optimization of polycondensation process of alkyd resin ...€¦ · alkyd polymers find use in most...
TRANSCRIPT
-
Available online at www.worldscientificnews.com
( Received 27 November 2018; Accepted 12 December 2018; Date of Publication 13 December 2018 )
WSN 116 (2019) 62-90 EISSN 2392-2192
Optimization of polycondensation process of alkyd resin synthesis from modified Picralima nitida seed
oil suitable for surface coating of metal
J. O. Ezeugo
Department of Chemical Engineering, Chukwuemea Odumegwu Ojukwu University, Anambra State, Nigeria
E-mail address: [email protected]
Phone: +2348063873600
ABSTRACT
The potential of synthesizing an air drying alkyd resin through polycondensation of non-drying
Picralima nitida seed oil (PNSO) was investigated. The structural elucidation of the raw PNSO, de-
saturated PNSO and PNSO based alkyd resin were evaluated using FTIR. Design matrix and response
Surface methodology were used to model the process variables. Artificial neural network was further
used to examine the sensitivity analysis of the studied process variables. The results obtained showed
that modified PNSO has high drying rate in the presence of drying agent, exhibit excellent adhesion,
abrasion and chemical properties. Optimum responses of 82% conversion, viscosity of 269 cp and
molecular weight average of 4434 g/mol were predicted at a temperature of 260 ºC, time of 134 mins,
oil ratio of 0.286, catalyst concentration of 0.062 wt and a stirring rate of 595rpm. Correspondent
experimental results using the same optimal conditions gave optimum conversion of 83% of alkyd resin,
viscosity of 271cp and MWA of 4440 g/mol.
Keywords: Picralima nitida seed oil, Response surface methodology, alkyd resin, Artificial Neural
Network
http://www.worldscientificnews.com/mailto:[email protected]
-
World Scientific News 116 (2019) 62-90
-63-
1. INTRODUCTION
One of the emerging major themes in polymer science for the 21st century is the
production of sustainable green polymeric materials and chemicals from renewable resources
[1]. Seeds and fruits of plants are veritable sources of oil for domestic and industrial utility. The
lipid-based raw materials for paints are vegetable oils. Many vegetable oils and some animal
oils are ‘drying’ or ‘semi-drying’ and it is this property that accounts for the suitability of many
oils such as linseed, tung and some fish oils as the base of paints and other coatings. Vegetable
sources occupy an important position in the provision of individual raw materials for paint
production. This is because they are readily renewable resources and contain high levels of
unsaturated fatty acids; a well sought property for oil paint production. They are also
environmental-friendly, less expensive, easy to obtain using conventional extraction techniques
and produced easily in rural areas.
Although vegetable sources of raw materials are readily renewable, the utilization of
wholly inedible and ‘unuseful’ seeds as sources of industrial raw materials will help in
sustaining the high demand for industrial raw materials and reduce the environmental pollutions
usually caused by the indiscriminate dumping of such wastes [2].There are several potentially
useful topical plant materials that have been left unutilized due to inadequate knowledge of
their compositions. The seed of P. nitida epitomized such plant material. In view of the need to
find renewable sources of raw materials of quality for the paint industry, this work is a study of
the seed of Picralima nitida which is known to contain oils and is also wholly inedible [3].
According to [4], all seeds contain oils. With no competing food uses, attention is on P. nitida
seed. Picralima nitida (Akuama) is a well known plant in West Africa. The fruit is inedible.
The bark, leaves, stems and roots of the trees are used as local medicine for the treatment of
diseases [5,6]. Previous studies have shown that the seeds from these fruits contain oil which
have considerable nutritional value. This study is aimed at optimizing P. nitida seed oils for
polycondensation process of alkyd resin synthesis. It intends to maximize percentage yield of
alkyd resin oil based P. nitida via application of optimal independent variables and
subsequently reduce dependence on resin imports for oil paint production.
Firstly, Picralima nitida (P. nitida) is commonly known as Akuamma or pile plant. It
belongs to the tribe of the apocynaceae family. The plant is widely found in forest region of
west-central Africa, from Nigeria to Uganda. It even extends to Cameroon and Congo basin
(Iwu et al, 1993). P. nitida is an under storey tree which reaches up to 4-35 meters in height,
crown dense, trunk 5-60m diameter; cylindrical, the wood is pale-yellow, hard, elastic, fine-
grained and assuming a high polish. It bears white flowers (about 3 cm long) with ovoid fruits
which at maturity are yellowish in color. The leaves are broad (3-10 cm) and oblong (6-20 cm)
long with tough tiny lateral nerves of about 14 to 24 pairs [7]. P. nitida has widely varied
applications in Nigeria and indeed West Africa traditional medicine. The seedis used as
antipyretic, aphrodisiac, for the treatment of malaria [8]. The seed – decoction is given as an
enema while the crushed seed is taken orally for chest problems, pneumonia and for
gastrointestinal disorders [9]. The idea was to make alkyd with a high acid value which would
be neutralized by amines that could be soluble in water [10]. Alkyd and chemically modified
alkyd polymers find use in most types of liquid organic coatings for architectural, air-dry, and
baked industrial and maintenance coatings [11]. Alkyds are a special class of polyesters that
often have vegetable oil or fatty acids co-reacted into the polyester, and these compounds
provide the distinctive air-cure feature of many of these compounds [12].
-
World Scientific News 116 (2019) 62-90
-64-
Figure 1. The general chemical structure of alkyd resin
The coatings made of long oil alkyd resins deliver properties such as ability to be air-
dryable and good interaction with polar substrates such as wood and steel [13]. In addition, the
coatings of conventional alkyd resin with any oil content or acid value are solvent based and
usually diluted in an organic solvent such as toluene, xylene and different oil cuts or a mixture
of these solvents [14].
Alkyd resins were very important paint in the past because of their strong strength, high
film hardness, and gloss retention [16]. They serve as the film-forming agent in some paints
and clear coatings [17]. Therefore alkyd resins are main product of poly-condensation reactions
between polycarboxylicacid and poly alcohol in present fatty acids or vegetable oils [18].This
kind of reaction had obtained by the following formula,
𝑃𝑜𝑙𝑦𝑐𝑎𝑟𝑏𝑜𝑥𝑦𝑙𝑖𝑐 𝑎𝑐𝑖𝑑𝑠 + 𝑃𝑜𝑙𝑦𝑎𝑙𝑐𝑜ℎ𝑜𝑙𝑠 + 𝐹𝑎𝑡𝑡𝑦 𝑎𝑐𝑖𝑑𝑠 𝐴𝑙𝑘𝑦𝑑 𝑟𝑒𝑠𝑖𝑛 + 𝐻2𝑂 … . . (1)
Poly-alcohols which are mainly used for condensational polymerization reactions of
alkyd resins are ethylene glycol, propylene glycol, diethylene glycol and pentaerythritol [19]
2. MATERIALS AND METHODS
2. 1. Sample collection and preparation
P. nitida fruit was plugged from its tree in nearby bush close to Ezeugos compound, Uke
in Idemili North Local Government Area, Nigeria. The pulps of the ripped fruits were consumed
and the seeds were cleaned by washing with distilled water and oven dried to constant weight
in a JP Selecta hot air at 100 ⁰C for 10 hours. The dried seeds were then ground and sieved with
a 450µm in order to increase the surface area prior to solvent extraction.
-
World Scientific News 116 (2019) 62-90
-65-
2. 2. Design of experiments (DOE)
The system prediction through response surface methodology (RSM) study was
implemented on model-based calibration (MBC) v2.0 tool found in MATLAB 7.1 tool box.
The system analysis via MBC tool requires initial coding of the independent variables into the
solution algorithm. This involves using unique notations (or codes) such as; A, B, C…, x1, x2,
x3 ….etc., to represent the variables, and specifying the actual range of values of the variables
using a suitable DOE template. In the current study, five process variables including
temperature, time, oil ratio, catalyst concentration stirring rate coded A, B, C, D and E,
respectively, were considered as major operating factors affecting reaction progress and product
quality in a typical alkyd polycondensation process. A useful DOE template which minimizes
the required number of system iteration for full system assessment is the classical central
composite design (CCD). However, considering the number of independent variables in the
present study, fractional factorial design was implemented at the factorial levels of a standard CCD to further reduce the required number of iterations. This results in a useful experimental
design template called combined array (CA) which is also known to give reliable result with
minimized number of system iteration in studying multi-variate systems [20-23]. The proposed
CA for the current design problem requires performing sixteen distinct system iterations at the
factorial points, ten iterations at the axial points and four iterations at the center points, giving
a total of thirty different iterations of the system. This step was accomplished for the present
study applying the underlisted codes and data ranges:
𝐴: [−𝛼, 𝛼] → Temperature (˚C): [230, 270] 𝐵: [−𝛼, 𝛼] → Time (Mins.): [60, 180] 𝐶: [−𝛼, 𝛼] → Oil ratio: [0.1, 0.5] 𝐷: [−𝛼, 𝛼] → Catalyst Conc.: [0.02, 0.1] 𝐸: [−𝛼, 𝛼] → Stirring rate (rpm): [500, 700]
2. 3. Sensitivity Analysis and System Prediction via Artificial Neural Network
(ANN) Model
Table 1. Design template of alkyd resin from Avocado seed oil (PNSO)
Run Design
space
Independent variables Responses
A B C D E Y1 Y2 Y3
1 Factorial -1 -1 1 1 1 61 216 2438
2 Factorial 1 1 1 1 1 91 228 4890
3 Factorial 1 1 1 -1 -1 81 291 4995
4 Factorial -1 -1 1 -1 -1 38 162 693.9
5 Factorial 1 -1 -1 1 1 77 249 3620
6 Factorial 1 -1 1 -1 1 68 239 3088
-
World Scientific News 116 (2019) 62-90
-66-
7 Factorial 1 -1 1 1 -1 70 273 4436
8 Factorial -1 1 -1 -1 -1 60 215 2533
9 Factorial 1 1 -1 1 -1 78 280 4292
10 Factorial 1 1 -1 -1 1 77 260 3790
11 Factorial -1 1 1 -1 1 76 259 3886
12 Factorial -1 -1 -1 -1 1 30 140 500
13 Factorial -1 -1 -1 1 -1 30 140 500
14 Factorial -1 1 -1 1 1 63 230 2993
15 Factorial 1 -1 -1 -1 -1 42 177 1270
16 Factorial -1 1 1 1 -1 79 275 4640
17 Axial -α 0 0 0 0 51 180 1200
18 Axial Α 0 0 0 0 70 240 3336
19 Axial 0 -α 0 0 0 42 166 839
20 Axial 0 Α 0 0 0 82 273 4445
21 Axial 0 0 -α 0 0 55 199 1947
22 Axial 0 0 α 0 0 65 228 2921
23 Axial 0 0 0 -α 0 60 214 2111
24 Axial 0 0 0 Α 0 81 269 4294
25 Axial 0 0 0 0 -α 75 250 3830
26 Axial 0 0 0 0 Α 78 258 3850
27 Center 0 0 0 0 0 74 255 3980
28 Center 0 0 0 0 0 74 256 3980
29 Center 0 0 0 0 0 75 256 3986
30 Center 0 0 0 0 0 75 255 3988
where: A = Temperature; B = Time; C = Oil Ratio; D = Catalyst Conc.; E = Stirring rate
aliased to highest order interaction effect of the other process variables; Y1 = Conversion;
Y2 = Viscosity; Y3 = Molecular weight average (MWA).
-
World Scientific News 116 (2019) 62-90
-67-
Sensitivity analysis was first conducted to determine the effectiveness of the parameters
by the constructed neural network model. In the analysis, the effects of different interactions of
the variables on the recorded conversion of acid functional group were studied and the results
were used as basis for evaluating the sensitivity of the neural network model. The performances
of the five interaction groups of; (one, two, three, four and five) variables were examined using
the LM (Levenberg-Marquardt) algorithm. Artificial neural network (ANN) model from
Toolbox V7.12 of MATLAB mathematical software is also constructed for prediction of the
system response. The effort is intended to compare the performance of the two popular
functional approximation methods in yielding adequate prediction of the studied alkyd
polycondensation process. The implementation followed the basic procedure for ANN model
(Ghazanfari et al, 2004; Haeri et al, 2000). The performance of ANN and RSM models were
statistically measured and compared by mean relative percent deviation (MRPD)
Picralima nitida seed oil extraction and synthesis of PNSO Oxy-polymerizable alkyd
resin were carried out following the published report of [24-26]
2. 4. Performance Evaluation of Modified Alkyd Resin PNSO
The properties and performance characteristics of the crude PNSO alkyd resin (Alkyd-
A), modified PNSO alkyd resin with PA only (Alkyd-B), modified PNSO alkyd resin with PA
and MA 1:1 (Alkyd-C) and modified PNSO alkyd resin with PA&MA 3:1 (Alkyd-D) of the
same class (50% oil) were to be evaluated in terms of their physico-chemico-mechanical
properties such as acid value, viscosity, colour, refractive index, chemical and abrasion
resistance, drying schedule, adhesion, impact, and scratch hardness test. There is no common
standard to compare alkyds resins as each alkyd resin has its own properties.
Alkyd resin that has acid number of less than 15 is suitable for application of paint,
according to literature (RMRDG, 2012). A higher acid value translates to reduced drying rate,
since acid group usually delay drying rate [27]. In addition, industrial products (such as paints)
formulated with alkyd resin with high acid value usually cause rusting or corrosion of substrate
surfaces.
3. RESULTS AND DISCUSSION
3. 1. Results
A percentage oil yield of 3.63% was obtained after extraction from the P. nitida seed
cake. (Uzoh and Onukwuli, 2018) reported a value of 50% for the avocado seed oil. The
difference in oil content could be attributed to differing climatic conditions, stage of
ripening/development of the fruit at the time of harvest and growth conditions.
3. 2. Statistical screening analysis of alkyd resin produced from the crude and modified
PNSO
To study the effects of the identified system parameters which include; temperature,
reaction time, oil ratio, catalyst concentration and stirring rate; A, B, C, D and E on the
molecular properties (conversion, viscosity and MW of PNSO modified alkyd resin for the
proposed method, a surrogate model of the system was derived from multi-regression analysis.
The global matrix equation (1) was fitted to the data provided by the combined array given in
Table 3 and the resulting models were adjusted in terms of the significant system variables to
-
World Scientific News 116 (2019) 62-90
-68-
obtain the predictive model equation (2-3). The coefficients of determination R2 values of 0.89,
0.97 and 0.95 obtained for Y1, Y2 and Y3 based esterification processes respectively showed
that more than 85% of the overall system variability can be explained by the empirical models
of equations (2-3) which are specific cases of the general predictive equation derived for
individual investigations from the multivariate regression analyses implemented on Design
expert.
𝑅1 = 75.20 + 7.70𝐴 + 10.95𝐵 + 5.32𝐶 + 4.89𝐷 + 3.04𝐸 − 3.12𝐴𝐵 − 2.28𝐴𝐶 + 1.16𝐴 +1.14𝐴𝐸 − 0.87𝐵𝐶 − 2.83𝐵𝐷 − 2.89𝐵𝐸 − 3.41𝐴2 − 2.92𝐵2 − 3.46𝐶2 − 0.92𝐷2 + 0.49𝐸2
(2)
𝑅2 = 253.45 + 19.86𝐴 + 27.24𝐵 + 12.87𝐶 + 10.89𝐷 + 1.07𝐸 − 12.79𝐴𝐵 − 7.58𝐴𝐶 −5.88𝐴𝐸 − 7.26𝐵𝐶 − 10.87𝐵𝐷 − 10.78𝐵𝐸 − 7.87𝐶𝐸 − 6.08𝐷𝐸 − 9.85𝐴2 − 7.46𝐵2 − 9𝐶2
(3)
𝑅3 = 3867.66 + 686.21𝐴 + 945.25𝐵 + 479.87𝐶 + 475.74𝐷 + 78.55𝐸 − 272.94𝐴𝐵− 65.94𝐴𝐸 − 239.44𝐵𝐷 − 227.94𝐵𝐸 − 173.18𝐶𝐸 − 106.19𝐷𝐸− 341.99𝐴2 − 248.43𝐵2 − 300.49𝐶2 − 108.28𝐷2 + 51.01𝐸2
(4)
where: Y1, Y2 and Y3 are the predicted values of the dependent variables investigated. The
coefficients A, B, C, D and E are main linear effects of the independent process variables. AB,
AC, AE, BC, BD, BE, CE and DE represent the linear interaction effects between the various
independent variables. A2, B2 and C2 are the quadratic effects of the respective process
variables.
The estimated coefficient terms revealed that quadratic interactions of system variables
showed negative effects while the main linear and first order interactive effects are positive.
Higher order interaction terms are not significant in the model. The “Predicted R-Squared” of
0.87, 0.97 and 0.95 are in reasonable agreement with the respective “Adjusted R-Squared” of
0.89, 0.94 and 0.96; and the Model F-values of 16.53, 28.71 and 27.65 further indicated that
the models are significant.
There is only a 0.01% probability that the “Model F- values’’ this large could occur due
to noise. P-values of less than 0.05 indicated that the model terms are significant. “Adequacy
Precision” measures the signal to noise ratio (SN). A ratio greater than 4 is desirable. SN values
13.88, 17.46 and 17.272 indicated adequate signal to noise ratios. This model can be used to
navigate the design space.
The ANOVA results derived from the predictive models showed that the main linear
effects due to individual control factors coded A, B, C and D respectively are significant
variables indicated with the observed P-values < 0.05 in the numerical analysis.
This is equally true with the linear interaction effects between temperature and time
(AB),time and catalyst concentration (BD), time and stirring rate (BE) and (for Y2 analysis)
AB, AC, AE, BC, BD, BE, CE and DE are valid while AB, BD, BE and CE is valid for Y3. The
quadratic effects of temperature, time and molar ratio denoted by A2, B2 and C2 respectively
are significant for the molecular properties (Y1, Y2 and Y3) shown on Table 2.
-
World Scientific News 116 (2019) 62-90
-69-
Table 2. ANOVA response surface reduced order quadratic model (RSROQM) in terms
of only the significant process parameters.
SOURCE
F-Value P-Value
Y1 Y2 Y3 Y1 Y2 Y3
Model 19.51 28.87 21.58
-
World Scientific News 116 (2019) 62-90
-70-
Conversion
(Y1)
Std. Dev. 6.32 R-Squared 0.8969
Mean 66.06 Adj R-Squared 0.8426
C.V.% 9.57 Pred R-Squared 0.7424
Press 1897.26 Adeq R-Squared 13.880
Viscosity (Y2)
Std. Dev. 10.39 R-Squared 0.9725
Mean 231.51 Adj R-Squared 0.9386
C.V.% 4.49 Pred R-Squared 0.7601
Press 12243.47 Adeq R-Squared 17.460
Molecular
weight average
(Y3)
Std. Dev. 399.70 R-Squared 0.9513
Mean 3109.10 Adj R-Squared 0.9169
C.V.% 12.86 Pred R-Squared 0.8187
Press 1.01x107 Adeq R-Squared 17.272
DF: Model = 9, Residual = 20, Total = 29.
3. 3. Interaction effects of process variables
The combined effects of adjusting the process variables within the design space were
monitored using 3-D surface plots. Every significant interaction effects on the system response
between two independent variables were analyzed in phases. The results are presented in Figs.
(2- 4).
The overall system behavior is characterized by various degrees of curvature which
reflect the levels of uncertainties associated with every interaction of the process variables. The
observed trends suggested that by proper adjustment of the system variables within the sampled
space, a valid optimal was attained. The results conformed largely to what is already known for
alkyd polycondensation processes [28-32]
Nevertheless, one noticeable unusual result may be the apparent linear (one-directional)
response observed in the system response with respect to D and E axes. This does not reflect a
typical behavior of batch alkyd polycondensation processes. Particle congestion which occurs
at high catalyst concentration.
The consistent linear behavior implies that no definite optimal solution could be obtained
for (D and E).
-
World Scientific News 116 (2019) 62-90
-71-
(a)
(b)
Design-Expert® SoftwareFactor Coding: ActualR1
Design points above predicted valueDesign points below predicted value91.1
31.1
X1 = A: AX2 = C: C
Actual FactorsB: B = 120D: D = 0.06E: E = 600
0.2
0.25
0.3
0.35
0.4240
245
250
255
260
30
40
50
60
70
80
90
100
R1
A: AC: C
Design-Expert® SoftwareFactor Coding: ActualR1
Design points above predicted valueDesign points below predicted value91.1
31.1
X1 = A: AX2 = B: B
Actual FactorsC: C = 0.3D: D = 0.06E: E = 600
90 100
110 120
130 140
150240
245
250
255
260
30
40
50
60
70
80
90
100
R1
A: AB: B
-
World Scientific News 116 (2019) 62-90
-72-
Design-Expert® SoftwareFactor Coding: ActualR1
Design points above predicted valueDesign points below predicted value91.1
31.1
X1 = A: AX2 = E: E
Actual FactorsB: B = 120C: C = 0.3D: D = 0.06
550
570
590
610
630
650240
245
250
255
260
30
40
50
60
70
80
90
100
R1
A: AE: E
(c)
(d)
Design-Expert® SoftwareFactor Coding: ActualR1
Design points above predicted valueDesign points below predicted value91.1
31.1
X1 = A: AX2 = D: D
Actual FactorsB: B = 120C: C = 0.3E: E = 600
0.04
0.05
0.06
0.07
0.08240
245
250
255
260
30
40
50
60
70
80
90
100
R1
A: AD: D
-
World Scientific News 116 (2019) 62-90
-73-
(e)
(f)
Figure 2. Interaction effects between: (a), A and B, (b), A and C, (c), A and D, (d), A and E,
(e), B and E, (f), B and D on conversion of acid functional group
Design-Expert® SoftwareFactor Coding: ActualR1
Design points above predicted valueDesign points below predicted value91.1
31.1
X1 = B: BX2 = E: E
Actual FactorsA: A = 250C: C = 0.3D: D = 0.06
550
570
590
610
630
65090
100
110
120
130
140
150
30
40
50
60
70
80
90
100
R1
B: BE: E
Design-Expert® SoftwareFactor Coding: ActualR1
Design points above predicted valueDesign points below predicted value91.1
31.1
X1 = B: BX2 = D: D
Actual FactorsA: A = 250C: C = 0.3E: E = 600
0.04
0.05
0.06
0.07
0.0890
100
110
120
130
140
150
30
40
50
60
70
80
90
100
R1
B: BD: D
-
World Scientific News 116 (2019) 62-90
-74-
(a)
(b)
Design-Expert® SoftwareFactor Coding: ActualR2
Design points below predicted value291
141
X1 = A: AX2 = B: B
Actual FactorsC: C = 0.2D: D = 0.04E: E = 650
90
100
110
120
130
140
150
240
245
250
255
260
100
150
200
250
300
R2
A: AB: B
261.746
Design-Expert® SoftwareFactor Coding: ActualR2
Design points above predicted valueDesign points below predicted value291
141
X1 = A: AX2 = C: C
Actual FactorsB: B = 150D: D = 0.04E: E = 650
0.2
0.25
0.3
0.35
0.4
240
245
250
255
260
100
150
200
250
300
R2
A: AC: C
261.746
-
World Scientific News 116 (2019) 62-90
-75-
(c)
(d)
Design-Expert® SoftwareFactor Coding: ActualR2
Design points below predicted value291
141
X1 = A: AX2 = E: E
Actual FactorsB: B = 150C: C = 0.2D: D = 0.04
550
570
590
610
630
650
240
245
250
255
260
100
150
200
250
300
R2
A: AE: E
261.746
Design-Expert® SoftwareFactor Coding: ActualR2
Design points above predicted valueDesign points below predicted value291
141
X1 = B: BX2 = C: C
Actual FactorsA: A = 260D: D = 0.04E: E = 650
0.2
0.25
0.3
0.35
0.4
90
100
110
120
130
140
150
100
150
200
250
300
R2
B: BC: C
261.746
-
World Scientific News 116 (2019) 62-90
-76-
Design-Expert® SoftwareFactor Coding: ActualR2
Design points above predicted valueDesign points below predicted value291
141
X1 = B: BX2 = D: D
Actual FactorsA: A = 260C: C = 0.2E: E = 650
0.04
0.05
0.06
0.07
0.08
90
100
110
120
130
140
150
100
150
200
250
300
R2
B: BD: D
261.746
(e)
(f)
Design-Expert® SoftwareFactor Coding: ActualR2
Design points below predicted value291
141
X1 = B: BX2 = E: E
Actual FactorsA: A = 260C: C = 0.2D: D = 0.04
550
570
590
610
630
650
90
100
110
120
130
140
150
100
150
200
250
300
R2
B: BE: E
261.746
-
World Scientific News 116 (2019) 62-90
-77-
(g)
(h)
0.020.04
0.060.08
0.1
0.1
0.2
0.3
0.4
0.5150
200
250
300
D [Catalyst Conc. (w /w )]C [Oil ratio]
Y2 [V
iscosity (
cP
)]
Design-Expert® SoftwareFactor Coding: ActualR2
Design points above predicted valueDesign points below predicted value291
141
X1 = C: CX2 = E: E
Actual FactorsA: A = 260B: B = 150D: D = 0.04
550
570
590
610
630
650
0.2
0.25
0.3
0.35
0.4
100
150
200
250
300
R2
C: CE: E
261.746
-
World Scientific News 116 (2019) 62-90
-78-
(i)
Figure 3. Interaction effects between: (a) A and B, (b) A and C, (c) A and E, (d) B and C,
(e) B and D, (f) B and E, (g) C and D, (h) C and E, (i) D and E on viscosity.
Fig. 2. (a) and (b) reveal some basic routes to optimum progress of reaction for the studied
alkyd polymerization process which requires tuning the key control variables including reaction
time, temperature or oil ratio to some values well above their center points while catalyst
concentration and stirring rate are fixed at their mean values. This framework stands as a viable
solution especially where oil as a major raw material is relatively available. Alternatively,
optimum conversion of acid functional group could be achieved by setting reaction time and
temperature at their mean values while either catalyst concentration, Fig. 2. (c) or steering rate,
Fig. 2.(d) is reviewed upwards.
By and large, the latter solution may be less economically viable since it possibly leads
to increased material/operational cost [33]. The most economically viable solution in terms of
achieving optimal progress of reaction may be traced to the possibility of reaching optimum
conversion of acid functional group either with low catalyst concentration demonstrated in
Figure 2 (e) or with low stirring rate presented in Figure 2 (f) by mere allowing a relatively high
reaction period. Hence reaction time seems to be a very important process variable whose
overall effects on the system responses must be investigated in details to enhance selection of
the most desirable optimal solution.
In a typical alkyd monitoring scheme, reactor performance is usually measured in terms
of the commercial values of the resulting molecular properties of the product including cold-
viscosity and/or molecular weight average. Thus these molecular properties are usually
monitored (preferably on-line) such that the final product lies substantially within specifications
Design-Expert® SoftwareFactor Coding: ActualR2
Design points above predicted valueDesign points below predicted value291
141
X1 = D: DX2 = E: E
Actual FactorsA: A = 260B: B = 150C: C = 0.2
550
570
590
610
630
650
0.04
0.05
0.06
0.07
0.08
100
150
200
250
300
R2
D: DE: E
261.746
-
World Scientific News 116 (2019) 62-90
-79-
34-36]. As part of the present modeling and optimization scheme, data obtained for cold-
viscosity and MWA at various conditions were analyzed and the effects of every significant
interaction of the variables on system response were evaluated systematically. The results of
this study are presented in Figure 2 for cold-viscosity and Figure 3 for MWA. In coating
industry alkyd resin whose cold-viscosity and MWA values lie in the range (243-600 CP) and
(4000-10000 g/mol.) satisfy most commercial needs (Fujita and Kishimoto, 1951). The results
presented in Fig. 2 highlight various ways to achieve this target.
Figures 2 (a-g) represent a typical behavior in which increasing the value of the system
variables initially leads to a corresponding increase in the value of the observed response before
an optimum value is attained, after which the trend is reversed. In Fig. 2 (c, f, h, and.i) high
viscosity response is recorded at low stirring speed by reviewing upwards the temperature, the
reaction time, the oil ratio and the catalyst concentration, respectively. For some economic
reasons, it is usually more elegant to pursue a good result with moderate oil ratio and reduced
catalyst concentration.
Thus, an important observation made in this regard is the possibility of obtaining high
viscous resin with low catalyst concentration and mean oil ratio presented in Fig. 2.(c) and (e)
by mere allowing relatively increased reaction time. This point again highlights the reaction
time as an important factor that must be monitored closely to control both reaction progress and
product quality without unnecessarily increasing the cost of operating the batch. The results
obtained for MWA properties of the system at various operation conditions are presented in
Fig. 3. However, Fig. 3(a) highlights the futility of operating the reactor for short reaction time
using low reactor temperature.
(a)
Design-Expert® SoftwareFactor Coding: ActualR3
Design points above predicted value4995
500
X1 = A: AX2 = B: B
Actual FactorsC: C = 0.2D: D = 0.04E: E = 650
90
100
110
120
130
140
150
240
245
250
255
260
0
1000
2000
3000
4000
5000
R3
A: AB: B
3625.86
-
World Scientific News 116 (2019) 62-90
-80-
Design-Expert® SoftwareFactor Coding: ActualR3
Design points above predicted valueDesign points below predicted value4995
500
X1 = A: AX2 = E: E
Actual FactorsB: B = 150C: C = 0.2D: D = 0.04
550
570
590
610
630
650
240
245
250
255
260
0
1000
2000
3000
4000
5000
R3
A: AE: E
3625.86
+
(b)
(c)
Design-Expert® SoftwareFactor Coding: ActualR3
Design points above predicted value4995
500
X1 = B: BX2 = D: D
Actual FactorsA: A = 260C: C = 0.2E: E = 650
0.04
0.05
0.06
0.07
0.08
90
100
110
120
130
140
150
0
1000
2000
3000
4000
5000
R3
B: BD: D
3625.86
-
World Scientific News 116 (2019) 62-90
-81-
Design-Expert® SoftwareFactor Coding: ActualR3
Design points above predicted value4995
500
X1 = B: BX2 = E: E
Actual FactorsA: A = 260C: C = 0.2D: D = 0.04
550
570
590
610
630
650
90
100
110
120
130
140
150
0
1000
2000
3000
4000
5000
R3
B: BE: E
3625.86
(d)
(e)
Figure 3. Interaction between: (a) A and B, (b) A and D, (c) B and D, (d) B and E,
(e) C and E, on molecular weight average
Design-Expert® SoftwareFactor Coding: ActualR3
Design points above predicted value4995
500
X1 = C: CX2 = E: E
Actual FactorsA: A = 260B: B = 150D: D = 0.04
550
570
590
610
630
650
0.2
0.25
0.3
0.35
0.4
0
1000
2000
3000
4000
5000
R3
C: CE: E
3625.86
-
World Scientific News 116 (2019) 62-90
-82-
With mean reaction time and moderate temperature a good result in MWA properties
could be achieved; either with increased catalyst concentration as described in Fig. 3(b) or with
high oil ratio as shown in Fig. 3 (e). However, both options may lead to increased material cost
making them less economically viable. Some good alternative routes to desirable response are
demonstrated in Fig. 3(a) and Fig. 3(d) where good MWA properties are rather obtained with
low catalyst concentration and low stirring rate, respectively, by mere increasing the reaction
time.
3. 4. Sensitivity Analysis and Reaction Prediction via ANN
The performances of the five interaction groups of; (one, two, three, four and five)
variables were examined using the LM (Levenberg-Marquardt) algorithm. The results are
presented in Table 3. The results show that while B (reaction time) is the most effective
parameter in the group of one variable, the interaction of A and B which shows relatively low
MSE value (0.299) has the greatest effect in the group of two variables. Similarly, the
interaction of five variables produced the most significant effect on the studied response with
the overall minimum MSE of 0.079. This suggests that the most accurate prediction of the
system via ANN model could be achieved by using the five independent variables in
formulation of the model.
3. 5. Reaction prediction using artificial neural network
The performance of the constructed artificial neural network model in yielding the
conversion of acid functional group (𝑌1), viscosity (𝑌2) and molecular weight average (𝑌3)were evaluated considering the highest order interaction of the system variables (ABCD E)
respectively.
The results are presented in Fig. (4-6) for 𝑌1, 𝑌2 and 𝑌3 respectively. The ANN results and that of equivalent RSM are compared with experimental data in Table 3 and 4.From the
comparative study, looking at the performances of the two functional approximation models it
seems that greater overall prediction accuracy is guaranteed by the RSM model. Further
analysis of the predictive efficiency of models using MRPD as shown in Table 5 clearly showed
that RSM is a better model for alkyd resin poly-condensation process. This may be because of
the limited number of experimental data used in the analysis. ANN generally performs better
when very large number of data points is used for training the network (Katarina et al, 2013).
Attempts to compare this observation overtly with earlier work from literature did not provide
much desire result. This is because there is scanty or no study on alkyd resin poly-condensation
process modeling using ANN and RSM.
However, the result contained in this essay are apparently different from literature;
methanol sis of sunflower oil using ANN and RSM in which there are 162 data points (Katarina
et al, 2013); RSM and ANN modeling of electro coagulation of copper from simulated
wastewater (Manpreet et al, 2011); optimization of recombinant of Oryza sativa non-symbiotic
hemoglobin using ANN and RSM. Optimization of biosorption process using ANN and RSM,
alkaline palm oil transesterification by RSM and ANN. From the foregoing, the prediction
accuracy was more prominent for ANN possibly because, it was built up from the experimental
results of RSM.
-
World Scientific News 116 (2019) 62-90
-83-
Table 3. Performance evaluation of the interaction of the process variables for the LM
algorithm with 5 neurons in the hidden layer for sensitivity analysis
No Interaction MSE 𝑅2 Gradient BLF
Group of one Variables
1 A 0.736 0.24 1.38 × 10−09 𝑌1 = 0.217𝑇 + 51.9 2 B 0.546 0.42 4.30 × 10−08 𝑌1 = 0.413𝑇 + 38.6 3 C 0.766 0.14 7.80 × 1004 𝑌1 = 0.15𝑇 + 56.9 4 D 0.836 0.09 3.40 × 1004 𝑌1 = 0.092𝑇 + 60.3 5 E 0.857 0.07 1.56 × 10−2 𝑌1 = 0.069𝑇 + 61.8
Group of two variables
6 A B 0.299 0.67 2.77 × 1000 𝑌1 = 0.682𝑇 + 20.5 7 A C 0.436 0.33 7.47 × 10−09 𝑌1 = 0.377𝑇 + 39.9 8 A D 0.758 0.28 1.30 × 10−10 𝑌1 = 0.179𝑇 + 55.6 9 A E 0.625 0.27 8.60 × 10−08 𝑌1 = 0.211𝑇 + 52.9 10 B C 0.386 0.54 1.59 × 10−09 𝑌1 = 0.475𝑇 + 37 11 B D 0.444 0.51 1.84 × 10−03 𝑌1 = 0.51𝑇 + 32.9 12 B E 0.439 0.48 9.23 × 10−12 𝑌1 = 0.597𝑇 + 26.9 13 C D 0.581 0.16 4.98 × 10−09 𝑌1 = 0.18𝑇 + 54.5 14 C E 0.629 0.13 2.35 × 10−11 𝑌1 = 0.167𝑇 + 58.2 15 D E 0.790 0.11 3.15 × 10−11 𝑌1 = 0.075𝑇 + 62.2
Group of three variables
16 ABC 0.165 0.82 5.91 × 10−10 𝑌1 = 0.81𝑇 + 11.7 17 ABD 0.217 0.76 8.86 × 10−09 𝑌1 = 0.746𝑇 + 17 18 ABE 0.236 0.73 1.35 × 10−10 𝑌1 = 0.632𝑇 + 24 19 ACD 0.410 0.46 1.05 × 1001 𝑌1 = 0.49𝑇 + 32.9 20 ACE 0.584 0.37 5.42 × 10−12 𝑌1 = 0.336𝑇 + 42.9 21 ADE 0.613 0.30 2. 11 × 10−13 𝑌1 = 0.311𝑇 + 46.3 22 BCD 0.330 0.64 8. 28 × 10−10 𝑌1 = 0.601𝑇 + 26.4 23 BCE 0.385 0.58 9. 01 × 10−09 𝑌1 = 0.586𝑇 + 26.6 24 BDE 0.419 0.57 1. 06 × 10−11 𝑌1 = 0.541𝑇 + 30.4 25 CDE 0.642 0.20 1.42 × 10−12 𝑌1 = 0.204𝑇 + 52
Group of four variables
26 ABCD 0.095 0.86 1. 35 × 10−11 𝑌1 = 0.823𝑇 + 11.7 27 ABCE 0.186 0.76 3.23 × 10−12 𝑌1 = 0.663𝑇 + 23.1 28 ABDE 0.152 0.77 8. 42 × 10−11 𝑌1 = 0.752𝑇 + 15.4 29 ACDE 0.475 0.29 1. 69 × 10−10 𝑌1 = 0.173𝑇 + 63.1 30 BCDE 0.166 0.69 1.52 × 10−12 𝑌1 = 0.717𝑇 + 19.1
Group of five variables
31 ABCDE 0.079 0.88 1. 16 × 10−11 𝑌1 = 0.853𝑇 + 10.1
-
World Scientific News 116 (2019) 62-90
-84-
0 5 10 15 200
1
2
3
4
Epoch
MS
E
TrainingValidationTest
20 40 60 80 10030
40
50
60
70
80
90
100
Experimentally measured conversion (%)
AN
N p
redic
ted c
onvers
ion (
%)
Data PointsBest Linear FitA = T
(a)
(b)
Figure 4. Predicted versus actual values of the conversion (b) performance of
the constructed ANN model in predicting the actual value of the conversion
-
World Scientific News 116 (2019) 62-90
-85-
Figure 5. Predicted (A) versus actual (T) values of (a) the viscosity
Figure 6. Predicted (A) versus actual (T) values of (b) the molecular weight average.
100 150 200 250 300100
150
200
250
300
T
A
Best Linear Fit: A = (0.911) T + (25.2)
R = 0.962
Data Points
Best Linear Fit
A = T
0 2000 4000 60000
1000
2000
3000
4000
5000
T
A
Best Linear Fit: A = (0.938) T + (188)
R = 0.96
Data Points
Best Linear Fit
A = T
-
World Scientific News 116 (2019) 62-90
-86-
Table 4. RSM and ANN results compared with the experimental data.
S/N
Independent variables
A B C D E
𝑌1(%) Actual RSM
ANN
𝑌2(𝑐𝑃) Actual RSM
ANN
𝑌3(𝑔/𝑚𝑜𝑙) Actual RSM
ANN
1a
2c
3a
4b
5a
6a
7a
8c
9a
10b
11a
12a
13a
14c
15a
16b
17a
18a
19a
20c
21a
22b
23a
24a
25a
26c
27a
28b
29a
30
240 90 0.4 0.08 650
260 150 0.4 0.08 650
260 150 0.4 0.04 550
240 90 0.4 0.04 550
260 90 0.2 0.08 650
260 90 0.4 0.04 650
260 90 0.4 0.08 550
240 150 0.2 0.04 550
260 150 0.2 0.08 550
260 150 0.2 0.04 650
240 150 0.4 0.04 650
240 90 0.2 0.04 650
240 90 0.2 0.08 550
240 150 0.2 0.08 650
260 90 0.2 0.04 550
240 150 0.4 0.08 550
270 120 0.3 0.06 600
230 120 0.3 0.06 600
250 180 0.3 0.06 600
250 60 0.3 0.06 600
250 120 0.5 0.06 600
250 120 0.1 0.06 600
250 120 0.3 0.1 600
250 120 0.3 0.02 600
250 120 0.3 0.06 700
250 120 0.3 0.06 500
250 120 0.3 0.06 600
250 120 0.3 0.06 600
250 120 0.3 0.06 600
250 120 0.3 0.06 600
62
91
81
38
77
68
70
60
78
77
76
30
30
63
42
79
71
51
83
43
66
55
81
61
74
75
74
75
75
75
64
86
80
40
77
66
69
63
79
75
77
34
37
64
46
80
76
45
85
41
71
50
84
64
81
68
74
74
74
74
61
79
78
37
72
71
75
65
78
78
76
34
40
77
48
78
78
48
79
36
77
63
77
59
76
69
73
73
73
73
227
229
291
163
250
240
274
216
280
261
260
141
141
231
178
276
241
181
274
167
229
210
271
215
259
251
256
256
256
256
216
225
280
161
253
231
270
219
281
257
258
145
145
236
174
274
253
174
278
169
243
191
275
231
255
251
253
253
253
253
217
231
267
176
255
242
267
244
279
270
257
145
156
267
185
276
267
186
278
165
232
205
268
214
248
242
264
264
264
264
2440
4890
4990
693
3621
3089
4436
2533
4292
3790
3886
501
505
2993
1270
4640
3336
1210
4445
839
2921
1947
4294
2111
3850
3830
3980
3980
3986
3988
2328
4759
4825
735
3746
3016
4089
2815
4111
3564
3463
392
746
3216
1229
4467
3803
1058
4695
840
2956
1637
4681
2777
4140
3825
3729
3729
3729
3729
2712
4146
4913
0795
3571
3062
3507
3041
4199
4167
3818
0340
0715
4066
1441
4281
3864
1302
4552
313
3117
1717
4191
2112
3818
4311
3869
3869
3869
3869
a -Training data set; b -Validating data set; c -Testing data set of the ANN mode
Table 5. Mean Relative Percent Deviation (MRPD) for Artificial Neural Net Work (ANN)
MRPD (%)
RSM ANN
Conversion % 3.13 ±4.18
Viscosity CP 2.91 ±3.11
Molecular weight average g/mol 7.01 ±9.24
-
World Scientific News 116 (2019) 62-90
-87-
3. 6. Optimization process
Significant economic benefit may be derived by optimizing the molecular properties
which relates to the end-use properties of PNSO modified alkyd resin during synthesis. These
molecular properties are required to lie within some desired optimal in the parameter design
space. As anticipation of a typical manufacturing process, the characterization-control-
optimization algorithm base on a 25-1 FFA adequately guaranteed the details process analysis
and optimization of the PNSO modified alkyd resin molecular properties. Of particular interest
in the end-use properties is the drying time of the alkyd which has been highlighted as a critical
issue associated with drying oil. Other advantages from the chosen optimization process are
through reduced reaction time and temperature which in effect minimized the overall cost of
the synthesis simultaneously.
The processing of high quality resin from PNSO, with short drying time reduced material
cost, improved color (appearance), viscosity, molecular weight and high commercial
importance necessarily require optimization of the poly-condensation process with particular in
focus on the molecular properties as the output responses. This was achieved in the present
study through formulation of a global optimization criteria based on RSM upon which the
necessary trade-off of system variables were implemented. Such systematic compromise was
particularly important in the process described above since the system responses show peak
value at non-unique locations within the variable design space. The entire exercise aided by
numerical optimization tool function of the design expert statistical software (trial version 9)
used for the experimental analysis. Equations (1-3) were solved for the best solution(s) such
that the response Y, X and Z were maximized. No unique solution was attainable.
The various solutions obtained were assessed based on their contribution to maximum
responses and other necessary economic consideration. From 20 best optimal solution obtained
as shown in Table 6 (a) credible optimum solution of 82.184% fractional conversion, viscosity
of 269.444 CP and MW (av) of 4434.213 predicted at temperature of 260 ºC, time of 134.043
min, oil ratio of 0.286, catalyst ratio of 0.062 and stirring rate of 595.385rpm at 1.00 desirability
were selected based on economic consideration and necessary trade-off. A repeated
correspondence investigation performed following the predicted optimal conditions recorded
83% fractional conversion, viscosity of 271CP and MW (av) of 4440 g/mol for the studied PNSO
modified alkyd resin. This figure represents 0.65% average maximum prediction error.
Table 6. Optimum values of process parameters for maximum responses
Process Parameters Optimum Values
Temperature (ºC) A 260.000
Time (mins) B 134.043
Oil Ratio C 0.286
Catalyst Conc. (wt.)D 0.062
Stirring Rate (rpm) 595.385
Conversion (%)Y1 82.184
Viscosity (CP) Y2 269.444
Molecular Weight Average (g/mol) Y3
4434.213
-
World Scientific News 116 (2019) 62-90
-88-
4. CONCLUSION
An auto-oxidative alkyd resin was synthesized from the chemically modified Picralima
nitida seed oil. The optimum yield of the P. nitida seed oil was determined through the
application of central composite design matrix Via response surface methodology. The optimal
conditions of the process parameters which gave the optimum responses were established
Acknowledgements
The authors sincerely acknowledged the staff and management of National Center for Energy Research and
Development, University of Nigeria Nsukka (UNN). Isa Yakubu of National Research Center for Chemical
Technology, Zaria, Nigeria. Our gratitude also goes to the staff and managements of Kappas Biotechnology,
Ibadan, Nigeria. However, we wish to state that no fund was received in the course of executing this research
work.
References
[1] Aigbodion A.I., Okieimen, F.E. 1996. Kinetics of the preparation of rubber seed oil alkyds, European Polymer Journal 32, 1105-1108, https://doi.org/10.1016/0014-
3057(96)00053-5
[2] Alvarez. J, Lopez. T, Robust dynamic state estimation of nonlinear plants, AIChE Journal 45 (1) (1999)
[3] Ansa-Asamoah, R. Kapadia, G.J., Lloyd, H.A., E.A. Sokoloski, E.A. 1990. Picratidine,
new indole alkaloid from Picralima nitida seeds, J. Nat. Prod. 153, 4, 975-977.
[4] Ezeugo J N O, Onukwuli O D, Omotioma M. 2018. Inhibition of Aluminium corrosion in 1.0 M HCl using Picralima nitida leaf extract, Der Pharma Chemica, 10 SI 7-13
[5] Etukudo, I. 2003. Conventional and traditional us e s of plants, Ethnobotany, Uyo The
Verdict Press, 1, 191.
[6] Ezeugo J N O, Onukwuli O D, Omotioma M. (2017). Optimization of corrosion inhibition of picralima nitida leaves extract as green corrosion inhibitor for zinc in 1.0
HCl. World News of Natural Sciences 15, 139-161
[7] Ghazanfari, Mehdi, Arkat, Jamal, 2004. Neural Networks (Principles and Applications) science and Industry University of Iran.
[8] Gogte, B.B, Dabhade, S.B, 1981. Alkyd Based on Non Edible Oils Karanja Oil (Pongamia glabra), Paint India, pp. 3–5.
[9] Golan, A.; Sandovski, Y.; Kahn, V., 2013. Evaluation of Browning Potential in Avocado Mesocarp, Journal of Food Science 42: 853-855.
[10] Gomez, R. F. and Bates, R.P. 2013. Storage Deterioration of Freeze- Dried Avocado Puree and Guacamole. Journal of Food Science 35: 472-475.
[11] Henerndez-Escoto. H, Lopez. T, Alvarez. J, 2010. Estimation of alkyd reactors with discrete-delayed measurements, Chemical Engineering Journal 160, 698-707
https://doi.org/10.1016/0014-3057(96)00053-5https://doi.org/10.1016/0014-3057(96)00053-5
-
World Scientific News 116 (2019) 62-90
-89-
[12] Hernández. H, Alvarez. J, 2003. Robust estimation of continuous nonlinear plants with discrete measurements, Journal of Process Control 13, 69-89
[13] Haeri, Seyed Mohsen, Sadati, Naser, Mahin Rousta, Reza, 2000. Using Neural Network for Anticipating Tension Behavior, Strain of Clay, Set of Articles at 5th International
Civil.
[14] Hester, O. C, and T. S. Stephens, 1997.Development and Preliminary Test of a Frozen Avocado Salad Base. Journal of Rio Grande Valley Horticultural Science 24: 176-180.
[15] Hlaing, N.N, Mya, O, 2008. Manufacture of alkyd resin from castor oil, World Academy of Science, Engineering and Technology 24, 115–161.
[16] Igwe, I.O, Ogbobe, O, 2000. Studies on the properties of polyester and polyester blends of selected vegetable oil, Journal of Applied Polymer Science 75, 1441–1446.
[17] Ikhuoria, E. U. and Maliki, M., 2007. Characterisation of Avocado Pear (Persea Americana) and African Pear (Dacryodes edulis) Extracts. African Journal of
Biotechnology 6(7): 950-952.
[18] Iwu, M., 1993. Hand book of African medicinal plants. U.S.A, CRC Press Inc., 219-221.
[19] Jirovetz, L.; Buchbawer, G.; Geissler, M.; Ngassoum, M. B.; Parmentler, M., 2003. Pulp Aroma Compounds of Untreated, Boiled and Roasted African Pear (Dacryodes
edulis (Don. G) Lam. H. J) Fruits from Cameroon by HS-SPME Analysis Complied
with GC/FID and GC/MS. European Food Resources Technology, 218: 40-43
[20] Kahn, V., 1997.Polyphenol Oxidase Activity and Browning of Three Avocado Varieties. Journal of Science Food Agriculture, 26: 1319-1324.
[21] Kapadia, G.J., Angerhofer, C.K., R. Ansa-Asamoah, R. 1993. Akuammine: an antimalarial indolemonoterpene alkaloid of Picralima nitida seeds. Planta Medica, 59,
6, 565-566
[22] Katarina, M.R., Jelena, M.A., Petar, S.M., Olivera, S.S., 2013. Optimization of ultrasound-assisted base-catalysed methanolysis of sunflower oil using response surface
and artificial neural network methodology. Chemical Engineering Journal, 2015-2016
pp.82-89..
[23] Manpreet, S.B., Dhriti K., Rajeev, K.K., Akepati, S.R., Ashwani, K.T., 2011. RSM and ANN modeling for electrocoagulation of copper from simulated wastewater: Multi
objective optimization using genetic algorithm approach. Desalination, 274 pp.74-80.
[24] Menkiti, M.C, Onukwuli, O.D, 2011. Utilization potentials of rubber seed oil for the production of alkyd resin using variable base oil lengths, New York Science Journal 4
(2) pp.51–58.
[25] Mirzazadeh, T., Mohammadi, F., Soltanieh, M., Joudaki, E., 2008. Optimization of caustic current efficiency in a zero-gap advanced chlor-alkali cell with application of
generic algorithm assisted by artificial neural networks. Chem. Eng. J. 140, 157-164.
[26] Morton JF., 2009. Avocado In Fruits of Warm Climates. Creative Resource Systems, Inc., Winterville, NC and Centre for New Crops & Plant Products, Department of
-
World Scientific News 116 (2019) 62-90
-90-
Horticulture and Landscape Architechure, Purdue University, West Lafayette, IN. pp.
91-102. ISBN 0-9610184-1-0
[27] Neuwinger, H. D., 2000. African Traditional Medicine. Medpharm, Stuttgert, Germany.
[28] Odetoye, T.E., D.S. Ogunniyi, G.A. Olatunji, 2012. Improving Jatropha curcas linnaeus oil alkyd drying properties. Progress in Organic Coatings 73, 374-381
[29] Ogunniyi, D.S. and Odetoye, T. E., 2008. Preparation and evaluation of tobacco seed oil-modified alkyd resins. Bioresource Technology, (99) 1300-1304.
[30] Okieimen, F.E. and Aigbodion, A.I., 1997. Studies in molecular weight determination of rubber seed oil alkyd resins. Industrial Crops and Products, 6, 155-161.
[31] Okolie, P.N., 2012. Extraction and characterization of oil from Jatrophacurcas seed. World J. Agric. Sci. 8(4), 359–365.
[32] Onuegbu, N. C.; Adedokun, I. I.; Kabuo, N. O. and Nwosu, J. N. 2011. Amino Acid Profile and Micronutrient Composition of the African Pear (Dacryodes edulis) Pulp.
Pakistan Journal of Nutrition 10(6): 555-557.
[33] Pablo, C.G., Hugo, D.M., Alberto, A.I., Alejandro, J.B., Hector, C.G., 2010. Application of response surface methodology and artificial neural networks for
optimization of recombinant Oryza sativa non-symbiotic hemoglobin 1 production by
Escherichia coli in medium containing byproduct glycerol. Bioresource Technology,
101, 7537-7544.
[34] RMRDC, 2012. Avocado Utilisation: Its Industrial and Economic Potential. Raw Materials Research and Development Council Publication, 76 pp.
[35] Werman, M. J. and Neeman, I. 1990. Avocado Oil Production and Chemical Characteristics. Journal of American Oil Chemists’ Society 64, 2.229-232
[36] J. N. O. Ezeugo, O. D. Onukwuli, M. Omotioma (2017). Optimization of corrosion inhibition of Picralima nitida leaves extract as green corrosion inhibitor for zinc in 1.0
M HCl. World News of Natural Sciences 15, 139-161