Novel Methodology for the Synthesis of Optimal ... - ACS Publications

Subscriber access provided by Yale University Library

Article

A NOVEL METHONODLOGY FOR THE SYNTHESIS OF OPTIMAL BIOCHEMICALS IN INTEGRATED BIOREFINERIES VIA INVERSE DESIGN TECHNIQUES Lik Yin Ng, Viknesh Andiappan, Nishanth Chemmangattuvalappil, and Denny K. S. Ng Ind. Eng. Chem. Res., Just Accepted Manuscript • DOI: 10.1021/acs.iecr.5b00217 • Publication Date (Web): 07 May 2015 Downloaded from http://pubs.acs.org on May 17, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Industrial & Engineering Chemistry Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

1

A NOVEL METHODOLOGY FOR THE

2

SYNTHESIS OF OPTIMAL BIOCHEMICAL IN

3

INTEGRATED BIOREFINERIES VIA INVERSE

4

DESIGN TECHNIQUES

5

Lik Yin Ng, Viknesh Andiappan, Nishanth G. Chemmangattuvalappil*, Denny K. S. Ng

6

Department of Chemical and Environmental Engineering/

7

Centre of Sustainable Palm Oil Research (CESPOR),

8

The University of Nottingham Malaysia Campus, Broga Road, Semenyih 43500, Malaysia.

9 10

ABSTRACT

11

The recent developments of process synthesis and design for integrated biorefineries have

12

significantly increased the potential of biomass to generate sustainable renewable energy as an

13

alternative source for fossil fuels. In addition, biomass can be converted into various value-

14

added products (e.g., biochemical, biomaterials, biosolvent etc.). To ensure the sustainable

15

production of energy and value-added products, biomass is converted into commodity and

16

specialty products in an integrated biorefinery. However, due to the increase in the number of

17

potential products, new reactions and technologies, determining of optimum products and

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 46

18

processing routes in an integrated biorefinery has become more challenging. Therefore, it is

19

essential to develop a systematic approach to address the abovementioned issues. In this work, a

20

novel two-stage optimisation approach has been developed to design optimal biochemical

21

products and synthesise optimum biomass conversion pathways in an integrated biorefinery. In

22

the presented approach, optimal biochemical products that meet the customer requirement are

23

first determined via signature based molecular design techniques.

24

conversion pathways that convert biomass into biochemical products which are identified in the

25

previous step can be determined via superstructural mathematical optimisation approach. A case

26

study of bio-based fuel production from palm-based biomass is solved to illustrate the proposed

27

approach.

In addition, optimum

28 29

KEYWORDS

30

Integrated biorefinery, process synthesis and design, product design, inverse design techniques,

31

integrated product and process design.

32 33

1. INTRODUCTION

34

In recent decades, declining fossil fuel (petroleum, natural gas and coal) reserves and increasing

35

environmental issues have fuelled the search of sustainable, renewable and clean sources of

36

energy1. One of the promising solutions is the use of biomass for producing fuels and energy.

37

Biomass is a biological material that is found in natural and derived materials2. It is a renewable,

38

potentially sustainable and relatively environmentally friendly source of energy3. Biomass can


2

Page 3 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


39

exist in different forms such as energy crops, forestry waste and municipal solid waste depending

40

on the local geographic conditions. Other than being utilised for power and heat generation,

41

these biomasses can be converted into various value-added products. For instance, palm-based

42

biomass (e.g. empty fruit brunch, palm kernel shell, palm oil fibre etc.) which are produced from

43

palm oil industry can be converted into bioplastic, biobriquettes, biomaterials, biochemical etc.4.

44

This can be done by a flexible processing facility known as biorefinery. According to Kamm et

45

al.5, biorefinery is a system made up from sustainable and environmentally benign technologies

46

to utilise biomass. A biorefinery is used to convert a wide range of biomass into fuels, power

47

and more importantly, value-added products through physical, biological/biochemical and

48

thermochemical conversion processes6.

49

characteristics, many possible processing technologies are available to convert biomass into

50

value-added products7.

As biomass is available in different forms and

51 52

In the last decade, many well established single conversion technologies have been developed for

53

processes that convert biomass into value-added products.

54

produced by transesterification of vegetable oils and methanol in the presence of catalyst. These

55

standalone plants are usually limited in product portfolios, and often result in low and under

56

satisfactory economic performance8. In order to increase the productivity and cost effectiveness,

57

an efficient and sustainable integrated biorefinery has been proposed by Fernando et al.9.

58

Integrated biorefinery is a facility which integrates multiple platforms such as biomass feedstock

59

handling, biomass pretreatments, biomass conversion processes and downstream processing as a

60

whole integrated system10. The waste generated from integrated biorefinery can be minimised,

61

while energy and material recovery can be maximised. As more biomass conversion reactions

For example, biodiesel can be


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 46

62

and technologies are being developed and established, more reaction pathways have to be

63

considered while designing and developing an integrated biorefinery.

64

screening tools are needed to reduce the numbers of available pathways and select the optimum

65

pathways that lead to the production of desired products based on different objectives. Ng et

66

al.11 presented a hierarchical approach to synthesise and screen the potential alternatives for an

67

integrated biorefinery. In the presented work11, two screening tools (evolutionary technique and

68

forward-reverse synthesis tree) are proposed to reduce the process alternatives systematically.

69

Ng12 presented an optimisation approach based on pinch analysis to synthesise an integrated

70

biorefinery with maximised biofuel production and economic performance. Later, automated

71

targeting approach was extended by Tay and Ng13 to handle multiple process parameters. On the

72

other hand, Tay et al.14 adapted the use of the C-H-O ternary diagram to determine the overall

73

performance of the synthesised integrated biorefineries.

Thus, systematic

74 75

In addition to the insight-based approaches, various mathematical optimisation approaches have

76

been developed for the synthesis of integrated biorefinery. For example, Bao et al.15 presented a

77

systematic approach based on technology pathway synthesis to determine the optimum pathway

78

that achieves the highest conversion. Later, Pham and El-Halwagi16 presented a systematic two-

79

stage approach for the synthesis and optimisation of biorefinery configurations. The presented

80

approach is based on the concept of “forward and backward” approach. In the presented work16,

81

forward synthesis has been used to identify the possible intermediates that can be synthesised

82

from biomass while the backward synthesis identifies the necessary species and pathways

83

leading to the desired components identified in the forward synthesis stage. Later, a new concept

84

of a palm oil processing complex which integrates the entire palm oil processing industry to


4

Page 5 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


85

maximise the material recovery has been developed17. Other than economic performance, a

86

number of other aspects such as environmental impact, safety and health impacts as well as

87

supply chain during the synthesis of integrated biorefineries have been considered by different

88

researchers while synthesising and designing the integrated biorefineries. Sammons et al.18

89

developed a framework to improve the product portfolio by evaluating the profitability of

90

different production routes and products based on maximisation of net present value and

91

minimisation of environmental impact. Zondervan et al.19 presented a mixed-integer nonlinear

92

programming (MINLP) model for the design of optimal processing routes for multi-product

93

biorefinery system by considering different feedstock, processing steps, final products and

94

optimisation objectives.

95

approach for optimal biorefinery pathway configuration for a given criterion such as economic,

96

environmental, and safety consideration. The proposed approach can systematically solve a

97

complex synthesis problem by decomposing the main problem into a set of subproblems. El-

98

Halwagi et al.21 introduced an approach that considers the effects of safety and economy into the

99

selection, sizing and supply chain network of a biorefinery. Amundson22 presented an approach

100

for the modelling of biorefinery supply chain economic performance by using discrete event

101

simulation. The approach utilises the outputs from chemical process simulation and optimisation

102

as well as supply chain optimisation in developing an integrated supply chain design framework.

103

The application of the developed supply chain design framework was shown by using an

104

assessment of region specific biorefineries.

105

disciplinary decision support tool for the evaluation of multiple biorefinery conversion

106

technologies and supply chain performance.

107

process systems engineering and supply chain optimisation have been included in the approach

Later, Ponce-Ortega et al.20 developed a disjunctive programming

Later, Sukumara et al.23 proposed a multi-

Design aspects such as feedstock assessment,


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 46

108

to estimate the production cost of chemicals, energy and fuel from various renewable resources

109

in a specific geographic region.

110 111

It can be seen that the focus of current research is mainly on identifying the optimal processing

112

routes that lead to the product but not on the product itself. Minimal effort has been done on

113

addressing the design of optimal products for an integrated biorefinery. However, a lot of

114

research has been done on integrating the process design with product design techniques.

115

Hostrup et al.24 developed a hybrid method which integrates mathematical modelling with

116

heuristic approaches for separation process flow sheet design by considering solvent and process

117

performance simultaneously. Simultaneous consideration of process constraints and property

118

requirements while designing blanket wash solvent has been developed by Sinha and Achenie25.

119

Papadopoulos and Linke26 presented a multi-objective optimisation approach for solvent design

120

with consideration of separation process performance. The complexity of chemical product

121

design problems has been addressed by problem decomposition strategy by different researchers.

122

Karunanithi et al.27 applied the method to crystallisation solvent design which is solved together

123

with performance objectives. The same method is also adapted by Conte el al.28 in developing a

124

virtual product-process design laboratory software for the design of formulated liquid products

125

which is able to design/verify a formulated product. Meanwhile, Bommareddy et al.29 developed

126

an algebraic approach for product design problems by solving two reverse problems. The

127

approach identifies the input molecules’ property targets based on the desired process

128

performance in the first step.

129

identified targets have been determined.

In the second step, the molecular structures that match the

130


6

Page 7 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


131

It is realised that most of the previous works do not consider customer needs in producing value-

132

added products in an integrated biorefinery. Most of the works have focused on process design

133

aspect of designing an integrated biorefinery where the attention is mainly on identifying and

134

designing the optimal processing routes that lead to the product without considering the product

135

design aspect of the integrated biorefinery.

136

biorefinery, the product design aspect has to be considered. Product design can be studied and

137

utilised to address the abovementioned issues. Therefore, this work aims to fill the research gap

138

by synthesising an integrated biorefinery that is able to produce value-added products that meet

139

customer requirements.

140

biorefinery with chemical product design.

In order to synthesise an optimal integrated

This can be achieved by integrating the design of an integrated

141 142

2. COMPUTER-AIDED MOLECULAR DESIGN

143

Chemical product design is the process of choosing the optimal product to be made for a specific

144

application30.

145

approach which is usually based on design heuristics, experimental studies and expert

146

judgements or experiences31. These methods start from the identification of molecules from raw

147

materials, and search for the required and preferred properties from the identified molecules. On

148

the contrary, chemical products can be designed via the top-down approaches.

149

approaches are reverse engineering approaches which begin the chemical product design

150

procedure with identification of product needs to fulfil, followed by the search for molecules that

151

possess properties which can meet the needs32. This can be considered as an inverse property

152

prediction problem where the desired attributes of the chemical product are represented in terms

153

of the physical properties of the molecule. Hence, in inverse property prediction problems, the

The usual practise in searching for new chemical products is a bottom-up

Top-down


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 46

154

objective of the problem is to determine the molecular structure that matches these properties33.

155

Since customer needs are one of the important sources of product specification and requirements,

156

it is required to translate the descriptive product attributes into measurable physical properties in

157

order to design a product34. This process of representing product attributes by using measurable

158

product properties is often done by using various computer-aided molecular design (CAMD)

159

tools.

160 161

CAMD techniques are important for chemical product design for their ability in predicting and

162

designing molecules within a set of predefined target properties35.

163

normally done by utilising property prediction models in predicting molecular properties from

164

structural descriptors36. Some of the commonly used structural descriptors to quantify molecular

165

structure include chemical bonds and molecular geometry

166

utilise property prediction models based on group contribution (GC) methods to verify that the

167

generated molecules possess the specified set of target properties38. By utilising molecular

168

groups as structural descriptors, GC methods estimate the property of a molecule by summing up

169

the contributions from the molecular groups in the molecule according to their appearance

170

frequency39. A general representation of property prediction model based on GC methods is

171

illustrated with eq 1.

37

. Most of the CAMD techniques

f ( X ) = ∑ N i C i + w∑ M j D j + z ∑ O k E k i

j

Property estimation is

k

(1)

172

In eq 1, f(X) is a function of the property X, w and z are binary coefficients depending on the

173

levels of estimation, Ni, Mj, Ok are the number of occurrence of first, second and third-order

174

molecular group correspondingly and Ci, Dj, Ek are contribution of first, second and third-order


8

Page 9 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


175

molecular groups subsequently. In addition to GC methods, property prediction of molecules

176

can be done based on different topological indices (TIs).

177

calculated based on the principles in chemical graph theory40. Chemical graph theory represents

178

the atoms and bonds in a molecule as vertices and edges in a graph. Based on the molecular

179

structure, important information such as total number of atoms, bonding between the atoms and

180

types of atoms and bonds can be determined. TIs describe a molecular graph as an index by

181

utilising the interactions among different atoms/molecular groups based on their connectivity and

182

the effects due to these interactions. The indices can be used to correlate the chemical structure

183

to physical properties of the molecule. As TI for a molecular graph is consistent, it is useful in

184

characterising a chemical structure40.

TIs are molecular descriptors

185 186

Due to their ability in predicting and designing molecules within a set of predefined target

187

properties, CAMD techniques are widely used during the pre-design stage for the screening of

188

possible molecular structures. Numerous CAMD techniques have been developed and applied in

189

the design of numerous chemical products. These includes the design of crystallisation solvent27,

190

environmental benign blanket solvent25, polymer products41, alternative refrigerants42 and

191

working fluid for organic Rankine cycle43. In some chemical product design problems, the

192

desired target properties could not be estimated by using a single class of property prediction

193

model. Hence, different classes of property prediction models are required for the estimation of

194

different target properties in the design problem.

195

exclusive for different property prediction models. While GC methods estimate property by

196

summing up the contributions from the molecular groups in the molecule according to their

197

appearance frequency, estimations of TIs involve the operations on vertex-adjacency matrix44.

However, mathematical formulations are


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 46

198

Therefore, it is difficult to utilise these different models by using a similar calculation method41.

199

This difficulty is addressed by utilising molecular signature descriptor as structural descriptors 45.

200

Signature is a systematic coding system to represent the atoms in a molecule by using the

201

extended valencies to a pre-defined height. Eq 2 represents the relationship between a TI and its

202

signature. TI ( G ) = k hα G ⋅ TI [ root ( h Σ )]

(2)

203

Here, hαG is the occurrence number of each signature of height h and TI[root (hΣ)] is the TI

204

values for each signature root and k is a constant specific to TI. Signature of a molecule can be

205

obtained as a linear combination of its atomic signatures by representing a molecule with atomic

206

signature. By writing a molecule in terms of signature, GC methods and TIs with different

207

mathematical formulations can now be expressed and utilised on a common platform. The

208

application of molecular signature is important for chemical product design problem which

209

involve multiple property targets which are required to be estimated by different classes of

210

property prediction models46. In order to utilise molecular signature descriptors in a molecular

211

design problem, signature-based molecular design technique developed by Chemangattuvalappil

212

et al.46 is applied in this work.

213 214 215

3. METHODOLOGY: TWO-STAGE OPTIMISATION APPROACH FOR SYNTHESIS OF OPTIMAL BIOCHEMICALS

216

In order to ensure the optimum pathways that convert biomass into biochemical products with

217

optimised properties of interest, a novel two-stage optimisation approach has been developed by

218

integrating molecular design technique with synthesis approach for integrated biorefineries. In


10

Page 11 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


219

the first stage, a chemical product design problem is formulated to determine the favourable

220

products that meet customer requirements. Based on the identified products in the first stage, the

221

optimum conversion pathways that convert biomass into the optimal products are determined in

222

the second stage based on superstructure mathematical optimisation approach. Figure 1 shows

223

the integration of synthesis of integrated biorefinery with molecular design.

224 225

Figure 1. Integration of synthesis of integrated biorefinery with product design.

226

As shown in Figure 1, in order to utilise this two-stage optimisation approach, the optimum

227

biochemical products that meet the customer requirements are first identified in the first stage of

228

the optimisation approach. Based on the customer requirements, the product needs are translated

229

into a set of property constraints which represent the product specifications.

230

constraints are applied to guarantee complete formation of the molecular structure of the product.

231

The optimal products which satisfy property and structural constraints are identified by utilising

Structural


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 46

232

the signature based molecular design technique developed by Chemmangattuvalappil et al.46

233

Once the optimal products are determined, identification of the optimum pathways that convert

234

biomass into the optimal products can then be determined in the second stage of the optimisation

235

approach. Based on the available conversion pathway and technologies, a superstructure is

236

constructed as a representation of integrated biorefinery.

237

mathematical optimisation approach, optimal conversion pathways based on different design

238

goals such as economic potential, production yield, environmental impact etc. can be determined

239

in this stage. By combining the strengths from both sides, this two-stage optimisation approach

240

is able to determine the optimum conversion pathways that convert biomass into biochemical

241

products that meet customer requirements.

242

identification of optimal biochemical as well as the conversion pathways is represented in a

243

flowchart as shown in Figure 2. Details of the proposed two-stage optimisation approach are

244

discussed in the following sub-sections.

By using the superstructural

The step by step procedure involved in the

245


12

Page 13 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


246 247

Figure 2. Procedure for solving a two-stage optimisation problem for the synthesis of optimal

248

biochemical product.


13


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 46

249

3.1.STAGE 1: DESIGN OF OPTIMAL BIOCHEMICAL

250

In this stage, the optimal biochemical product is designed by utilising signature based molecular

251

design techniques. Note that the procedure is designed specifically for product design problems

252

where different classes of property prediction models are used and the molecular structure of the

253

product is represented by using molecular signature descriptor. The details for the design of

254

optimal biochemical product are discussed as follows.

255 256

3.1.1. DEFINE OBJECTIVE FOR THE PRODUCT DESIGN PROBLEM

257

As shown in Figure 2, the first step in solving the product design problem is to define the

258

objective. This is done by identifying the product needs. These product needs can be extracted

259

from the operating conditions of an industrial process or from customer requirements. The

260

product needs cover the physical properties which are responsible for a particular functionality of

261

the product as well as properties that make sure that the product fulfils the environmental and

262

safety regulations. For example, in order to design an effective refrigerant, the performance of

263

the refrigerant should be high while the power requirement for the refrigerant is preferred to be

264

low. In addition, the refrigerant should not be harmful to the environment and should be safe to

265

use. Hence, the objective of the design problem can be the optimisation of any target property or

266

performance criterion.

267 268 269

3.1.2. IDENTIFY TARGET PROPERTIES AND DETERMINE TARGET PROPERTY RANGES


14

Page 15 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


270

Once the product needs and the objective of the product design problem have been identified, the

271

identified descriptive product needs are translated into measurable physical properties. For

272

example, during the design of a refrigerant, the performance of the refrigerant can be expressed

273

as volumetric heat capacity while the power requirement of the refrigerant can be measured as

274

viscosity. The volumetric heat capacity should be high so that the amount of refrigerant required

275

is reduced for the same refrigeration duty whilst the viscosity is preferred to be low to achieve

276

low pumping power requirement. On the other hand, ozone depletion potential (ODP) and

277

median lethal dose/concentration (LD50/LC50) can be used to ensure that the designed refrigerant

278

is environmentally benign and safe to be used. These target properties are then expressed as a

279

property specifications, which can be written as a set of property constraints bounded by an

280

upper and lower limit. For example, while designing a gasoline blend, the Reid vapour pressure

281

is designed to fall within 45 kPa and 60 kPa and the desired viscosity should fall within 0.30 cP

282

and 0.60 cP. The property specifications for a product design problem can be generalised and

283

shown in eq 3.

v Lp ≤ V p ≤ v Up

∀p ∈ P

(3)

284

Here, p is the index for the target property, Vp is the target property value, v Lp is the lower limit

285

and v Up is the upper limit for product target property. By following eq 3, an optimal solution is

286

identified within the predefined target property ranges while solving a product design problem.

287 288

3.1.3. IDENTIFY APPROPRIATE PROPERTY PREDICTION MODELS


15


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 46

289

After the identification of target properties from the product needs, property prediction models

290

which estimate the target properties of the product can be identified. This work utilises signature

291

based molecular design techniques which allow the utilisation of different classes of property

292

prediction models. Hence, property prediction models developed from the GC method or TIs

293

can be utilised for the prediction of target properties. The target properties can be written as

294

functions of property prediction models developed from GC method or TIs, as shown in the

295

following equation. θ p = f (GC / TI )

∀p ∈ P

(4)

296

In eq 4, θ is the property function corresponding to target property p. For target properties where

297

property prediction models are unavailable, models which combined experimental data and

298

available property prediction models can be developed to estimate the respective property.

299 300

3.1.4. SELECT MOLECULAR BUILDING BLOCKS

301

Suitable molecular building blocks for the product design problem are determined in this step.

302

The molecular building blocks have to be chosen such that the properties and molecular structure

303

of the new product are similar to the available product from where the molecular building blocks

304

are selected. It is assumed that by designing a new molecule with the chosen molecular groups

305

as building blocks, the designed product will possess the properties and functionalities of the

306

desired product. For example, in order to design an alcohol solvent, molecular group –OH is

307

chosen as one of the molecular building blocks as it is the functional group of alcohol. As the


16

Page 17 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


308

product design methodology employs signature based molecular design technique, signatures

309

corresponding to the selected molecular groups are then generated.

310 311 312

3.1.5. FORM

THE

PROPERTY

PREDICTION

MODELS

AS

NORMALISED

PROPERTY OPERATORS

313

The next step is to express the property prediction models as normalised property operators.

314

Normalised property operators are dimensionless property operators, which are required so that

315

different target properties can be expressed and compared together on the same property

316

platform47. According to Shelley and El-Halwagi47, property operators are functions of the

317

original properties tailored to obey linear mixing rules. Hence, a property operator will follow

318

simple linear mixing rules regardless of the linearity of the original property.

319

specifications in eq 3 can be written as normalised property operators as shown in eq 5.

Ω pL ≤ Ω p ≤ Ω pU

∀p ∈ P

Property

(5)

320

Here, Ωp is the normalised property operator for property p, Ω Lp is the lower limit and Ω pU is the

321

upper limit for the normalised property operator. As the signature based molecular design

322

technique is employed in this developed methodology, normalised property operators are used to

323

express molecules as linear combinations of atomic signatures.

324 325

3.1.6. DEVELOP STRUCTURAL CONSTRAINTS


17


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 46

326

Apart from satisfying the properties constraints, the targeted molecule should have a feasible and

327

stable chemical structure which is formed from a collection of molecular signatures46. Hence,

328

structural constraints are generated based on graph theory principles in order to enable the

329

formation of a complete molecule. Firstly, it is ensured that the generated molecule is complete

330

without any free bond in the structure40. This is illustrated with eq 6. n2 n3 n4 N Mi NTi  N S   1 N Di  + + + = + + + x 2 x 3 x 4 x 2 x x x  ∑ ∑ ∑ ∑ i i i i  ∑ i 2 ∑ i ∑ i ∑ xi  − 1 + R  i =1 n1 n2 n3 i =0 i =0 i =1   i =1  n1

(6)

331

In eq 6, n1, n2, n3 and n4 are the number of signatures of valency one, two, three and four

332

respectively, Ns is the total number of signatures in the molecule, NDi, NMi and NTi are the

333

signatures with one double bond, two double bonds and one triple bond, R is the number of

334

circuits in the molecular graph. In addition, it must be ensured that the signatures in the solution

335

set is consistent. In order to ensure the consistency of signatures, one of the properties of

336

digraph known as handshaking dilemma is used. By following handshaking dilemma, the sum of

337

the in-degrees of all the vertices of a digraph will be equal to the sum of their out-degrees48. This

338

can be shown in eq 7.

∑ (l

i

→ l j )h =

∑ (l

j

→ l i )h

(7)

339

In eq 7, (li → lj)h is one colouring sequence li → lj at a level h. Eq 7 must be obeyed for all

340

colour sequences at each height. This guarantees that the number of bonds in each signature will

341

match with the bonds in the other signatures. This is explained in Figure 3.

342


18

Page 19 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


343 344

Figure 3. Explanation of handshaking dilemma.

345

In Figure 3, the presented signatures consist of signatures with carbon atom as root atom with

346

one (C1), two (C2), three (C3) and four (C4) neighbouring atom(s). As shown in Figure 3, the

347

edges of the signatures have the colours of 1 and 2. The reading of colouring sequence for

348

signature S1 will be 1→2, 2→2 and 2→1 for signature S2, 2→3 and 2→1 for signature S3 and

349

2→4 and 2→1 for signature S4. According to the handshaking dilemma, each colour sequence

350

(e.g. 1→2) has to be complemented with another colouring sequence in reverse order (e.g. 2→1)

351

to ensure linkage and consistency of the signatures. By obeying the structural constrains, a

352

complete molecular structure without any free bonds can be formed from the combination of

353

signatures. The using of molecular signatures in molecular design and the connectivity rules of

354

signatures are discussed in detail by Chemmangattuvalappil and Eden 49.

355 356 357

3.1.7. GENERATE FEASIBLE SOLUTIONS BY SOLVING THE MATHEMATICAL MODEL

358

Now, mathematical model can be formulated to solve the molecular design problem, where the

359

objective function is to maximise/minimise the preferred target property Ωp, as shown in eq 8.

Maximise/Minimise Ωp

(8)


19


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 46

360

For example, in order to design a refrigerant with high volumetric heat capacity, the objective

361

function for the mathematical model can be formulated to maximise the volumetric heat capacity

362

while fulfilling other target property ranges. On the other hand, the objective function can be

363

formulated to minimise the viscosity in order to achieve low pumping power requirement.

364

Subject to property and structural constraints, the objective function is solved to determine the

365

solution for the product design problem. The solution is obtained in terms of the number of

366

appearances of signatures.

367 368

3.1.8. ENUMERATE THE MOLECULAR STRUCTURE

369

With the signatures obtained by solving the design problem, a molecular graph can now be

370

generated from the set of signatures based on the graph signature enumeration algorithm by

371

Chemmangattuvalappil and Eden49.

372

structures are generated from the list of signatures, and the names of the new molecules are

373

identified. Figure 4 shows an example for the enumeration of molecular structure for propan-1-

374

ol.

By using the graph enumeration algorithm, molecular

OH H3C

375 376

Figure 4. Enumeration of structure for propan-1-ol.


20

Page 21 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


377

Four signatures are presented in Figure 4. Signature C2(C2(CC)O1(C)) has signatures C2(CC)

378

and O1(C) as its neighbouring signatures; signature C1(C2(CC)) has signature C2(CC) as its

379

neighbouring signature; signature O1(C2(OC)) has signature C2(OC) as its neighbouring

380

signature while signature C2(C1(C)C2(OC)) has signatures C1(C) and C2(OC) as its

381

neighbouring signatures. By using the graph enumeration algorithm, these signatures can be

382

enumerated and the molecular structure of propan-1-ol can be formed. Detailed instructions for

383

the enumeration of molecular structure are provided in the work of Chemmangattuvalappil and

384

Eden49.

385 386

3.2.STAGE 2: DESIGN OF INTEGRATED BIOREFINERY

387

Once the optimal biochemical products which meet the customer requirements are identified in

388

the first stage, the optimal biomass conversion pathways to produce the biochemical products are

389

identified in the second stage of the optimisation approach.

390

superstructural mathematical optimisation approach. First, the objective of this second stage of

391

the optimisation approach is defined, as shown in Figure 2. The objective for the design of

392

integrated biorefinery can be aimed to maximise the yield of the desired product. Other than

393

maximising the yield of the desired product, maximum economic performance can also be

394

targeted as one of the design objectives during the configuration of the integrated biorefinery.

395

Next, information of all the possible conversion pathways and technologies that convert biomass

396

to intermediates and from intermediates to the final products are compiled. Based on the

397

considerations such as product manufacturability and process feasibility, the compiled

398

conversion pathways are screened to filter the redundant processes and select the feasible ones to

399

be included in the construction of superstructure. Now, a superstructure which includes all the

This is done by utilising the


21


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 46

400

conversion pathways and technologies can then be constructed as the representation of an

401

integrated biorefinery, as shown in Figure 5.

402 403

Figure 5. Superstructure as representation of integrated biorefineries.

404

Figure 5 illustrates a general superstructure of an integrated biorefinery with biomass feedstock b

405

converted through pathways q to produce intermediates s, and further processed via pathways q’

406

to produce products s’. The mathematical model which relates the flow of biomass through

407

different conversion pathways to produce the products can now be formulated. This is explained

408

and discussed as follows.

409 410

Biomass feedstock b can be split to biomass conversion pathway q with their respective flowrate

411

FbqI by using eq 9.


22

Page 23 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


BbBio = ∑ FbqI

∀b

(9)

q

412

In eq 9, BbBio is the available total flowrate of biomass feedstock b. After going through the

413

biomass conversion pathway q, intermediate s is generated based on the conversion rate of

414

Inter conversion pathway q, Rbqs . This gives a total intermediate production rate of Ts , as shown in

415

eq 10.

I

(

I TsInter = ∑∑ FbqI Rbqs q

)

∀s

(10)

b

416

Subsequently, the intermediate s is further converted to product s’ via upgrading pathway q’.

417

The splitting of total production rate of intermediate TsInter to all possible pathway q’ with

418

flowrate Fsq' can be represented by eq 11.

II

TsInter = ∑ FsqII'

∀s

(11)

q'

419

The total production rate of product s’, TsProd can be determined based on the given conversion '

420

rate of pathway q’, Rsq's' via eq 12.

II

(

TsProd = ∑∑ FsqII' RsqII 's' ' q'

)

∀s'

(12)

s

421

By following eqs 9 – 12, the material balance of the biomass, intermediates and final products

422

can be performed. Thus, an integrated biorefinery can be represented by using the developed

423

superstructure.

424


23


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 46

425

After the construction of superstructure, methods to measure the process performances are

426

identified. The objective of this second stage of the optimisation approach is to determine the

427

optimal conversion pathways that convert biomass into the optimal products identified in the first

428

stage of the methodology. As mentioned earlier, the optimality of the conversion pathways can

429

be aimed to maximise the yield of the desired product. This can be determined by using the

430

equation as shown in eq 13. Maximise T sProd '

(13)

431

In addition, maximum economic performance can also be aimed as one of the design goals

432

during the configuration of the integrated biorefinery. Economic performance can be defined

433

with the following equations. Maximise GP Total

(14)

Prod GPTotal = ∑TsProd − ∑BbBioGbBio − TAC ' Gs' s'

TAC = TACC + TAOC TACC =

∑∑F G

(16)

∑∑F

I Cap bq bq CRF +

q

b

II Cap sq'Gsq' CRF

q'

s

Opr TAOC = ∑∑ FbqI Gbq +∑∑ FsqII' GsqOpr' q

(15)

b

b

q'

s

(17)

(18)

434

In eqs 14 – 18, GPTotal is the gross profit of the overall integrated biorefinery configuration, TAC

435

is the total annualised cost, TACC is the total annualised capital cost, TAOC is the total

436

Prod Bio annualised operating cost, CRF is the capital recovery factor, Gs' is the cost of product s’, Gb


24

Page 25 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Cap

Cap

437

is the cost of biomass feedstock b, Gbq is the capital cost for conversion of biomass b, Gsq' is

438

the capital cost for conversion of intermediate s, Gbq is the operating cost for conversion of

439

biomass b, Gsq' is the operating cost for conversion of intermediate s.

Opr

Opr

440 441

Now, a mathematical model for the design of integrated biorefinery can be formulated. By

442

solving the developed mathematical model based on different objective functions, the optimal

443

conversion pathways that lead to the desired optimal products can be determined in this stage.

444

For cases where the conversion pathway leads to the formation of products as a mixture of

445

several components, separation processes are included. These separation processes are taken

446

into account to refine and separate the final product from the other by-products based on the

447

results obtained from the design of product in stage 1 of the methodology. With the available

448

information, different objectives (e.g. economic performance, environmental impact, process

449

safety etc.) can be considered and included in the development of the superstructure. For

450

situations where the identified products cannot be produced in a feasible method (e.g. in terms of

451

economic potential, manufacturability etc.), an iterative identification of optimal product and its

452

conversion pathways is required, as shown in Figure 2. In such situations, the overall design

453

problem has to be repeated from Step 2 of the product design problem, where the target property

454

and target property ranges are re-evaluated. Other than identifying the optimal product and

455

feasible optimal conversion pathways, the iterative process also provides comparison and

456

tradeoff between multiple options of the products and conversion pathways.

457

pathways can be generated based on different design goals such as manufacturability of the

458

product and economic feasibility of the processing routes. For instance, when the optimal

Alternative


25


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 46

459

product in terms of customer requirements cannot be produced economically, iteration of the

460

design problem can be applied to identify the best product which can be manufactured in a cost

461

effective manner.

462 463

The developed methodology decomposes the integrated product and process design problem into

464

two design problems and solves them sequentially in two stages. This approach offers the

465

identification of optimal biochemical products in terms of target product properties as well as

466

optimal conversion pathways that convert the biomass into the biochemical products. Although

467

a feedback loop might be required, the computational complexity of this developed approach is

468

lower compared with solving the product and process design problem simultaneously. An

469

algebraic approach for the simultaneous solution of process and molecular design problems

470

developed by Bommareddy et al. (2010) can be utilised to solve the product and process design

471

simultaneously. However, simultaneous solution is not considered in this work. In addition, it is

472

aware that the composition of biomass is complex, and the conversion reactions involved are

473

difficult to be defined straightforwardly. Please note that this approach serves as a general

474

representation and idea to integrate the synthesis of integrated biorefinery with product design.

475

Hence, for ease of illustration, side reactions, additional reactants required and intermediate

476

products with complex chemical structure are not considered in the presented approach.

477 478

This developed novel methodology provides the identification of optimal biochemical products

479

in terms of target product properties as well as the selection of optimal conversion pathways that

480

convert biomass to the biochemical products. By utilising molecular signature descriptor, this


26

Page 27 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


481

methodology offers the simultaneous utilisation of different classes of property prediction

482

models for the design of optimal biochemical products. In addition, an integrated biorefinery

483

which produces biochemical products which fulfil the required product needs can be synthesised

484

by using the developed methodology. In order to show the efficacy of the methodology, a case

485

study is presented.

486 487

4. CASE STUDY

488

A product design problem of producing bio-based fuel from biomass is solved to illustrate the

489

proposed two-stage optimisation approach. In the first stage, molecular signatures are used to

490

represent different classes of property prediction models in designing bio-based fuel with optimal

491

target properties. In the second stage, the optimum conversion pathways in terms of different

492

production objectives that convert biomass into the designed bio-based fuel are identified. In

493

order to demonstrate the efficacy of the methodology, the conversion pathways of an integrated

494

biorefinery are synthesised for two scenarios: conversion pathways for maximum product yield

495

and conversion pathways for maximum economic potential. For the ease of illustration, the bio-

496

based fuel is targeted and designed as a single component bio-based fuel in this case study.

497 498

4.1.DESIGN OF OPTIMAL PRODUCT

499

The bio-based fuel is designed in terms of different product needs. The first is engine efficiency,

500

which can be measured as octane rating. Octane rating is a measure of a fuel’s ability to resist

501

auto-ignition and knock in a spark-ignited engine conditions. Higher octane rating helps vehicle

502

to run smoothly and keep the vehicles’ fuel system clean for optimal performance. In addition,


27


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 28 of 46

503

octane rating is the main property that sets the price of bio-based fuel. In this case study,

504

research octane number (RON) is used as the indication of octane rating of the fuel.

505

Furthermore, it is very important for a bio-based fuel to be safe to use.

506

flammability characteristics and toxicity of the bio-based fuel are also taken into consideration,

507

which are measured as flash point (Tf) and lethal concentration (logLC50) respectively.

508

Meanwhile, latent heat of vaporisation (Hv) and viscosity (η) of the bio-based fuel are the other

509

target properties that are considered during the product design stage to ensure the consistency of

510

the fuel flow as well as the stability of the bio-based fuel. Since higher RON bio-based fuel is

511

desirable as it enables improved engine efficiency, the objective of this case study is to design a

512

bio-based fuel with maximised RON. Hence, RON is target property to be optimised while Tf,

513

logLC50, Hv and η are property constraints to be fulfilled. The target property ranges for each of

514

the target property are shown as follows.

515

Table 1. Upper and Lower Bounds for Solvent Design.

Property

Therefore, the

Target property range v Lp

v Up

516

Tf (K) 230.00 350.00 logLC50 1.00 2.00 Hv (kJ/mol) 25.00 55.00 η (cP) 0.10 3.00 After identifying the target properties for the product, property prediction models for each target

517

properties are identified. In order to illustrate the ability of the methodology to utilise different

518

classes of property prediction models in a design problem, property prediction models based on

519

GC methods and connectivity index (CI), which is one of the commonly used TIs are chosen to

520

estimate the target properties. For RON, a reliable group contribution is available50. This is

521

illustrated with eq 19.


28

Page 29 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


2

3

4

          f ( X ) = a + b ∑ Ni Ci  − c ∑ N i Ci  + d  ∑ N i Ci  + e ∑ N i Ci  + f / ∑ N i Ci   i   i   i   i   i 

(19)

522

In eq 19, a, b, c, d, e, f are the correlation constants. Since the values of the constant c, d, e, f are

523

relatively insignificant, only the first two terms of eq 19 will be considered for this case study.

524

CI of order zero is available for the prediction of Tf

525

in eq 21.

51

as shown in eq 20 and logLC5052 as shown

T f = 33.638 0 χ + 164.386

( )

(20)

log LC50 = 4.115 − 0.762(0χ )

(21)

526

where 0χ is the zeroth order CI. GC model developed by Marrero and Gani53 is utilised to predict

527

Hv while GC model developed by Conte et al.54 is used for the estimation of η. These property

528

prediction models are shown in eqs 22 and 23. Hv0 in eq 22 is an adjustable parameter.

H v = H v0 + ∑ N i Ci + w∑ M j D j + z ∑ Ok Ek i

j

k

lnη = ∑ N i Ci + w∑ M j D j + z ∑ Ok Ek i

j

k

(22)

(23)

529

With the identification of property prediction models, the next step is to select the suitable

530

molecular building blocks for the design problem. As the objective of this design problem is to

531

design a bio-based fuel, the target molecule category is identified as alkanes. Therefore, only

532

carbon (C) and hydrogen (H) atoms are considered. As molecular signature descriptor is utilised

533

in solving the chemical product design problem, only signatures with single bonds are considered

534

in this design problem to design the bio-based fuel. Signatures of height one is required since

535

property prediction models of zeroth order CI are utilised. The generated signatures can be


29


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 30 of 46

536

classified into first order groups of carbon with zero (C-), one (CH-), two (CH2-) and three (CH3-

537

) hydrogen atoms. For signature C-, as it is bonded with zero hydrogen atoms, it can be

538

connected to four other matching signatures. Same concept applies for others signatures as well,

539

where signature CH- can be connected to three other matching signatures, signature CH2- can be

540

connected to two other matching signatures and signature CH3- can be connected to one

541

matching signature. The generated signatures for the design problem are shown in Table 2.

542

Table 2. List of signatures.

543

No. Signature 1. C(C) 2. C(CC) 3. C(CCC) 4. C(CCCC) The next step is to transform the property prediction models into their respective normalised

544

property operators.

545

normalised property operators.

546

property ranges are shown in Table 3.

547

Table 3. Normalised property operators and normalised target property ranges for the design

548

problem.

Property prediction models as shown in Eqs 19 – 23 are written as Normalised property operators and the normalised target

Normalised target property range Property

Ωp

RON

Ω RON =

Tf (K)

ΩT f =

logLC50 Hv (kJ/mol) η (cP)

Ωlog LC50 ΩH v

RON − 103.6 0.231

T f − 164.386

33.638 4.115 − log LC50

Ω pL

Ω pU

To be maximised 1.95

5.52

0.762 = H v −11.733

2.78

4.09

13.27

43.27

Ωη = lnη

-2.30

1.10


30

Page 31 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


549

Now, the molecular design problem can be written as a mixed integer linear programming

550

(MILP) model. As the objective of this case study is to design a bio-based fuel with maximised

551

RON, objective function for the case study can be written as shown in eq 24.

Maximise Ω RON

(24)

552

To ensure the formation of a complete molecule with no free bonds in the final molecular

553

structure, structural constraints are employed. Eqs 6 and 7 are modified according to the case

554

study and used to guarantee that a complete structure can be generated. The objective function

555

eq 24 can now be solved together with property constraints and structural constraints to generate

556

the optimal bio-based fuel. Commercial optimisation software LINGO version 13, in Asus

557

N56V with Intel Core i5-3210M (2.50 GHz) processor and 4 GB DDR3 RAM is used to solve

558

the MILP model.

559

solutions is 0.1 s. The solution obtained is the bio-based fuel with maximised RON which fulfils

560

other target properties. Four additional solutions are generated by using integer cuts. Integer

561

cuts work by adding additional constraints in the mathematical programming model to ensure the

562

generated solution (in terms of combination of molecular signatures) will not appear again when

563

the model is solved. This step may be continued until no feasible solution can be found. The list

564

of possible combinations of signatures is shown in Table A in Supporting Information.

565

Molecular graphs can be generated from the signatures obtained in Table A based on the graph

566

signature enumeration algorithm by Chemmangattuvalappil and Eden49.

567

molecules is performed on all five solutions. The list of products and their respective properties

568

are given in Table 4 while the generated molecular structures of the bio-based fuel are shown in

569

Table 5.

The average Central Processing Unit (CPU) time for the generation of

Enumeration of


31


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

570

Page 32 of 46

Table 4. Possible design of bio-based fuel.

Sol.

Name

A B C D E

2,2,3,3-tetramethylbutane 2,2,4,4-tetramethylpentane 2,2,3-trimethylbutane 2,2,3-trimethylpentane 2,3,4-trimethylpentane

RON 105.91 103.96 103.64 101.69 100.80

Tf (K) 282.12 294.01 266.60 278.49 277.61

Property logLC50 1.45 1.18 1.80 1.53 1.55

Hv (kJ/mol) η (cP) 33.19 0.58 38.10 0.72 30.43 0.37 35.34 0.45 36.85 0.31

571 572

Table 5. Molecular structures for the possible designs of bio-based fuel. Sol.

1

2

Name

Molecular structure

2,2,3,3tetramethylbutane

2,2,4,4tetramethylpentane

3

2,2,3-trimethylbutane

4

2,2,3-trimethylpentane

5

2,3,4-trimethylpentane

573

In this case study, the bio-based fuel is designed to possess maximised RON while having other

574

properties fall within the preferred target property ranges. It should be noted that these target

575

ranges represent the product specification that customers desire and prefer. From Table 4, it can


32

Page 33 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


576

be seen that the solutions are arranged with decreasing RON. The optimal bio-based fuel

577

designed for the case study is 2,2,3,3-tetramethylbutane with RON of 105.91 while the fifth best

578

bio-based fuel identified is 2,3,4-trimethylpentne with RON of 100.80. It can be seen from Table

579

4 that all of the bio-based fuel properties fall within the target properties range, as shown in

580

Table 1.

581 582

4.2.SELECTION OF OPTIMAL CONVERSION PATHWAY

583

With the identification of optimal bio-based fuel in the first stage of the methodology, the

584

optimal conversion pathways that convert biomass into the bio-based fuel can now be identified

585

in the second stage of the methodology. In this case study, palm-based biomass known as empty

586

fruit bunches (EFB) is chosen as feedstock of the integrated biorefinery. The lignocellulosic

587

composition of the EFB is shown in Table 6.

588 589 590

Table 6. Lignocellulosic composition of EFB.

591

Components Composition (% of Dry Matter) Lignin 39.00 Cellulose 22.00 Hemicellulose 29.00 From the first stage of the methodology, the optimal product identified is 2,2,3,3-

592

tetramethylbutane, which is an alkane with carbon number 8. Hence, it is known that the end

593

product of the integrated biorefinery is alkane with carbon number of 8. For illustration purpose,

594

the end products alkanes of the integrated biorefinery are represented as straight-chain products

595

without considering the formation of isomers. For example, the optimal bio-based fuel, 2,2,3,3-

596

tetramethylbutane is represented as straight-chain alkane with carbon number of 8 (Alkane C8) in


33


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 34 of 46

597

this case study. After the screening of possible conversion pathways and technologies, selected

598

conversion pathways that produce alkanes from biomass are compiled and shown in Table B in

599

the Supporting Information. Figure 6 shows a superstructure developed based on the conversion

600

pathways in Table B.

601

categorised into reactions from biochemical and thermochemical platforms. It is noted that the

602

developed superstructure can be revised to include more conversion pathways and technologies

603

in synthesising an integrated biorefinery.

As shown in Figure 6, the selected conversion pathways can be

604 605

Figure 6: Production of additives made from alkane and alcohol from lignocellulosic biomass.


34

Page 35 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


606

In order to demonstrate the efficacy of the developed methodology in accommodating different

607

production objectives, two scenarios of different production objectives are considered in

608

synthesising the integrated biorefinery:

609

1. Design for maximum product yield

610

2. Design for maximum economic potential

611

The market prices of the products and biomass feedstock as well as the capital and operating

612

costs for each conversion pathway are tabulated in Tables C and D respectively in the Supporting

613

Information. In this case study, other than the revenue generated by producing the bio-based

614

fuel, the revenue obtained from the generation of by-products along with the product is included

615

in the overall economic potential of the integrated biorefinery as well. For the simplicity of

616

illustration, the impact of market saturation on fluctuation of products revenue is not considered

617

in this case study. The list of capital costs provided in Table D are the capital costs for nominal

618

capacity of each conversion technology available in the market assuming 50000 tonnes per year

619

of initial biomass feed. Hence, the flow rate determined by the mathematical model is the

620

operating flow rate into the selected conversion technology with a fixed nominal capacity. It

621

should be noted that the prices of the products, feedstock as well as the capital and operating

622

costs for each conversion technology can be revised from time to time to provide an up-to-date

623

economic analysis. With a feed of 50000 tonnes per year of EFB, superstructural optimisation

624

model is formulated and solved for the following scenario.

625 626

Scenario 1: Design for maximum product yield


35


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 36 of 46

627

In this scenario, an integrated biorefinery is synthesised by solving the optimisation model using

628

the optimisation objective in eq 25.

629

tetramethylbutane is represented as straight chain Alkane C8 in this case study.

630

commercial optimisation software with similar software and hardware specification is utilised in

631

solving the mathematical model for both scenario 1 and 2. The average CPU time for the

632

generation of solutions is 0.1 s for both scenarios.

Note that the optimum bio-based fuel 2,2,3,3Similar

Prod Maximise TAlkane C

(25)

8

633

Based on the obtained result, the maximum yield for Alkane C8 is 5645.74 t/y. Along with

634

Alkane C8, alkanes with different carbon number are produced as by-products in the integrated

635

biorefinery. The GPTotal for the scenario is found to be U.S. $23.63 million (per annum). The

636

conversion pathways selected for the scenario is illustrated in the synthesised integrated

637

biorefinery as shown in Figure 7.

638 639

Figure 7: Flow diagram of synthesised integrated biorefinery (maximum product yield).


36

Page 37 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


640

From Figure 7, it can be seen that Alkane C8 is produced from biomass in the conversion

641

pathway sequence of pyrolysis, Fischer-Tropsch process 1 and 2 followed by fractional

642

distillation of alkanes, which are all thermochemical pathways. It is worth mentioning that

643

specific separation processes that suit the identified product can be chosen and included in the

644

integrated biorefinery to refine and separate the final product from the other generated by-

645

products. Hence, separation processes for alkanes are chosen based on the results of the product

646

design identified in stage 1 of the methodology. The performance of the separation processes are

647

then taken into consideration in identifying the product yield and economic potential of the

648

overall conversion pathway.

649 650

Scenario 2: Design for maximum economic potential

651

In this scenario, an integrated biorefinery configuration with maximum economic potential is

652

determined by solving the optimisation objective as shown in eq 26.

Maximise GP Total

(26)

653

Based on the generated optimisation result, the maximum GPTotal for the scenario is identified to

654

be U.S. $24.07 million (per annum) with the annual production for Alkane C8 of 2831.00 t. As

655

the objective of this scenario is to synthesise an integrated biorefinery with maximum economic

656

potential, alcohols are produced and sold as by-products together with the main product Alkane

657

C8 and other alkane by-products. The conversion pathways chosen for the scenario is presented

658

in the synthesised integrated biorefinery as shown in Figure 8.


37


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 38 of 46

659 660

Figure 8: Flow diagram of synthesised integrated biorefinery (maximum economic potential).

661

From Figure 9, it can be seen that alcohols are produced from biomass in the conversion pathway

662

sequence of ammonia explosion, Organosolv separation, dehydration of sugars, hydrogenation of

663

furfural and hydrogenation of THFA 1. Some of the generated alcohol is further converted into

664

alkanes via dehydration of alcohols 3. Main product Alkane C8 is produced from fractional

665

distillation of alkanes, which are produced from pyrolysis of biomass followed by Fischer-

666

Tropsch process 2 together with dehydration of alcohols 3. The selected conversion pathways

667

consist of both biochemical and thermochemical pathways. The comparison of the results

668

generated for scenario 1 and 2 are summarised in Table 7.

669 670


38

Page 39 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

671


Table 7. Comparison of results for scenario 1 and 2. Scenario GPTotal (U.S $/y) Alkane C8 production rate (t/y) Alkane by-product production rate (t/y) Alcohol production rate (t/y)

1 23.63 × 106 5645.74 41354.26 0.00

2 24.07 × 106 2831.00 22697.58 6908.48

672 673

From Table 7, it can be seen that the production rate of Alkane C8 in scenario 1 is higher

674

compared with scenario 2.

675

conversion pathways which produce maximum yield of Alkane C8. In scenario 2, it is noticed

676

that significant amount alcohols are generated together with main product Alkane C8 and other

677

alkane by-products. Although the production rate of Alkane C8 in scenario 2 is lower compared

678

to scenario 1, the GPTotal generated is higher as the objective of scenario 2 is to synthesise an

679

integrated biorefinery with maximum economic potential. As shown in Table C in Supporting

680

Information, the market price for alcohols is higher than the market price of alkanes. Thus,

681

generation of by-products alcohols along with main product Alkane C8 and alkane by-products

682

results in an integrated biorefinery with higher economic potential.

This is because the objective of scenario 1 is to identify the

683 684

In addition, as shown in Table 7, the conversion pathways for the conversion of biomass into the

685

biochemical products identified for both scenarios are feasible in terms of economic potential.

686

Since the optimal biochemical products in terms of target product properties can be produced via

687

feasible optimal conversion pathways in terms of economic potential, an integrated biorefinery

688

can be synthesised and configured based on the identified conversion pathways. Please be noted

689

that different process performances such as product manufacturability can be utilised in assessing


39


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 40 of 46

690

and analysing the generated results. If necessary, the overall design problem has to be repeated

691

from Step 2 of the product design problem, as discussed in Section 3.2.

692 693

5. CONCLUSIONS

694

This paper introduces a novel two-stage optimisation approach to integrate the product with

695

process design aspects in integrated biorefineries to convert biomass feedstock into products

696

which meet customer needs. In the first stage, signature based molecular design techniques have

697

been employed to determine the optimum product in terms of target product properties. By using

698

molecular signature descriptor, different classes of property prediction models can be utilised

699

simultaneously for the design of optimal biochemical products.

700

pathways that convert biomass into the product which meets the required product needs have

701

been determined via a superstructural mathematical optimisation approach in the second stage of

702

the optimisation approach. A case study of design a bio-based fuel with optimised RON from

703

palm-based biomass is solved to illustrate the proposed approach.

The optimum conversion

704 705

As mentioned beforehand, this approach serves as a general representation and idea to integrate

706

the synthesis of integrated biorefinery with product design. Hence, future efforts in this area will

707

include more details in various process/product design steps. In order to reduce the gap between

708

the presented approach and the reality of integrated biorefinery, information such as side

709

reactions, additional reactants required and intermediate products with complex chemical

710

structure involved will be considered and included. Besides, newly developed and enhanced

711

biomass conversion technologies can be included in the superstructure model to determine the


40

Page 41 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


712

optimum pathways. Moreover, in order to enhance the methodology, a detailed business model

713

could be incorporated into the methodology to study the effects of economic performance to the

714

integrated product and process design. This is possible by investigating the impact of market

715

needs, manufacturability of the products as well as the influence of corporate, business and

716

engineering stakeholders to the overall product and process design.

717

simplicity of illustration, the case study presented a bio-based fuel design problem based on the

718

assumption that the targeted bio-based fuel is a single component bio-based fuel. The design

719

problem can be formulated as mixture design problem where information such as properties of

720

different components and mixture stability will be considered and analysed during the design

721

process.

In addition, for the

722 723

AUTHOR INFORMATION

724

Corresponding Author

725

*E-mail: [email protected]

726

ACKNOWLEDGMENT

727

The financial support from the Ministry of Higher Education, Malaysia through the LRGS Grant

728

(Project Code: LRGS/2013/UKM-UNMC/PT/05) is gratefully acknowledged.

729

SUPPORTING INFORMATION AVAILABLE

730

The list of solutions in terms of signatures, list of conversion pathways and yields, list of prices

731

of products and raw material as well as list of capital and operating costs for conversion


41


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 42 of 46

732

pathways are provided in Supporting Information. This information is available free of charge

733

via the Internet at http://pubs.acs.org/.

734

NOMENCLATURE

735

CI, Connectivity index; GC, Group contribution; QSPR/QSAR, Quantitative structure

736

property/activity relationships; TI, Topological index; χ, Connectivity index; G, Molecular sub-

737

graph; h, Height of signature; Ni, Number of occurrence of first order group of type-i; Mj,

738

Number of occurrence of second order group of type-j; Ok, Number of occurrence of third order

739

group of type-k; Ci, Contribution of the first order group of type-i; Dj, Contribution of the second

740

order group of type-j; Ek, Contribution of the third order group of type-k; Ωp, Normalised

741

property operator for target property p;

742

target property p; Vp, Value for target property p;

743

I Fbq , Flow rate of bioresource b to pathway q in t/y; FsqII' , Flow rate of intermediate s to pathway

744

q’ in t/y;

745

to product s’;

746

product s’ in t/y; GPTotal, Total gross profit in U.S $/y;

747

GsProd , Cost of product s’; G bqCap , Capital cost for conversion of bioresource b through pathway q; '

748

Cap G sq '

749

conversion of bioresource b through pathway q;

750

intermediate s through pathway q’; TAC, Total annualised cost; TACC, Total annualised capital

751

cost; TAOC, Total annualised operating cost; CRF, Capital recovery factor;

752

REFERENCES

I R bqs

v Lp

, Lower limit for target property p; B bBio

II R sq 's ' ,

, Total production rate of intermediate s in t/y; G bBio

,

Conversion of intermediate s

TsProd '

, Total production rate of

Cost of biomass feedstock b;

, Capital cost for conversion of intermediate s through pathway q’; Opr G sq '

, Upper limit for

, Flow rate of biomass feedstock b in t/y;

, Conversion of bioresource b to intermediate s; T sInter

v Up

Opr G bq

, Operating cost for

, Operating cost for conversion of


42

Page 43 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


753 754

(1)

Panwar, N. L.; Kaushik, S. C.; Kothari, S. Role of Renewable Energy Sources in Environmental Protection: A Review. Renew. Sustain. Energy Rev. 2011, 15, 1513.

755 756

(2)

Demirbas, M. F. Biorefineries for Biofuel Upgrading: A Critical Review. Appl. Energy 2009, 86, S151.

757 758

(3)

Saxena, R. C.; Adhikari, D. K.; Goyal, H. B. Biomass-Based Energy Fuel through Biochemical Routes: A Review. Renew. Sustain. Energy Rev. 2009, 13, 167.

759 760 761

(4)

Hassan, M. A.; Yee, L.-N.; Yee, P. L.; Ariffin, H.; Raha, A. R.; Shirai, Y.; Sudesh, K. Sustainable Production of Polyhydroxyalkanoates from Renewable Oil-Palm Biomass. Biomass and Bioenergy 2013, 50, 1.

762 763 764

(5)

Kamm, B.; Kamm, M.; Soyez, K. The Green Biorefinery, Concept of Technology. In First International Symposium on Green Biorefinery; Neuruppin Society of Ecological Technology and System Analysis: Berlin, 1998.

765 766

(6)

Huber, G. W.; Corma, A. Synergies between Bio- and Oil Refineries for the Production of Fuels from Biomass. Angew. Chemie Int. Ed. 2007, 46, 7184.

767 768

(7)

Kamm, B.; Kamm, M. Principles of Biorefineries. Appl. Microbiol. Biotechnol. 2004, 64, 137.

769 770

(8)

Martin, M.; Eklund, M. Improving the Environmental Performance of Biofuels with Industrial Symbiosis. Biomass and Bioenergy 2011, 35, 1747.

771 772

(9)

Fernando, S.; Adhikari, S.; Chandrapal, C.; Murali, N. Biorefineries: Current Status, Challenges, and Future Direction. Energy & Fuels 2006, 20, 1727.

773 774 775

(10)

Gravitis, J.; Zandersons, J.; Vedernikov, N.; Kruma, I.; Ozols-Kalnins, V. Clustering of Bio-Products Technologies for Zero Emissions and Eco-Efficiency. Ind. Crops Prod. 2004, 20, 169.

776 777 778 779 780

(11)

Ng, D. K. S.; Pham, V.; El-Halwagi, M. M.; Jiménez-Gutiérrez, A.; Spriggs, H. D. A Hierarchical Approach to the Synthesis and Analysis of Integrated Biorefineries. In Design for Energy and the Environment: Proceedings of Seventh International Conference on Foundations of Computer-Aided Process Design; CRC Press: Boca Raton FL, 2009.

781 782

(12)

Ng, D. K. S. Automated Targeting for the Synthesis of an Integrated Biorefinery. Chem. Eng. J. 2010, 162, 67.

783 784

(13)

Tay, D. H. S.; Ng, D. K. S. Multiple-Cascade Automated Targeting for Synthesis of a Gasification-Based Integrated Biorefinery. J. Clean. Prod. 2012, 34, 38.


43


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 44 of 46

785 786

(14)

Tay, D. H. S.; Kheireddine, H.; Ng, D. K. S.; El-Halwagi, M. M. Synthesis of an Integrated Biorefinery via the C-H-O Ternary Diagram. Chem. Eng. Trans. 2010, 21, 1411.

787 788 789

(15)

Bao, B.; Ng, D. K. S.; El-Halwagi, M. M.; Tay, D. H. S. Synthesis of Technology Pathways for An Integrated Biorefinery. In 2009 AIChE Annual Meeting; Nashville, Tennessee, United States, 2009.

790 791

(16)

Pham, V.; El-halwagi, M. Process Synthesis and Optimization of Biorefinery Configurations. AIChE J. 2012, 58, 1212.

792 793

(17)

Ng, R. T. L.; Ng, D. K. S. Systematic Approach for Synthesis of Integrated Palm Oil Processing Complex. Part 1: Single Owner. Ind. Eng. Chem. Res. 2013, 52, 10206.

794 795

(18)

Sammons, N.; Eden, M.; Yuan, W.; Cullinan, H.; Aksoy, B. A Flexible Framework for Optimal Biorefinery Product Allocation. Environ. Prog. 2007, 26, 349.

796 797

(19)

Zondervan, E.; Nawaz, M.; de Haan, A. B.; Woodley, J. M.; Gani, R. Optimal Design of a Multi-Product Biorefinery System. Comput. Chem. Eng. 2011, 35, 1752.

798 799 800

(20)

Ponce-Ortega, J. M.; Pham, V.; El-Halwagi, M. M.; El-Baz, A. A. A Disjunctive Programming Formulation for the Optimal Design of Biorefinery Configurations. Ind. Eng. Chem. Res. 2012, 51, 3381.

801 802 803

(21)

El-halwagi, A. M.; Rosas, C.; Ponce-Ortega, J. M.; Jiménez-Gutiérrez, A.; Mannan, M. S.; El-halwagi, M. M. Multiobjective Optimization of Biorefineries with Economic and Safety Objectives. AIChE J. 2013, 59, 2427.

804 805

(22)

Amundson, J. S. Modeling of Biorefinery Supply Chain Economic Performance with Discrete Event Simulation, University of Kentucky, 2013.

806 807 808

(23)

Sukumara, S.; Faulkner, W.; Amundson, J.; Badurdeen, F.; Seay, J. A Multidisciplinary Decision Support Tool for Evaluating Multiple Biorefinery Conversion Technologies and Supply Chain Performance. Clean Technol. Environ. Policy 2014, 16, 1027.

809 810 811

(24)

Hostrup, M.; Harper, P. M.; Gani, R. Design of Environmentally Benign Processes: Integration of Solvent Design and Separation Process Synthesis. Comput. Chem. Eng. 1999, 23, 1395.

812 813

(25)

Sinha, M.; Achenie, L. E. K.; Gani, R. Blanket Wash Solvent Blend Design Using Interval Analysis. Ind. Eng. Chem. Res. 2003, 42, 516.

814 815

(26)

Papadopoulos, A. I.; Linke, P. Multiobjective Molecular Design for Integrated ProcessSolvent Systems Synthesis. AIChE J. 2006, 52, 1057.

816 817

(27)

Karunanithi, A. T.; Achenie, L. E. K.; Gani, R. A Computer-Aided Molecular Design Framework for Crystallization Solvent Design. Chem. Eng. Sci. 2006, 61, 1247.


44

Page 45 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


818 819 820

(28)

Conte, E.; Gani, R.; Malik, T. I. The Virtual Product-Process Design Laboratory to Manage the Complexity in the Verification of Formulated Products. Fluid Phase Equilib. 2011, 302, 294.

821 822 823

(29)

Bommareddy, S.; Chemmangattuvalappil, N. G.; Solvason, C. C.; Eden, M. R. Simultaneous Solution of Process and Molecular Design Problems Using an Algebraic Approach. Comput. Chem. Eng. 2010, 34, 1481.

824 825

(30)

Cussler, E. L.; Moggridge, G. D. Chemical Product Design; Cambridge University Press: New York, 2001.

826 827

(31)

Odele, O.; Macchietto, S. Computer Aided Molecular Design: A Novel Method for Optimal Solvent Selection. Fluid Phase Equilib. 1993, 82, 47.

828 829

(32)

Gani, R.; Nielsen, B.; Fredenslund, A. A Group Contribution Approach to ComputerAided Molecular Design. AIChE J. 1991, 37, 1318.

830 831

(33)

Gani, R.; O’Connell, J. P. Properties and CAPE: From Present Uses to Future Challenges. Comput. Chem. Eng. 2001, 25, 3.

832 833

(34)

Achenie, L. E. K.; Gani, R.; Venkatasubramanian, V. Computer Aided Molecular Design: Theory and Practice; Elsevier: Amsterdam, 2003; Vol. 20.

834 835

(35)

Harper, P. M.; Gani, R. A Multi-Step and Multi-Level Approach for Computer Aided Molecular Design. Comput. Chem. Eng. 2000, 24, 677.

836 837

(36)

Gani, R.; Pistikopoulos, E. N. Property Modelling and Simulation for Product and Process Design. Fluid Phase Equilib. 2002, 194-197, 43.

838 839

(37)

Randić, M.; Mihalić, Z.; Nikolić, S.; Trinajstić, N. Graphical Bond Orders: Novel Structural Descriptors. J. Chem. Inf. Comput. Sci. 1994, 34, 403.

840 841 842

(38)

Harper, P. M.; Gani, R.; Kolar, P.; Ishikawa, T. Computer-Aided Molecular Design with Combined Molecular Modeling and Group Contribution. Fluid Phase Equilib. 1999, 158160, 337.

843 844 845

(39)

Ambrose, D. Correlation and Estimation of Vapour-Liquid Critical Properties: I, Critical Temperatures of Organic Compounds, Volume 1; National Physical Laboratory: Teddington, United Kingdom, 1978.

846

(40)

Trinajstić, N. Chemical Graph Theory; CRC Press: Boca Raton, 1992.

847 848

(41)

Camarda, K. V; Maranas, C. D. Optimization in Polymer Design Using Connectivity Indices. Ind. Eng. Chem. Res. 1999, 38, 1884.


45


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 46 of 46

849 850

(42)

Samudra, A.; Sahinidis, N. V. Design of Heat-Transfer Media Components for Retail Food Refrigeration. Ind. Eng. Chem. Res. 2013, 52, 8518.

851 852

(43)

Papadopoulos, A. I.; Stijepovic, M.; Linke, P. On the Systematic Design and Selection of Optimal Working Fluids for Organic Rankine Cycles. Appl. Therm. Eng. 2010, 30, 760.

853 854

(44)

Raman, V. S.; Maranas, C. D. Optimization in Product Design with Properties Correlated with Topological Indices. Comput. Chem. Eng. 1998, 22, 747.

855 856 857

(45)

Visco, D. P.; Pophale, R. S.; Rintoul, M. D.; Faulon, J.-L. Developing a Methodology for an Inverse Quantitative Structure-Activity Relationship Using the Signature Molecular Descriptor. J. Mol. Graph. Model. 2002, 20, 429.

858 859 860

(46)

Chemmangattuvalappil, N. G.; Solvason, C. C.; Bommareddy, S.; Eden, M. R. Reverse Problem Formulation Approach to Molecular Design Using Property Operators Based on Signature Descriptors. Comput. Chem. Eng. 2010, 34, 2062.

861 862 863

(47)

Shelley, M. D.; El-Halwagi, M. M. Component-Less Design of Recovery and Allocation Systems: A Functionality-Based Clustering Approach. Comput. Chem. Eng. 2000, 24, 2081.

864 865

(48)

Wilson, R. J. Introduction to Graph Theory; 4th ed.; Pearson Education Limited: Harlow, 1986.

866 867 868

(49)

Chemmangattuvalappil, N. G.; Eden, M. R. A Novel Methodology for Property-Based Molecular Design Using Multiple Topological Indices. Ind. Eng. Chem. Res. 2013, 52, 7090.

869 870

(50)

Albahri, T. A. Structural Group Contribution Method for Predicting the Octane Number of Pure Hydrocarbon Liquids. Ind. Eng. Chem. Res. 2003, 42, 657.

871 872 873

(51)

Patel, S. J.; Ng, D.; Mannan, M. S. QSPR Flash Point Prediction of Solvents Using Topological Indices for Application in Computer Aided Molecular Design. Ind. Eng. Chem. Res. 2009, 48, 7378.

874 875

(52)

Jurić, A.; Gagro, M.; Nikolić, S.; Trinajstić, N. Molecular Topological Index: An Application in the QSAR Study of Toxicity of Alcohols. J. Math. Chem. 1992, 11, 179.

876 877

(53) Marrero, J.; Gani, R. Group-Contribution Based Estimation of Pure Component Properties. Fluid Phase Equilib. 2001, 183-184, 183.

878 879 880

(54)

Conte, E.; Martinho, A.; Matos, H. A.; Gani, R. Combined Group-Contribution and Atom Connectivity Index-Based Methods for Estimation of Surface Tension and Viscosity. Ind. Eng. Chem. Res. 2008, 47, 7940.

881


46

Novel Methodology for the Synthesis of Optimal ... - ACS Publications

Recommend Documents