Effect of Systematic and Random Errors in Thermodynamic Models on Chemical Process Design and Simulation:  A Monte Carlo Approach

Abstract

A Monte Carlo method is presented to separate and study the effects of systematic and random errors present in thermodynamic data on chemical process design and simulation. From analysis of thermodynamic data found in the literature, there is clear evidence of the presence of systematic errors, in particular, for liquid−liquid equilibria data. For systematic errors, the data are perturbed systematically with a rectangular probability distribution, and to analyze random errors, the perturbation is carried out randomly with normal probability distributions. Thermodynamic parameters are obtained from appropriate regression methods and used to simulate a given unit operation and to obtain cumulative frequency distributions, providing a quantitative risk assessment and a better understanding of the role of uncertainty in process design and simulation. The results show that the proposed method can clearly distinguish when one type of error is more significant. Potential applications are safety factors, process modeling, and experimental design.

New association schemes for mono-ethylene glycol: Cubic-Plus-Association parameterization and uncertainty analysis

“…Whiting and co-workers extensively studied the sensitivity of process design to experimental errors and propose ways to quantify and apply uncertainty analysis [77][78][79][80]. While this list of studies is not exhaustive, it provides a good sampling of the type of investigations completed.…”

On Monte Carlo simulations of the LAser RElativity Satellite experiment

“…Actually, it does not seem a plausible choice since their values are affected not only by random measurement errors 1 but also-to a predominant level-by systematic errors of several types. In this cases, systematics cannot be treated with Gaussian distributions in Monte Carlo simulations [74,79,80].…”

ThermoData Engine (TDE): Software Implementation of the Dynamic Data Evaluation Concept. 8. Properties of Material Streams and Solvent Design

“…The methods used were not described in sufficient detail to reproduce their approach, but, through the article series, their approach evolved from variation of parameters based on parameter variance without consideration of covariance 43 to sampling of raw data with an assumed normal distribution. 46 Hajipour and Satyro 47 used a similar approach in uncertainty analysis applied to thermodynamic models and process design involving pure components. Recently, Theodorou et al 48 published an attempt to consider all uncertainty contributions to the raw (experimental) values with different distributions and correlation of errors for a single-valued property: enthalpy of combustion of fuels at fixed conditions and composition.…”