The Effects of Uncertainties in Physical Property Data and Models on Chemical Process Design

古普塔9321865对流体相平衡的测量必然包含实验误差。根据这些数据开发的当前模型(状态方程和活度系数模型)通常比数据的精确度低得多。因此，基于这些模型的过程模拟结果的质量是有限的。本项目的目标是开发一种合理而严谨的方法来评估热力学实验、模型和设计中的不确定度之间的关系。方法是使用蒙特卡罗模拟来研究基本情况(与学术界和工业界同事合作选择)的敏感性和不确定性，从而制定和评估用于实验设计、模型开发和过程设计的启发式方法。初步结果表明，当考虑到汽液平衡数据中看似很小的不确定性时，工艺的设计和操作中存在着非常大的不确定性。在化学加工工业中，一种常见的方法是将安全系数应用于设计，以解决这种不确定性。然而，蒙特卡罗不确定度分析可以量化不确定度，从而使安全系数法更加合理。从实验数据建立热力学模型需要对模型参数进行回归，这些回归的结果可以作为蒙特卡罗模拟的输入。然而，最大似然回归方法的经典应用假设模型本质上比数据更准确。对于相平衡数据，通常情况正好相反。因此，将开发一种方法来消除这种不一致，并允许人们在合理的基础上比较模型及其精度。这项工作的一个最终目标是将物理性质的不确定性分为两类：由模型不确定性引起的不确定性和由实验不确定性引起的不确定性。将进行蒙特卡洛模拟，以制定实现这种分离的策略。***

Gupta 9321865 Measurements of fluid-phase equilibrium necessarily involve experimental error. Current models developed from these data (both equations of state and activity coefficient models) typically are significantly less precise than are the data. The quality of process simulation results based on these models is, thereby, limited. The objective of this project is to develop a rational and rigorous method for evaluating the relationships between the uncertainties in thermodynamic experiments, models, and design. The approach is to use Monte Carlo simulation to study the sensitivities and uncertainties of base cases (chosen in collaboration with academic and industrial colleagues) and thereby to formulate and to evaluate heuristics for experimental design, model development, and process design. Preliminary results show that there is a very significant uncertainty in the design and operation of processes when one considers the seemingly small uncertainties in data for vapor-liquid equilibrium. A common approach in the chemical processing industries is to apply safety factors to designs to account for this uncertainty. However, Monte Carlo uncertainty analysis can quantify the uncertainty and lead to rationalization of the safety factor approach. The development of thermodynamic models from experimental data requires the regression of model parameters, and the results of these regressions can serve as input to the Monte Carlo simulations. However, the classical application of the maximum likelihood regression method assumes that the model is inherently more accurate than are the data. For phase-equilibrium data, often the inverse is true. Therefore, an approach will be developed that eliminates this inconsistency and allows one to compare models and their accuracy's on a rational basis. One ultimate goal of this work is to separate the uncertainty of physical-property into two categories: those caused by model uncertainty and those caused by experimental uncertainty. Mon te Carlo simulations will be performed to develop strategies to make this separation. ***