Dec-2020

Should you rely on your simulation results?

Avoiding some of the pitfalls of process simulation

ATUL CHOUDHARI
Tata Consulting Engineers

Article Summary

Process simulation is a useful and powerful tool to model chemical process flowsheets of varying complexities. Modern day simulators are built with a comprehensive, pure component databank, an exhaustive library of thermodynamic systems, physical property estimation methods, initial estimate generators, and inbuilt algorithms for every unit operation with a user-friendly graphical interface. These simulators can solve and optimise virtually any flowsheet synthesis problem. However, despite the simulator’s sophisticated and rigorous modelling techniques, process simulations at times fail to represent real life plant data. In most cases, the user shows blind faith in the inbuilt configurations and default selection of methods in simulators, which can introduce erroneous results for specific systems. Since simulators can generate multiple and, at times, conflicting solutions for the same set of external input data parameters, this raises doubts about their effectiveness and reliability. There are a number of reasons for simulation failure. This article discusses various issues that may help users to derive meaningful results from simulation and thus enhance the reliability of their simulations.

Elements of process simulators
With advances in computer technology and the availability of modern tools, commercial steady state simulation software packages have become an integral part of process design practices. Process simulation is utilised in all stages of process plants, from concept and feasibility, to basic design, to detail design, even extending to commissioning and revamps. Process flowsheet synthesis using simulation software yields a heat and mass balance that is consistent with thermodynamics. This heat and mass balance is a firm basis for all downstream process design activities including preparation of equipment and instrument specifications.

Mathematical simulation models offer a drag-drop type of graphical interface for setting up process flowsheet configurations. Graphical user interfaces have matured to the extent that they make it easy to build large and complex models; they also provide a step-wise menu and online help to enable users to define input parameter specifications easily.

Simulators serve a wide variety of unit operations in refining and include mathematical solver algorithms specific to every unit operation model. For example, distillation columns can be solved by selecting any of the suitable prebuilt algorithms like Inside Out or chemdist. In short, virtually all unit operations are well supported and can be easily defined using a typical simulation suite. These mathematical models are supported with internal databases covering physical and thermodynamic property databanks. In a bid to enhance the capability of the software, these simulation packages present multiple choices for the user to configure and then to solve the flowsheet. A few of the important user choices include selection of thermodynamic methods (for calculating pure component and mixture physical properties) and/or specific algorithms for solving individual unit operations, deciding calculation sequences, methods for generating initial estimates, methods for defining a non-library component, and so on.
More choice, more confusion?

As more options become available, it is expected and assumed that the user is knowledgeable enough to select appropriate methods and employ correct choices to solve the flowsheet. Since simulation software uses advanced high-end computer hardware for such sophisticated and advanced simulation programs, there is no programming limit with respect to presenting multiple choices to the user. In fact, they are increasingly interpreted, falsely, as ‘more choice, more capability’ products. However, so many choices surely pose a challenge for process simulation engineers to check the applicability and suitability of every choice to model a given process, verify the advantage or disadvantage of the selection, and determine its effectiveness over the operating range. Thus, to utilise the capability of the simulation software, adequate knowledge and specific process experience become a prerequisite. However, this basic and fundamental issue could affect the quality of simulation output results.

With increased availability of simulation models, it has become easier for users to study various design cases. Normally, these design cases include modelling the same flowsheet for various case studies, for instance the effect of processing different feedstocks, or checking performance under different operating conditions. In today’s world of high computing speeds, computation time is no longer a constraint while selecting or adding as many check cases as required. The user can check performance for multiple cases, conduct what-if analysis, perform sensitivity studies, and optimise process designs.

Using a simulation tool, one can significantly increase the profitability of a process by optimising design and operating parameters. Thus, simulators provide a reliable platform by solving any process flowsheet using an inbuilt modelling approach. However, a few problems still exist with respect to multiple solutions that can be generated for the same flowsheet by different users using the same simulator and using the same set of input data. Multiple solutions are possible as multiple choices are available to users while configuring the flowsheet. As an example from experience, a typical glycol dehydration tower for the same input and output specifications will require eight theoretical stages if an equation of state based thermodynamic model is selected against 12 theoretical stages if a liquid activity coefficient based thermodynamic model is selected. In both cases, the simulator will converge but will yield two different hardware configurations for the same mass balance.

Blind faith by inexperienced engineers
Knowing how to configure a flowsheet in a simulation environment and familiarity operating the simulator does not necessarily confirm the availability of skills required to effectively solve and analyse flowsheet synthesis results. Here is the main hitch. Although simulator programs provide early warnings for missing input data or inadequate input data, there may be no warning if the user selects an inappropriate calculation method. This is especially true when blind faith is exhibited by users on the applicability of various choices such as selection of a thermodynamic system. As Figure 1 shows, modern day simulators carry a wide variety of unit operation libraries, expanded thermodynamic data libraries, comprehensive pure component and binary interaction parameter databases, initial estimate generators, and so on. Looking at these capabilities, it seems that most chemical plant modelling problems or process design and optimisation problems can be dealt with accurately and reliably. Unfortunately, this is not a reality.

Accuracy of simulation output is critical, irrespective of the reputation of the simulation package deployed, or the speed to obtain a solution. Unless the simulation model accurately describes the interaction of different components at varying temperatures or pressures using reliable methods, simulation results will not represent reality. Simulators are just automated mathematical model solvers based on inputs provided by users. Therefore, simulation input parameters require skilful scrutiny. Also, the results need to be carefully analysed based on fundamental principles and the specific objectives of a simulation.

Differences in simulation and design
It is now widespread practice to employ process simulators to solve design cases. Although a simulator can be configured to provide solutions to a design problem, there is a difference in approach and method using simulators to arrive at a solution. In order to obtain a uniform solution, any design problem is solved in a step-wise manner utilising a fixed and defined method or procedure for every step. However, when it comes to solving the same design problem using a simulator, an iterative or case study type of approach has to be adopted. Before dealing with a simulation’s pitfalls, it is important to understand the differences between a design problem and a simulation problem.