terms of Eq. (5) and Eq. (7) which resulted in
statistically significant values.
Table 3.4 Benzyl alcohol - CO2 binary interaction parameter values for a PR-EOS
derived from the VLE data of Walther et al. [12].
Parameter 𝑘𝑖𝑗(1)a 𝑙𝑖𝑗
(1) 𝑙𝑗𝑖
(1) 𝑙𝑖𝑗
(2) 𝑙𝑗𝑖
(2)
Value b 0.1321 -0.1882 -0.4729 0.00054 0.0012
a component i is solute and component j is CO2, b in SI units
The PR-EOS model utilising the interaction parameters given in Table 3.4
provided a good fit (see Figure 3-4) for the regressed data [12]. The deviations of
the PR-EOS model predictions relative to the experimental VLE data [12] were
about 15% AARD for vapour phase, and less than 1% AARD for liquid phase
compositions.
Figure 3-4 Predicted (PR-EOS) and experimental (Walther et al.[12]) composition
- pressure phase diagram for a benzyl alcohol-CO2 binary system
64 Chapter 3: Fundamental Experimental Data and Equation of State Model
The model was effective in predicting the mole fraction composition of both
phases over the regressed temperature range. At pressures typically ≥ 100 bar,
vapour phase average AARD (14.5%) is acceptable, given the difficulty, generally,
encountered in determining the experimental vapour phase data. When data
below 100 bar is also included in the comparison, the vapour phase composition
varied by 45.8 %AARD. Slightly higher vapour phase AARD in solubility data is a
result of making the binary interactions parameter independent of temperature,
which gives same parameter values over the whole temperature range used in
regression. Regressed model predicted the liquid phase composition of Walther
et al. [12] very well, with maximum AARD less than 1%.
The regressed model was then used to predict the solubility of benzyl alcohol at
temperature, pressure conditions other than that used in regressing the binary
interaction parameters and used to validate solubility data in this study and from
other literature. This study has determined extended experimental solubility
data of benzyl alcohol at conditions of pressures from 93 bar up to 284 bar. Figure
3-5 is a graphic comparison between the regressed model’s predictions and the
actual experimental solubility data determined in this study.
Figure 3-5 Comparison of experimental solubility data of benzyl alcohol
determined in this study, with that of PR-EOS model predictions. The model was
first optimized with the help of experimental VLE data of Walther et al. [12]
Model predictions were in good agreement with the laboratory solubility data
also determined from this study (Figure 3-5). Relative discrepancies between the
model and experimental data produced in this study were found to be within 4.8
– 12.1 %AARD. It was also found in this study that our model regression based on
Chapter 3: Fundamental Experimental Data and Equation of State Model 65
the 313.2K and 353.2K isotherms reported by Walther et al. [12] provided an
accurate prediction (1.8 %AARD) of the liquid phase composition of the 393.1K
isotherm of the same study [12]. Hence, the regressed model proved capable of
correctly predicting both the liquid phase and vapour phase (solubility data of
this study) compositions.
Comparison of the regressed model with other literature studies [10, 11]
revealed that vapour phase compositions of benzyl alcohol and CO2 system varied
by 42 % - 58 % AARD (Figure 3-6).
Figure 3-6 Comparison of solubility data of benzyl alcohol in CO2 vapour phase
from literature [10, 11] and the regressed model of this work based on Walther
et al. [12] data
However, the regressed model was quite successful in predicting the liquid phase
compositions of the same binary system in literature [10, 11], with maximum
AARD less than 5% (Figure 3-7).
66 Chapter 3: Fundamental Experimental Data and Equation of State Model
Figure 3-7 Comparison of solubility data of benzyl alcohol in CO2 liquid phase
from literature [10, 11] and the regressed model of this work based on Walther
et al. [12] data
3.7 Process design and techno-economic evaluation using Aspen Plus® to extract bio-oil from the aqueous hydrothermally liquefied product
The regressed model was then incorporated into simulating process scenarios in
Aspen Plus® process simulation software. A pseudo binary bio-oil (the crude
aqueous mixture product following liquefaction) was assumed to be the feed to a
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics 79
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics
This chapter will use the PR-BM property method, successfully implemented
previously in Chapter 3, to model phase behaviour of our bio-crude compounds
with CO2. The model is then successfully validated on pilot plant trials of SFE of
bio-crude and subsequent two-stage fractionation of extract stream. Aspen
Plus® simulation scenarios are then constructed to evaluate the techno-
economics of SFE of our bio-crude mixture for varying solvent/bio-crude ratios.
The SFE process economics were also compared with a conventional distillation
process for bio-crude.
4.1 Title: Extraction and purification of renewable chemicals from hydrothermal liquefaction bio-oil using supercritical carbon dioxide: A techno-economic evaluation
Wahab Maqbool, Kameron Dunn, William Doherty, Neil Mckenzie, Dylan Cronin, Philip
Hobson*
Queensland University of Technology (QUT), 2 George Street, Gardens Point, 4000 Brisbane,
Australia
4.2 Abstract
Supercritical fluid extraction (SFE) and fractionation of products from a complex
mixture such as bio-oil, where many compounds are present in low
concentrations, is a difficult process to model. This difficulty arises from the
uncertainty associated with those interactions between mixture components for
which fundamental vapour-liquid equilibrium (VLE) data is not available. In this
work a novel extraction and purification concept is investigated using a
predictive model developed from VLE data of binary solute-solvent systems;
solute-solute interactions in the supercritical carbon dioxide (scCO2) phase are
neglected. The predictive component of the work employs an equation of state
(EOS) model to achieve the above task. The results of pilot plant trials utilising a
80 Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics
bio-crude feedstock were shown to be in good agreement with the model
predictions. Aspen Plus® process simulations were developed for the extraction
process which comprised of supercritical extraction and subsequent purification
steps utilising distillation and multistage evaporation. A techno-economic
analysis of different process designs were evaluated for comparison. In
particular, distillation as the primary separation process with and without
multistage evaporation were simulated to compare the economics of
supercritical extraction to distillation. It was found from simulation results that
distillation is a very energy intensive process, and total operating costs for it are
always greater than supercritical extraction counterparts. Combining multistage
evaporation with distillation will reduce the total operating cost to a slightly
lower value than that required for a supercritical extraction processes. However,
the internal rate of return (IRR) value was similar for both SFE and distillation
combined with multistage evaporation processes. Whilst the solvent/bio-oil
(S/B) ratio will have a considerable impact on total profits of SFE process in
relation to distillation.
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics 81
QUT Verified Signature
82 Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics
4.3 Introduction
Supercritical fluid extraction is currently in use for a number of niche applications
[1, 2] such as the decaffeination of coffee or the recovery of essential oils and
bioactive compounds from plant materials. The use of SFE for the extraction of
compounds from bio-oil has been the subject of a limited number experimental
studies [3-13]. The lack of fundamental investigations into SFE of bio-oil can be
attributed to the highly complex nature of bio-oil and the difficulty this presents
in describing this process in terms of phase equilibria. Bio-oils are made up of
large portions of water and many other individual chemical compounds, but the
latter only in small quantities [4, 6, 7, 13].
A fundamental modelling approach based on an equation of state was adopted in
the current study to investigate the novel SFE and subsequent purification of bio-
oil compounds. The developed model for multicomponent mixture was used to
determine the subsequent staged depressurization conditions required for the
recovery of individual compounds or groups of compounds from the supercritical
extract phase. In this work, in-house produced bio-oil from hydrothermal
liquefaction (HTL) of black liquor, also known more commonly as bio-crude, was
first extracted with scCO2 and subsequently fractionated in to two product
fractions with the use of stage-wise pressure reduction techniques.
In the currently proposed extraction process the highly dilute bio-crude in water
feedstock is first fed into the top of a SFE extraction column. Literature review
and preliminary experiments helped to determine the conditions of temperature,
pressure and bio-crude pH at which the SFE of our bio-crude from the aqueous
phase will produce equilibrated extract samples in the pilot plant trials. The
supercritical extract stream emerging from the top of the extraction column will
be loaded with different bio-crude compounds. As the bio-crude compounds are
absorbed in scCO2 medium, solute-solute interaction effects will be negligible in
this phase as compared to the liquid bio-crude phase. Exclusion of solute-solute
interactions will simplify the system such that only solute-solvent binary
interaction effects will now play the determining role in the phase behaviour
description of supercritical extract phase. The application of stage-wise pressure
reduction techniques for the purification of bio-compounds have been reported
in the literature[2, 14] but for mixtures other than bio-oils.
A Peng-Robinson equation of state[15] (PR-EOS) model was developed to
investigate the phase behaviour of the solutes-rich supercritical phase. This
model was subsequently validated against pilot plant scale trials for predicting
the stage-wise pressure reduction fractionation. This study is aimed at validating
the model predictions at preferably saturated (low S/B) conditions. That is why
a range of S/B ratios will be investigated in subsequent simulation work to give
reader some information on variation of SFE plant economics with change in
solvent usage. Another aim of this study is to, for the first time, compare the
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics 83
techno-economics of scCO2 separation of bio-crude with that of a conventional
distillation process. This has been achieved by simulating both the SFE and
distillation separation processes in Aspen Plus® and then evaluating the
respective process economics.
4.4 Experimental methodology
4.4.1 Materials
Carbon dioxide was purchased from Supagas (Australia), with purity ≥ 99.9 wt%.
Bio-crude was produced in-house from the HTL of black liquor, where the black
liquor was a lignin-rich by-product of a bagasse pulping process. Phenol, p-cresol,
catechol, 4-ethylphenol, acetic acid, docosane, sulphuric acid and acetone were
purchased from Sigma-Aldrich (Australia), each with purity ≥ 99.0 wt% except
for sulphuric acid and 4-ethylphenol which were ≥ 98.0 wt% and ≥ 97.0 wt% pure
respectively.
4.4.2 Bio-crude preparation and its characteristics
About 50 litres of bio-crude was produced from black liquor using the HTL
continuous reactor facility at QUT. HTL liquefaction of the black liquor was
performed at a temperature and pressure of 290oC and 220 bar respectively. The
HTL reactor residence time was 60 minutes. The bio-crude product was stored at
2oC in a closed container prior to the SFE pilot plant extraction trials. The
produced bio-crude was homogenous, had a dense blackish appearance and a
viscosity similar to water. When physical settling and separation is possible, the
oil fraction should be separated from aqueous fraction of bio-oil as a first option.
Doing so will not only reduce the SFE plant footprint but will also be very helpful
in reducing the operating costs of SFE separation as a result of working with
relatively small volumes.
The bio-crude produced in this work was quite thin and didn’t show any signs of
phase separation upon settling and weeks long storage. However, in relevant
future works, it is advised to look for any possible scenario of physical settling
and separation as a first resort for bio-oil separation. The native HTL bio-crude
had a pH of 9.0 but preliminary SFE pilot plant trials revealed that the extraction
at such a high pH was problematic as it caused foaming, clogging and carry-over
of water from the extraction column. To lower the pH of the bio-crude, sulphuric
acid was incrementally added and then vigorously agitated with an electric mixer
until a final pH of 4.4 was achieved. This pH-lowered bio-crude (pH=4.4) was
centrifuged at 3300 rpm (Beckman GS-6R centrifuge, Marshall Scientific, USA) for
5 minutes, to remove precipitates and suspended solids.
84 Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics
4.4.3 The SFE pilot plant setup
The SFE pilot plant used to determine the initial extraction and verify the
predicted stage-wise fractionation of bio-crude components is shown
schematically in Figure 4-1. The pilot plant was purchased from Applied
Separations (USA) and installed and commissioned at the QUT Pilot Plant
Precinct. It consisted of a CO2 reservoir, bio-crude feed tank, CO2 pre-heater,
temperature-regulated extraction column and two separators in series. Pressure
in the separators was controlled with back pressure regulating valves. The shut-
off valve in-between extraction column and first separator was used as a back-
pressure regulator to control the pressure in extraction column. Extraction
column was a 4 Litre vessel. Column ID is about 50 mm, whilst column height is
2.2 m. Separator 1 and 2 had a volume of 300 mL and 4 L respectively.
Figure 4-1 Pilot plant setup used in this work for supercritical extraction and
fractionation of bio-crude (T: temperature control, Sep: separator, MV:
micrometering valve). Sep-1 and Sep-2 were wrapped in trace heaters to
compensate for the cooling effects resulting from depressurisation of the extract
streams.
4.4.4 Extraction and Fractionation Procedure
Carbon dioxide from the reservoir cylinder is supplied at a set flow rate by a high
pressure pneumatic pump (Haskel, USA). This high-pressure CO2 is then passed
through a 1250 watts pre-heater to bring the CO2 up to the desired extraction
temperature, before entering into the extraction column. The extraction column
is a 4 litre stainless steel tubular vessel in which CO2 enters from bottom and bio-
crude from the top. The CO2 and bio-crude streams flow in counter current over
a densely packed bed made up of small tubular elements. The CO2 stream absorbs
the majority of non-aqueous bio-crude and then leaves from the top of the
extraction column where it is fed into the two separators in series. The remaining
bio-crude and the majority of water is continuously drained from the bottom of
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics 85
the extraction column as raffinate. Micrometering and back pressure regulating
valves are positioned so as to produce the required pressures in column and
separators respectively. Once the steady state operation has been reached and no
more fluctuations in temperatures, pressures and flowrates are observed,
sampling procedures are initiated. Separator fractions and column raffinate
samples were collected every 15-30 minutes once continuous operation was
achieved.
From the bio-oil solubility in scCO2 reported in the literature[3, 16] and
preliminary trials on a lab scale solubility cell, it was determined that minimum
mass flow ratios of CO2 to bio-oil of under 10 could be used, in the pilot plant
trials, to ensure getting saturated extractions and consistent solubility data for
analysis. Normally S/B should be greater than 10 to maximise yields but was
limited in the pilot plant trials to less than this value because of pump cavitation
issues. Run conditions used for the pilot plant extraction and fractionation trials
are summarized in Table 4.1.
Table 4.1 Parameters used in this work for the supercritical CO2 pilot plant
extraction and fractionation of bio-crude produced from HTL of sugarcane
bagasse black liquor. Extraction was performed at 55oC temperature and 206.4
bar pressure, and Sep-2 was maintained at 18.4oC temperature and 46.8 bar
pressure.
No.
Sep-1 CO2 flow1 (mL/min)
Bio-crude flow
(mL/min)
S/B ratio2 (mass basis)
Extract Yield (%)
Press. (bar)
Temp. (oC)
1
137.6 49
217 89 2.5 0.4
2 202 88 2.3 0.7
3 203 89 2.3 0.6
4
116.3 47
307 68 4.5 1.1
5 284 50 5.7 0.9
6 299 52 5.7 1.0
7
91.5 43
260 41 6.3 1.2
8 291 41 7.1 1.2
9 303 42 7.2 1.7
10 300 42 7.1 1.7
1 Flow rate is given for CO2 at extraction column inlet. Corresponding CO2 inlet
temperature and pressure conditions were 48.6oC and 206.4 bar respectively. 2
Bio-crude density was 1.09 g/mL.
After reviewing the temperature and pressure conditions commonly found in the
literature[4, 16] for such an extraction process and to ensure the density
86 Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics
difference between the two phases inside the extraction column was at least 150
g/L[17] to avoid flooding, the conditions in Table 4.1 were chosen in this work to
make a comparison between our experimental fractionation results and the
model predictions. Maximum CO2 density used in the pilot plant trials was 763
g/L.
4.4.5 Gas chromatography mass spectrometry (GC-MS) analysis
The quantities of several key compounds present in the bio-crude and extraction
products were determined by GC-MS analysis. This process was performed on an
Agilent (US) 6890 Series Gas Chromatograph and a HP 5975 mass spectrometer
detector, employing helium as the carrier gas. The installed column was a
dimethyl polysiloxane Agilent DB 5-MS, 30 m x 0.32 mm x 0.25 μm. A split-less
injection of 2 μL was delivered to the injection port set at 250 °C. The
temperature program commenced at 70 °C and was heated at a rate of 5 °C.min-1
to a temperature of 320 °C. Compounds were identified from the spectra by
means of the Wiley library-HP G1035A and NIST mass spectra libraries and
subsets-HP G1033A (a criteria quality value >90% was used). Analytical samples
were prepared in acetone at a concentration of 0.05 mg/mL. Standard solutions
of pure chemicals were also prepared in acetone, in order to produce a 5-point
calibration curve over a concentration range of 0.025 to 0.3 mg/mL. All standards
and analytical samples were spiked with Docosane at a concentration of 0.06
mg/mL, to act as an internal standard.
4.4.6 Nuclear magnetic resonance (NMR) spectroscopy
Each sample (100 mg) of the collected oil fraction was dissolved in 0.9 mL of
deuterated water (D2O)) and filtered. The 1H spectra were then recorded at 25
°C on a Bruker AVANCE III HD 600 MHz NMR spectrometer (Agilent, US)
equipped with a cooled 5 mm TCI Cryoprobe. A total of 8 transients having an
acquisition time of 1.7 seconds and a spectral width of 9 kHz were recorded using
the Bruker pulse sequence noesygppr1d which features water suppression. The
triplet phenol reference peak was used as an internal chemical shift reference
point (δH = 7.25). Processing used shifted squared sine bell Gaussian apodization
in 1H. Data processing and plots were carried out using ACD/NMR processing
software, with automatic phase and baseline correction.
4.5 Thermodynamic modelling
Modelling was implemented in Aspen Plus® software, using the Peng-Robinson-
Boston-Mathias (PR-BM) property method [18]. The Peng-Robinson Equation of
State (PR-EOS)[15] forms the basis of the PR-BM property method, and BM alpha
function and asymmetric mixing rules are used in conjunction with the EOS to
make it suitable for modelling polar, non-ideal chemical systems. Eqs 1-14 are
mathematical expression of PR-BM model with asymmetric mixing rules.
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics 87
𝑃 =𝑅𝑇
𝑉𝑚−𝑏−
𝑎
𝑉𝑚(𝑉𝑚+𝑏)+𝑏(𝑉𝑚−𝑏) (1)
𝑏 = ∑ 𝑥𝑖𝑏𝑖𝑖 (2)
𝑎 = 𝑎0 + 𝑎1 (3)
𝑎0 = ∑ ∑ 𝑥𝑖𝑥𝑗(𝑎𝑖𝑎𝑗)0.5
(1 − 𝑘𝑖𝑗)𝑗𝑖 (4)
Eq 4 is the standard quadratic mixing term, where 𝑘𝑖𝑗 has been made
temperature-dependent
𝑘𝑖𝑗 = 𝑘𝑖𝑗(1)
+ 𝑘𝑖𝑗(2)
𝑇 + 𝑘𝑖𝑗(3)
𝑇⁄ (5)
Where 𝑘𝑖𝑗 = 𝑘𝑗𝑖 and superscripts (1), (2) and (3) are numbered
terms in eq 5
𝑎1 = ∑ 𝑥𝑖[∑ 𝑥𝑗((𝑎𝑖𝑎𝑗)1 2⁄ 𝑙𝑖,𝑗)1 3⁄𝑛𝑗=1 ]
3𝑛𝑖=1 (6)
Eq 6 is an additional asymmetric term used to model highly non-linear systems
𝑙𝑖𝑗 = 𝑙𝑖𝑗(1)
+ 𝑙𝑖𝑗(2)
𝑇 + 𝑙𝑖𝑗(3)
𝑇⁄ (7)
Where 𝑙𝑖𝑗 ≠ 𝑙𝑗𝑖 and superscripts (1), (2) and (3) are numbered
terms in eq 7
The pure component parameters for PR-EOS are calculated as follows:
𝑎𝑖 = 𝛼𝑖0.45724𝑅2𝑇𝑐𝑖
2
𝑃𝑐𝑖 (8)
𝑏𝑖 = 0.07780𝑅𝑇𝑐𝑖
𝑃𝑐𝑖 (9)
The parameter 𝛼𝑖 in Eq. 8 is used to improve the accuracy of predicted
temperature response of the pure component vapour pressure. In standard PR-
EOS, this parameter is expressed with eqs 10-11.
𝛼𝑖(𝑇) = [1 + 𝑚𝑖(1 − 𝑇𝑟𝑖1 2⁄
)]2 (10)
𝑚𝑖 = 0.37464 + 1.54226𝜔𝑖 − 0.26992𝜔𝑖2 (11)
𝛼𝑖 defined in eq 10 is used when 𝑇𝑟 < 1 (subcritical temperature), otherwise
Aspen BM alpha function (eqs 12-14) is used.
𝛼𝑖(𝑇) = [𝑒𝑥𝑝[𝐶𝑖(1 − 𝑇𝑟𝑖𝑑)]]
2
(12)
𝑑𝑖 = 1 + 𝑚𝑖 2⁄ (13)
𝐶𝑖 = 1 − 1 𝑑𝑖⁄ (14)
Binary interaction parameters (𝑘𝑖𝑗 , 𝑙𝑖𝑗) must be determined from regression of
phase equilibrium data. The optimized values of these binary interaction
parameters were obtained by maximum-likelihood algorithm (eq 15), defined
within the Aspen Plus® data regression system.
88 Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics
𝑄 = ∑ 𝑤𝑛 ∑ [(𝑇𝑒,𝑖−𝑇𝑚,𝑖
𝜎𝑇,𝑖)
2
+ (𝑃𝑒,𝑖−𝑃𝑚,𝑖
𝜎𝑃,𝑖)
2
+ ∑ (𝑥𝑒,𝑖,𝑗−𝑥𝑚,𝑖,𝑗
𝜎𝑥,𝑖,𝑗)
2𝑁𝐶−1𝑗=1 +𝑁𝑃
𝑖=1𝑁𝐷𝐺𝑛=1
∑ (𝑦𝑒,𝑖,𝑗−𝑦𝑚,𝑖,𝑗
𝜎𝑦,𝑖,𝑗)
2𝑁𝐶−1𝑗=1 ] (15)
Table 4.2 provides the standard pure component properties of critical
temperature (Tc), critical pressure (Pc) and acentric factor (ω), used in the Aspen
Plus® modelling of the binary systems.
Table 4.2 Critical properties of pure compounds used in the Aspen Plus®
modelling of the binary systems
Component Tc (oC) Pc (bar) ω
Carbon dioxide
31.06 73.83 0.2236
p-Cresol 431.5 51.5 0.5072
4-Ethylphenol 443.3 42.9 0.5154
Phenol 421.1 61.3 0.4435
Catechol 490.85 74.9 0.6937
Acetic acid 318.8 57.9 0.4665
Water 373.9 220.6 0.3449
The default binary interaction parameters available in in Aspen Plus® were
adjusted in this study such that the PR-BM property method used in the analysis
produced predictions which agreed more closely with experimental solubility
data published in the open literature. Table 4.3 shows the deviations between the
default Aspen Plus® predictions and experimental vapour-liquid equilibrium
(VLE) data from literature, for all our binary systems. Regressed values of binary
interaction parameters for all our binary systems are given in Table 4.4. Acetic
acid in Aspen Plus® showed relatively poor agreement with experimental vapour
phase solubility data giving an average absolute relative deviation (AARD) of
about 30% when compared to Bamberger et al. (2000)[19] and about 35% to
Jonasson et al. (1998)[20] data. On the other hand, liquid phase composition data
of this system was reasonably represented with the same model, where the AARD
between model predictions and both experimental studies[19, 20] was within
10%. Bamberger et al. (2000)[19] also pointed out towards difficulty in
modelling the VLE data of acetic acid, whence his selected model represented the
vapour phase composition with yet 18% deviation to experimental data, but only
when more sophisticated modelling approach of taking into account the
dimerization of acetic acid was adopted. Yet, the model predictions of Bamberger
et al. (2000)[19] were 50% smaller than reported by Jonasson et al. (1998)[20].
This means the model chosen in this work, and which represents all our other
binary systems very well, can be reasonably extended to acetic acid and CO2
Chapter 4: Bio-oil Mixture Model, Pilot Plant Validation, Aspen Plus® Simulation and Techno-economics 89
binary system too, as the average deviation between our model predictions and
experimental data of different sources[19, 20] is on average 25-35% AARD. For
catechol experimental VLE data was not available, as catechol will be present in
solid phase and will exhibit solid-fluid equilibrium at our interested supercritical
extraction conditions. For this binary system no regression was done, and it was
found that the default model predictions were in reasonable agreement to
experimental solid-fluid data of Garcia et al. (2001)[21], with average deviation
of less than 20% AARD for data determined under 200 bar pressure. Similarly, no
experimental VLE data was available for 4-ethylphenol and CO2 binary system, so
no regression could be performed on this system as well, rendering the model
description of this system totally predictive in nature based upon critical
properties of pure components listed in Table 4.2 above.
Table 4.3 Percent AARD between predicted and experimental VLE data for
different solute-CO2 binary systems using the default regression coefficients for
the PR-BM property method model available in Aspen Plus®
Binary system Experimental data Isotherms (Temp. in K) Model deviation
(% AARD)
Phenol Pfohl et al. (1997)[22] 373.15 4.8
Yau et al. (1992)[23] 348, 373, 398 15.5, 8.4, 4.9