Until the early 1980's, most refining, chemical and petrochemical (process) companies in the US and elsewhere employed a combination of decentralized manual and computer-based methods, to meet their process calculation needs. Some of the majors even developed powerful in-house process simulation programs for this purpose, typically tailored to processes of interest. During the last 2-3 decades, however, third-party process simulators have gained in popularity, and become the de facto standard tools for most industry-wide process calculations. Over time, these programs have become more versatile, more reliable and easier to use. In other words, process simulators have become an extremely attractive choice.

Unfortunately, this has introduced an additional layer of separation between the practicing engineer and the chemical engineering process she / he is attempting to describe. A key feature in this modeling process is that the quality of results obtained from a process simulator is strongly dependent on simple semi-empirical thermophysical methods at the heart of all its calculations. At the same time, most thermophysical methods implemented in a process simulator are limited in applicability. This limitation often arises from the fact that these methods need to be simple to ensure robustness and fast calculations, often at the expense of accuracy. The other limitation, which is quite common, is the scarcity of measured data to derive model parameters from. Typically, these simple models are unable to span wide ranges in pressure, temperature and composition (PTx)

However, it is vitally important for a process simulator program to be continuous across PTx as well as time. In most cases, it is not feasible for a process calculation to be terminated because the underlying science is weak. For example, an approximate answer may be a reasonable step in an iterative calculation. Approximations involving techniques such as linear extrapolation, scaling, localized curve-fitting and renormalization are routinely employed inside process simulation calculations. Such assumptions can lead to annoyance and wasted hours, but also to bad numbers which may be unacceptable. Unfortunately, the underlying limitations are often not recognized or underestimated by many process engineers because, simply, a process simulator is too easy to use!

This investigation takes a closer behind-the-scene look at the “science” of thermodynamics approximations that are routinely employed inside process simulators. The information is presented as several real-world cases encompassing heat effects, density solving, flash calculations, distillation analysis, and regression of thermodynamic data. The assumptions leading to unacceptable answers are analyzed, along with the improved techniques employed to maintain robustness while improving accuracy.