基于梯度提升决策树的三组元精馏流程结构最优化

doi:10.11949/0438-1157.20230345

摘要/Abstract

摘要：

针对三组元精馏流程结构最优选择问题，提出一种基于梯度提升决策树(gradient boosting decision tree，GBDT)的机器学习方法，利用7种精馏流程结构分离9种三组元混合物的经济评价结果数据用于模型训练。分别应用传统决策树和GBDT模型对多个实际案例进行验证，结果显示无论是对已知物系还是未知物系，提出的GBDT模型相较传统决策树模型具有更高的预测准确率。

关键词: 三组元精馏, 精馏最优流程结构, 机器学习, 梯度提升树, 蒸馏, 计算机模拟, 优化设计

Abstract:

A machine learning method based on gradient boosting decision tree (GBDT) is proposed for the optimal selection of ternary-distillation process structure. Economic evaluation data of 9 ternary mixtures separated by 7 distillation process structures were used for model training. By applying both traditional decision tree and GBDT models to multiple cases, the results show that the proposed GBDT model has higher prediction accuracy than the traditional decision tree model for both known and unknown ternary mixtures.

Key words: ternary-distillation, optimal distillation configuration, machine learning, gradient boosting decision tree, distillation, computer simulation, optimal design

中图分类号:

TQ 062.2

刘尚豪, 贾胜坤, 罗祎青, 袁希钢. 基于梯度提升决策树的三组元精馏流程结构最优化[J]. 化工学报, 2023, 74(5): 2075-2087.

Shanghao LIU, Shengkun JIA, Yiqing LUO, Xigang YUAN. Optimization of ternary-distillation sequence based on gradient boosting decision tree[J]. CIESC Journal, 2023, 74(5): 2075-2087.

图/表 12

图1 三组元精馏结构

Fig.1 Ternary-distillation sequences

表1 三组元物系和相关参数

Table 1 Ternary mixtures and related parameters

编号	物系(A/B/C)	ESI	$α A B$	$α B C$	$α A C$	meanT_b/K	ΔH/(kJ/mol)
A	异戊烷/正戊烷/正己烷(i-C₅/C₅/C₆）	0.4306	1.2992	3.0174	3.9203	317.3647	79.2729
B	正戊烷/正己烷/正辛烷(C₅/C₆/C₈)	0.4159	2.6132	6.2832	16.4194	349.9767	89.1309
C	苯/甲苯/临二甲苯(B/T/oX)	0.8799	2.3406	2.6601	6.2262	384.8667	100.5916
D	乙醇/丙醇/丁醇(C₂OH/C₃OH/C₄OH)	0.9210	2.0076	2.1798	4.3762	371.2300	123.7694
E	正己烷/正庚烷/正辛烷(C₆/C₇/C₈)	1.0157	2.3264	2.2906	5.3288	370.7633	95.1157
F	正戊烷/正己烷/正庚烷(C₅/C₆/C₇)	1.0251	2.7074	2.6412	7.1510	340.8933	86.2980
G	苯/甲苯/乙苯(B/T/EB)	1.1238	2.3604	2.1004	4.9578	382.1233	99.6300
H	甲苯/乙苯/临二甲苯(T/EB/oX)	1.5840	1.9853	1.2533	2.4882	403.5700	105.4700
I	正戊烷/正庚烷/正辛烷(C₅/C₇/C₈)	2.5696	6.1263	2.3841	14.6059	359.8767	92.0696

表1 三组元物系和相关参数

Table 1 Ternary mixtures and related parameters

编号	物系(A/B/C)	ESI	$α A B$	$α B C$	$α A C$	meanT_b/K	ΔH/(kJ/mol)
A	异戊烷/正戊烷/正己烷(i-C₅/C₅/C₆）	0.4306	1.2992	3.0174	3.9203	317.3647	79.2729
B	正戊烷/正己烷/正辛烷(C₅/C₆/C₈)	0.4159	2.6132	6.2832	16.4194	349.9767	89.1309
C	苯/甲苯/临二甲苯(B/T/oX)	0.8799	2.3406	2.6601	6.2262	384.8667	100.5916
D	乙醇/丙醇/丁醇(C₂OH/C₃OH/C₄OH)	0.9210	2.0076	2.1798	4.3762	371.2300	123.7694
E	正己烷/正庚烷/正辛烷(C₆/C₇/C₈)	1.0157	2.3264	2.2906	5.3288	370.7633	95.1157
F	正戊烷/正己烷/正庚烷(C₅/C₆/C₇)	1.0251	2.7074	2.6412	7.1510	340.8933	86.2980
G	苯/甲苯/乙苯(B/T/EB)	1.1238	2.3604	2.1004	4.9578	382.1233	99.6300
H	甲苯/乙苯/临二甲苯(T/EB/oX)	1.5840	1.9853	1.2533	2.4882	403.5700	105.4700
I	正戊烷/正庚烷/正辛烷(C₅/C₇/C₈)	2.5696	6.1263	2.3841	14.6059	359.8767	92.0696

表2 三组元分离产品纯度要求

Table 2 Product purity for the ternary separations

序号	φ_A	φ_B	φ_C
1	0.95	0.95	0.95
2	0.95	0.99	0.99
3	0.99	0.95	0.99
4	0.99	0.99	0.95
5	0.999	0.95	0.999
6	0.999	0.99	0.95
7	0.999	0.999	0.99
8	0.99	0.99	0.99
9	0.999	0.999	0.999

图2 SVR辅助选取最优精馏序列及TAC

Fig.2 SVR-assisted selection of optimal distillation sequence and TAC

图3 严格验证的简捷优化结果（φA∶φB∶φC=0.95∶0.95∶0.95）

Fig.3 Rigorously verified short-cut optimization results

图4 GBDT子树（CART树）可视化

Fig.4 Visualization of GBDT subtree (CART tree)

图5 GBDT流程图

Fig.5 The flowchart of GBDT algorithm

图6 各物系作为独立测试集的模型预测准确率

Fig.6 Model prediction accuracy of each mixture as an independent test set

图7 ESI<1物系最优精馏序列区域划分（φA∶φB∶φC=0.95∶0.95∶0.95）

Fig.7 Regionalization of the optimal distillation sequence for mixtures with ESI<1

表3 三组元物系和相关参数

Table 3 Ternary mixtures and related parameters

物系(A/B/C)	ESI	$α A B$	$α B C$
i-C₄/C₄/i-C₅	0.5173	1.506	2.911
i-C₆/C₆/C₇	0.5069	1.209	2.545
i-C₅/C₅/C₆	0.4306	1.2992	3.0174
C₅/C₆/C₈	0.4159	2.6132	6.2832

表3 三组元物系和相关参数

Table 3 Ternary mixtures and related parameters

物系(A/B/C)	ESI	$α A B$	$α B C$
i-C₄/C₄/i-C₅	0.5173	1.506	2.911
i-C₆/C₆/C₇	0.5069	1.209	2.545
i-C₅/C₅/C₆	0.4306	1.2992	3.0174
C₅/C₆/C₈	0.4159	2.6132	6.2832

图8 增减物系对i-C5/C5/C6预测准确率的影响

Fig.8 Effect of increasing or decreasing mixtures on i-C5/C5/C6 prediction accuracy

图9 特征重要性

Fig.9 Importance of features

参考文献 33

1	余国琮, 袁希钢. 我国蒸馏技术的现状与发展[J]. 现代化工, 1996, 16(10): 7-13.
	Yu G C, Yuan X G. The present status and development of distillation technology in China[J]. Modern Chemical Industry, 1996, 16(10): 7-13.
2	Tsirlin A M, Balunov A I, Sukin I A. Estimates of energy consumption and selection of optimal distillation sequence for multicomponent distillation[J]. Theoretical Foundations of Chemical Engineering, 2016, 50(3): 250-259.
3	Caballero J A. Thermally coupled distillation[J]. Computer Aided Chemical Engineering, 2009, 27: 59-64.
4	Petlyuk F B, Platonov V M, Slavinskii D M. Thermodynamically optimal method for separating multi component mixtures[J]. International Chimerical Engineering, 1965, 5(3): 555-561.
5	Glinos K, Malone M F. Minimum reflux, product distribution, and lumping rules for multicomponent distillation[J]. Industrial & Engineering Chemistry Process Design and Development, 1984, 23(4): 764-768.
6	Glinos K, Malone M F. Minimum vapor flows in a distillation column with a sidestream stripper[J]. Industrial & Engineering Chemistry Process Design and Development, 1985, 24(4): 1087-1090.
7	Glinos K N, Nikolaides I P, Malone M F. New complex column arrangements for ideal distillation[J]. Industrial & Engineering Chemistry Process Design and Development, 1986, 25(3): 694-699.
8	朱怀工, 王燕, 张敏卿. 进料性质对立式隔板塔操作特性的影响[J]. 化工进展, 2009, 28(4): 579-583.
	Zhu H G, Wang Y, Zhang M Q. Influence of feed property on the operation of dividing wall column[J]. Chemical Industry and Engineering Progress, 2009, 28(4): 579-583.
9	Pleşu V, Bonet Ruiz A E, Bonet J, et al. Shortcut assessment of alternative distillation sequence schemes for process intensification[J]. Computers & Chemical Engineering, 2015, 83: 58-71.
10	Takase H, Hasebe S. Synthesis of ternary distillation process structures featuring minimum utility cost using the IDEAS approach[J]. AIChE Journal, 2018, 64(4): 1285-1294.
11	Tedder D W, Rudd D F. Parametric studies in industrial distillation (Ⅰ): Design comparisons[J]. AIChE Journal, 1978, 24(2): 303-315.
12	Agrawal R, Fidkowski Z T. Are thermally coupled distillation columns always thermodynamically more efficient for ternary distillations[J]. Industrial & Engineering Chemistry Research, 1998, 37(8): 3444-3454.
13	吕向红, 陆恩锡. 热耦蒸馏及其选用原则[J]. 化学工程, 2005, 33(2): 9-12, 29.
	Lü X H, Lu E X. Thermally coupled distillation and its selection principles[J]. Chemical Engineering (China), 2005, 33(2): 9-12, 29.
14	袁野. 分离三组分物系精馏方式的模拟和优化选择[D]. 天津: 天津大学, 2015.
	Yuan Y. Simulation and optimization of distillation mode for separation of three components[D]. Tianjin: Tianjin University, 2015.
15	田芳, 袁野, 袁希钢, 等. 物系和分离要求对最优三组元精馏结构选择的影响[J]. 化工学报, 2017, 68(2): 708-715.
	Tian F, Yuan Y, Yuan X G, et al. Influence of mixture and separation requirements on optimal configuration for ternary distillation[J]. CIESC Journal, 2017, 68(2): 708-715.
16	王磊. 三组元最优精馏结构筛选的通用分离因子与定量化规则[D]. 天津: 天津大学, 2019.
	Wang L. General separation factors and quantitative rules for screening the optimal distillation structure of three components[D]. Tianjin: Tianjin University, 2019.
17	陈熙理, 孙国铭, 贾胜坤, 等. 基于决策树的三组元精馏序列结构最优合成规则识别[J]. 化工学报, 2021, 72(3): 1430-1437.
	Chen X L, Sun G M, Jia S K, et al. Identification of rules for optimal synthesis of ternary-distillation configuration based on decision tree[J]. CIESC Journal, 2021, 72(3): 1430-1437.
18	Quinlan J R. Induction of decision trees[J]. Machine Learning, 1986, 1(1): 81-106.
19	De’Ath G, Fabricius K E. Classification and regression trees: a powerful yet simple technique for ecological data analysis[J]. Ecology, 2000, 81(11): 3178-3192.
20	Kotsiantis S B. Decision trees: a recent overview[J]. Artificial Intelligence Review, 2013, 39(4): 261-283.
21	Wankat P C. Multieffect distillation processes[J]. Industrial & Engineering Chemistry Research, 1993, 32(5): 894-905.
22	Jiang Z Y, Mathew T J, Zhang H B, et al. Global optimization of multicomponent distillation configurations: global minimization of total cost for multicomponent mixture separations[J]. Computers & Chemical Engineering, 2019, 126: 249-262.
23	Ma Y J, Luo Y Q, Yuan X G. Simultaneous optimization of complex distillation systems with a new pseudo-transient continuation model[J]. Industrial & Engineering Chemistry Research, 2017, 56(21): 6266-6274.
24	Ma Y J, Luo Y Q, Zhang S, et al. Simultaneous optimization of complex distillation systems and heat integration using pseudo-transient continuation models[J]. Computers & Chemical Engineering, 2018, 108: 337-348.
25	Ma Y J, Luo Y Q, Ma X, et al. Fast algorithms for equation-oriented flowsheet simulation and optimization using pseudo-transient models[J]. Industrial & Engineering Chemistry Research, 2018, 57(42): 14124-14142.
26	Engelien H K, Skogestad S. Multi-effect distillation applied to an industrial case study[J]. Chemical Engineering and Processing: Process Intensification, 2005, 44(8): 819-826.
27	Carlberg N A, Westerberg A W. Temperature-heat diagrams for complex columns (3): Underwood's method for the Petlyuk configuration[J]. Industrial & Engineering Chemistry Research, 1989, 28(9): 1386-1397.
28	Friedman J H. Greedy function approximation: a gradient boosting machine[J]. The Annals of Statistics, 2001, 29(5): 1189-1232.
29	Kearns M, Valiant L G. Learning Boolean formulas[J]. Journal of the Association for Computing Machinery, 1994, 41(6): 1298-1328
30	Zhang M L, Zhou Z H. A review on multi-label learning algorithms[J]. IEEE Transactions on Knowledge and Data Engineering, 2014, 26(8): 1819-1837.
31	Friedman J, Hastie T, Tibshirani R. Additive logistic regression: a statistical view of boosting[J]. The Annals of Statistics, 2000, 28(2): 337-407.
32	Stone M. Cross-validatory choice and assessment of statistical predictions[J]. Journal of the Royal Statistical Society Series B: Statistical Methodology, 1976, 38(1): 102.
33	Stone M. An asymptotic equivalence of choice of model by cross-validation and akaike's criterion[J]. Journal of the Royal Statistical Society: Series B (Methodological), 1977, 39(1): 44-47.

[1]	陈哲文, 魏俊杰, 张玉明. 超临界水煤气化耦合SOFC发电系统集成及其能量转化机制[J]. 化工学报, 2023, 74(9): 3888-3902.
[2]	齐聪, 丁子, 余杰, 汤茂清, 梁林. 基于选择吸收纳米薄膜的太阳能温差发电特性研究[J]. 化工学报, 2023, 74(9): 3921-3930.
[3]	刘远超, 关斌, 钟建斌, 徐一帆, 蒋旭浩, 李耑. 单层XSe₂（X=Zr/Hf）的热电输运特性研究[J]. 化工学报, 2023, 74(9): 3968-3978.
[4]	文兆伦, 李沛睿, 张忠林, 杜晓, 侯起旺, 刘叶刚, 郝晓刚, 官国清. 基于自热再生的隔壁塔深冷空分工艺设计及优化[J]. 化工学报, 2023, 74(7): 2988-2998.
[5]	邵远哲, 赵忠盖, 刘飞. 基于共同趋势模型的非平稳过程质量相关故障检测方法[J]. 化工学报, 2023, 74(6): 2522-2537.
[6]	江锦波, 彭新, 许文烜, 门日秀, 刘畅, 彭旭东. 泵出型螺旋槽油气密封泄漏特性及参数影响研究[J]. 化工学报, 2023, 74(6): 2538-2554.
[7]	孙永尧, 高秋英, 曾文广, 王佳铭, 陈艺飞, 周永哲, 贺高红, 阮雪华. 面向含氮油田伴生气提质利用的膜耦合分离工艺设计优化[J]. 化工学报, 2023, 74(5): 2034-2045.
[8]	周必茂, 许世森, 王肖肖, 刘刚, 李小宇, 任永强, 谭厚章. 烧嘴偏转角度对气化炉渣层分布特性的影响[J]. 化工学报, 2023, 74(5): 1939-1949.
[9]	王泽栋, 石至平, 刘丽艳. 考虑气泡非均匀耗散的矩形反应器声流场数值模拟及结构优化[J]. 化工学报, 2023, 74(5): 1965-1973.
[10]	李纪元, 李金旺, 周刘伟. 不同扰流结构冷板传热性能研究[J]. 化工学报, 2023, 74(4): 1474-1488.
[11]	许文烜, 江锦波, 彭新, 门日秀, 刘畅, 彭旭东. 宽速域三种典型型槽油气密封泄漏与成膜特性对比研究[J]. 化工学报, 2023, 74(4): 1660-1679.
[12]	陈俊先, 姬忠礼, 赵瑜, 张倩, 周岩, 刘猛, 刘震. 基于微波技术的天然气管道内颗粒物在线检测方法研究[J]. 化工学报, 2023, 74(3): 1042-1053.
[13]	袁海鸥, 叶方俊, 张硕, 罗祎青, 袁希钢. 考虑中间换热器的能量集成精馏序列合成[J]. 化工学报, 2023, 74(2): 796-806.
[14]	魏进家, 刘蕾, 杨小平. 面向高热流电子器件散热的环路热管研究进展[J]. 化工学报, 2023, 74(1): 60-73.
[15]	袁妮妮, 郭拓, 白红存, 何育荣, 袁永宁, 马晶晶, 郭庆杰. 化学链燃烧过程Fe₂O₃/Al₂O₃载氧体表面CH₄反应：ReaxFF-MD模拟[J]. 化工学报, 2022, 73(9): 4054-4061.