粒子群响应面建模法在ASPEN多因素优化中的应用

发布时间：2018-08-23 14:33

【摘要】：响应面优化法可以建立试验因素与响应值之间近似的数学模型,从而进一步分析响应值的最优化问题,所以该方法目前被频繁应用到工程和科学试验领域。其精准度受试验次数以及试验点结构的影响较为明显,而试验次数又将影响试验周期的长短以及资金的投入。因此,如何利用有限的试验数据来获得到更加精准的响应面模型将显得尤为重要。经过研究响应面模型的求解方法—最小二乘拟合法,得知其本质是根据试验数据,以方差的平方和最小为前提来求解回归方程中的系数。在深入了解这一数学机理后,本文结合粒子群算法PSO随机性强、收敛速度快、处理非线性能力强、稳定性高等特点,提出一种新的响应面模型建立方法—粒子群响应面建模法。本文以醋酸水溶液为研究对象,采用粒子群响应面法对分离醋酸的试验建立响应面模型。利用流程模拟软件Aspen Plus模拟了萃取精馏醋酸的两塔工艺流程,对影响能耗和产品中醋酸质量分数的萃取精馏塔原料进料位置X1、萃取剂进料位置X2、回流比X3、萃取剂进料量X4以及溶剂回收塔原料进料位置X5、回流比X6等6因素条件进行了灵敏度分析。然后选取这六个主要因素的最佳取值范围,将其分为五个水平,进行正交试验,该方法有效减地少了设计点的数量。分别利用待定系数法和粒子群响应面法对产品中醋酸的含量R1和两塔再沸器总能耗R2进行回归方程的拟合。结果显示待定系数法在处理某些参数较多、多项式较复杂情况时不能得到正确的拟合方程,而粒子群响应面法可以处理这种复杂的拟合问题,且得到的响应面精度符合要求。根据工业的上要求,本文以产品中醋酸质量分数不得低于99.5%为约束条件,以总能耗最小为优化目标,利用PSO算法进行约束寻优,得到了满足产品质量分数要求的最小能耗以及相应的操作工艺条件。结果显示,产品醋酸的质量分数为0.9982时,两塔再沸器的总热负荷为5372kW。与文献中利用Aspen Plus的Sensitivity模块进行单因素分析的结果6545kW相比,在节能方面具有很大优势。该方法可以对复杂的精馏塔系统进行更好的优化,对工业设计和生产具有一定的指导意义。
[Abstract]:The response surface optimization (RSM) method can be used to establish an approximate mathematical model between the experimental factors and the response values to further analyze the optimization problem of the response values, so this method is frequently used in engineering and scientific experiments. The accuracy of the test is influenced by the number of times of test and the structure of test site, and the number of times of test will affect the length of test period and the investment of funds. Therefore, it is very important to obtain a more accurate response surface model by using limited experimental data. By studying the least square fitting method of response surface model, it is found that its essence is to solve the coefficients of regression equation based on the experimental data and the minimum sum of square variance. After deeply understanding this mathematical mechanism, this paper proposes a new response surface modeling method, which is called particle swarm optimization (PSO), which is characterized by strong randomness, fast convergence speed, strong nonlinear ability and high stability of particle swarm optimization (PSO) algorithm. In this paper, the particle swarm response surface (PSO) method is used to establish a response surface model for the separation of acetic acid from acetic acid aqueous solution. The two-column process flow of extractive distillation of acetic acid was simulated by using the process simulation software Aspen Plus. The feed position of extraction distillation column X1, the feed position of extractant X2, the ratio of reflux X3, the feed quantity of extractant X4, the feed position of solvent recovery column X5, the ratio of reflux X6 and so on, which affect the energy consumption and the content of acetic acid in the product, are discussed. Sensitivity analysis was carried out. Then the optimum value range of the six main factors is selected and divided into five levels. The orthogonal experiment is carried out. The method effectively reduces the number of design points. The undetermined coefficient method and particle swarm response surface method were used to fit the regression equations of the content of acetic acid R1 and the total energy consumption R2 of two reboiler. The results show that the undetermined coefficient method can not get the correct fitting equation when some parameters are more and the polynomial is more complex, while the particle swarm response surface method can deal with the complex fitting problem, and the precision of the response surface can meet the requirements. According to the requirements of industry, this paper takes the acetic acid content in the product not less than 99.5% as the constraint condition, takes the minimum total energy consumption as the optimization goal, and uses the PSO algorithm for the constraint optimization. The minimum energy consumption and the corresponding operating conditions were obtained to meet the requirement of product mass fraction. The results show that the total heat load of the reboiler is 5372kW when the mass fraction of acetic acid is 0.9982. Compared with the results of single factor analysis using Sensitivity module of Aspen Plus in literature, 6545kW has a great advantage in energy saving. This method can better optimize the complex distillation column system and has certain guiding significance for industrial design and production.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP18

【相似文献】