Detailed Optimisation Results for a Batch Extractive Distillation Process Obtained by Hybrid Optimisation Methods

The data presented in the four Excel files were used in the paper "Laszlo Hegely: Optimisation of a Batch Extractive Distillation Process with Hybrid Methods, 2025". They show the data points in which the simulation of the batch extractive distillation process was performed, as well as the simulation results. By the data points, we refer to points in the seven-dimensional space of the optimisation variables, which were the reflux ratios of Steps 1-3 of the procedure (R1, R2, R3), the stopping criteria of Steps 1 and 2 (Cr1 and Cr2), and the flow rate (FE) and duration (tF) of the water feeding. (Values of tF/2 are given in the files.) The stopping criteria are the values of tetrahydrofuran concentration (in mass fraction) in the (instantaneous) distillate when the given step is finished. Selected simulation results, the ones necessary to calculate the profit of the process, are also shown: duration of the process (t), mass of the Fore-cut 1 (mfc1) and of the main cut (mmc), the concentration of methanol, THF and toluene in Fore-cut 2 (xfc2,B, xfc2,C and xfc2,E, respectively), the final (xmc,B) and maximal (xmc,B,max) concentration of methanol in the main cut, as well as the profit (OF).
The different files correspond to the different optimisation methods used. They contain various sheets that correspond to different optimisation runs; they are identified with the notation used in the corresponding paper.
The file “GA run data.xlsx” contains the data of the best individual of each generation when only the genetic algorithm is used, with the first line being the first generation. Apart from the data mentioned above, it also contains the ratios C1=xfc2,C/xfc2,B and C2=xfc2,E/xfc2,B, which served as constraints. The elements of profit, the income, incineration and steam cost are also given.
The results of Latin hypercube sampling are in the file “LHS run data.xlsx”. Simulations prematurely terminated are indicated with the symbol “x” in the column “t, min”.
The file “Simplex run data.xlsx” contains the results of the optimisation performed with the Nelder-Mead simplex algorithm. Each run has two sheets. The one whose name starts with “History” contains the initial simplex, including its side length (a), and the list of new points generated by the algorithm. In the latter list, the average (Avg. OF) and standard deviation of the OF values (SD OF) of the points in the simplex are also given. The sheet with the name starting with “AS” shows all the simplexes in sequence, the average and standard deviation of OF (columns Q and R), the simplex size (column S) and the operation performed.
The file “Complex run data.xlsx” presents the data of the current best point of the complex in the rows of each sheet. “n” is the number of simulations performed, “d_avg” is the average scaled distance between the point of the complex.

本数据集包含四个Excel文件中的数据，已被应用于论文《Laszlo Hegely: Optimisation of a Batch Extractive Distillation Process with Hybrid Methods, 2025》中。这些数据涵盖了分批萃取精馏（Batch Extractive Distillation）过程模拟所采用的采样点，以及对应的模拟结果。所谓采样点，指的是优化变量七维空间中的点，这些优化变量包括流程步骤1至3的回流比（R1、R2、R3）、步骤1与步骤2的终止判据（Cr1、Cr2），以及供水流量（FE）与供水时长（tF）。（文件中给出的是tF/2的数值。）终止判据为对应步骤结束时，（瞬时）馏出物中四氢呋喃（tetrahydrofuran）的质量分数。此外还提供了计算该过程收益所需的核心模拟结果：总流程时长（t）、第一前馏分质量（mfc1）与主馏分质量（mmc）、第二前馏分中甲醇（methanol）、四氢呋喃（THF）与甲苯（toluene）的浓度（分别对应xfc2,B、xfc2,C与xfc2,E）、主馏分中甲醇的最终浓度（xmc,B）与最大浓度（xmc,B,max），以及过程收益（OF）。

不同Excel文件对应所采用的不同优化方法，每个文件包含多个工作表，分别对应不同的优化运行任务，其命名采用对应论文中使用的标注规则。

文件"GA run data.xlsx"包含仅使用遗传算法（Genetic Algorithm, GA）时每一代的最优个体数据，首行对应第一代种群。除上述数据外，该文件还包含作为约束条件的比值C1=xfc2,C/xfc2,B与C2=xfc2,E/xfc2,B，同时提供了收益构成项：总收入、焚烧成本与蒸汽成本。

拉丁超立方抽样（Latin Hypercube Sampling, LHS）的结果存储于文件"LHS run data.xlsx"中。提前终止的模拟结果在"t, min"列中以符号"x"标注。

文件"Simplex run data.xlsx"包含使用Nelder-Mead单纯形算法完成的优化结果。每次优化运行对应两个工作表：名称以"History"开头的工作表包含初始单纯形信息（包括其边长a）以及算法生成的新采样点列表，该列表中还会给出单纯形内所有点的收益值平均值（Avg. OF）与标准差（SD OF）；名称以"AS"开头的工作表按顺序展示所有迭代单纯形，以及OF值的平均值与标准差（分别对应Q列与R列）、单纯形尺寸（S列）与本次迭代执行的操作。

文件"Complex run data.xlsx"的每个工作表的行数据均对应复合形法（Complex Algorithm）当前的最优采样点。其中"n"为已执行的模拟次数，"d_avg"为复合形法采样点间的平均缩放距离。