(ANN) has been investigated as an effective approach in
the modeling and optimization of bioprocesses. Employing
neural network models would allow the optimal perfor-
mance conditions to be predicted at minimum cost and
with the lowest number of experiments possible. ANN has
successfully been applied to the study and modeling of the
enzymatic synthesis of several esters [7,9].
The objective of the present work was to employ a
solvent-free system for the enzymatic synthesis of adipate
ester in batch and continuous-flow stirred tank reactors.
ANN was used to predict the percentage yield of esteri-
fication between oleyl alcohol and adipic acid. The effects
of four reaction parameters (temperature, reaction time,
impeller speed, and amount of enzyme) on the degree of
esterification were also evaluated, and the optimal condi-
tions for the synthesis of ester were obtained.
2. Materials and Methods
2.1. Materials
Novozym 435, C. antarctica lipase B (EC 3.1.1.3) immo-
bilized on a macro porous acrylic resin (10,000 propyl
laurate units/g) was purchased from NOVO Nordisk A/S
(Bagsvaerd, Denmark). Adipic acid and oleyl alcohol were
purchased from Merck Co. (Darmstadt, Germany). All
other chemicals used in this study were of analytical grade.
2.2. Lipase-catalyzed esterification in a batch reactor
Synthesis of dioleyl adipate was carried out in a 500 mL
stirred tank reactor at a working volume of 350 mL. The reactor was equipped with a Rushton turbine (six-bladed
disc turbine) impeller, temperature control system (Julabo
MB-13, Germany), and sampling ports. Oleyl alcohol and
adipic acid were mixed in the reactor at a molar ratio of
5.3:1. According to our preliminary studies (unpublished
results), this ratio was found to be optimal for achieving
maximum yield in a solvent-free system. Different amounts
of Novozym 435, which were generated by central com-
Table 1. Experimental data of training, validating, and testing of
artificial neural network
Temperature
(
o
C)
Time
(min)
Enzyme
amount
(% w/w)
Impeller
speed
(rpm)
Actual
yield
(%)
Training data
65.0 142.5 3.25 400 72.1
55.0 255.0 5.50 300 93.6
65.0 367.5 3.25 400 94.7
35.0 255.0 5.50 300 85.4
55.0 480.0 5.50 300 91.8
45.0 367.5 3.25 200 87.6
65.0 367.5 7.75 400 93.3
45.0 142.5 7.75 400 89.4
55.0 30.0 5.50 300 60.8
55.0 255.0 10.000 300 92.2
65.0 142.5 7.75 400 92.6
55.0 255.0 1.00 300 69.5
65.0 142.5 7.75 200 88.6
45.0 142.5 3.25 400 65.0
65.0 142.5 3.25 200 69.4
75.0 255.0 5.50 300 91.8
45.0 367.5 3.25 400 92.0
45.0 367.5 7.75 400 94.8
45.0 142.5 3.25 200 60.9
55.0 255.0 5.50 500 96.4
45.0 142.5 7.75 200 85.5
45.0 367.5 7.75 200 92.8
65.0 367.5 7.75 200 91.2
55.0 255.0 5.50 100 89.9
65.0 367.5 3.25 200 92.6
Validating data
55.0 255.0 5.5 500 96.4
65.0 309.0 6.2 500 96.3
50.0 230.0 3.7 300 82.7
60.0 350.0 5.0 500 95.8
48.0 200.0 5.0 200 81.0
Testing data
60.0 437.9 2.5 500 95.5
55.0 320.0 4.0 200 92.6
60.0 323.7 6.0 500 96.2
57.3 328.2 6.2 100 93.5
45.0 255.0 5.5 250 89.9 对于己二酸酯合成在带搅拌的
釜式反应器的操作条件优化
摘要：在无溶剂的系统中进行二酸和油醇的酯化，其系统特点是含有市售固定化南极洲假丝酵母脂肪酶B中的搅拌罐反应器。这过程进行了使用一个人工神经网络（ANN），而ANN是通过文伯格-马夸特（LM）算法训练出来的。人们对四个操作变量：温度，时间，酶的用量和反应产率的影响进行了研究。通过检查不同的神经网络结构，最好的网络被发现由七个隐藏节点组成，而这七个隐藏点采用双曲正切S型传递函数。在实际和预测的响应之间的判定系数和均方根误差值被分别确定为1和0.0058178的培训和0.99467和0.622540的测试数据集。这些结果意着，在开发模式下是能够预测酯化率的。操作变量影响产量做出的贡献的大小顺序依次为：时间>酶用量>温度>叶轮转速。我们通过使用水平低的酶（2.5%W / W），以及温度，时间的控制和叶轮速度分别为66.5℃，354分钟（约6小时），和500转来获得高比例的产量（95.7%）。对于在通过高酶稳定性证明的无溶剂体系中底物转化的一个简单的协议是成功的酯合成的指示。23602 釜式反应器英文文献和中文翻译(2):http://www.youerw.com/fanyi/lunwen_16766.html