FLOWSHEETING PACKAGES: RELIABLE OR FICTITIOUS PROCESS MODELS? F. Benyahia Department of Chemical and Petroleum Engineering College of Engineering United Arab Emirates University
Introduction and background Cheaper computing power and increasing software standards shifting towards Microsoft Windows made process simulation within reach to more people than ever before. The trend towards graphical user interfaces made commercial flowsheeting packages more “user-friendly” and promoted intuitive learning for beginners. Both steady-state and dynamic versions of flowsheeting packages are now available although steady-state codes have been around for much longer, albeit on different platforms. Computationally efficient algorithms are incorporated into flowsheeting codes as they become available. The benefits being faster simulation runs and reduced convergence problems. Flowsheeting packages can be useful in both chemical engineering education and commercial exploitation. Commercial and home-grown flowsheeting packages can be used to generate information needed to optimise processes, develop new processes, study the behaviour of processes under unusual circumstances, etc… However, it is the opinion of the author that without careful training and a good overall chemical engineering background, the process simulation results can turn out to be no better than a series of "science fiction" scenarios instead of credible "what if" scenarios. The preceding statement does not diminish the undisputed power of flowsheeting packages, but simply draws to the attention the possibility of misuse and misinterpretation of simulation results.
Building Process Models Select Thermodynamic Property Method Select Process Components From Database Specify Process Conditions and Feed Specifications Install Unit Operations Select Performance Equations Select Solution Method Export or Analyse Simulation results
Tricks and traps of process simulation using flowsheeting packages Selection of the thermodynamics property method is probably the first most important task. The selection process must be done in conjunction with the components involved. The quality of input information can be more determining than the selection of the equipment in a simulated process. Broadly speaking, kinetic information in reactors and thermodynamic property methods in separation units are the likely sources of problems regarding reliability of simulation results. Use reactor conversion data from actual plant sources whenever available and avoid fragmented kinetic data from the literature. Convergence problems can be severe in column operations but usually can be resolved by tuning damping factors in the numerical solution procedure. Bugs are always present in new versions of the simulation package (however careful vendors are). PREPARE cases on BOTH versions and compare the results. IT IS NOT ENOUGH TO SAVE AN OLD CASE AND OPEN IT ON THE NEW SOFTWARE VERSION. Keep a close supervision on undergraduate students and new process engineers who start using flowsheeting packages until they become “analytically mature” (what’s that?!).
Case Study: the VCM Process How reliable are simulation results obtained from commercial flowsheeting packages? We have simulated a VCM process based on some actual plant input information but employed four recommended thermodynamic property methods. These are: The PRSV (Peng-Robinson-Stryjek-Vera) method. It is an equation of state that is both accurate and recommended for handling moderately non-ideal systems, including those in which water is present. The Kabadi-Daner method. It is a modification of the original SRK equation of state and enhanced to improve VLLE calculations for water-hydrocarbon systems. Extended NRTL (Non-Random-Two-Liquid) method. It is particularly useful for wide boiling range between components and when simultaneous VLE and LLE data are required. It has more binary interaction parameters than the general NRTL method. Antoine vapour pressure model. Assumes ideal mixtures at low pressures and is recommended for heavy components. Petrochemical processes are complex because of the diversity of components and unit operations involved. This approach forms an excellent vehicle for identifying simulation problems and potential misuse and misinterpretation of simulation results. The VCM process modelled has 3 reactors with a total of 16 chemical reactions producing a wide range of chlorinated compounds ranging from light to heavy, creating at least one azeotropic mixture. The separation specifications in the real process are very tight because of the narrow tolerance in the cracking furnace and the polymer grade Vinyl Chloride Monomer. Because of this constraint, we have expressed simulation variations on a percentage basis. The comparison of predicted flowrates, mole fractions and reboiler/condenser energies from a selection of streams are shown in tables 1 and 2. We have used the PRSV method as a base case.
CONCLUSIONS AND RECOMMENDATIONS –Flowsheeting packages are useful computational tools that can aid process design and understanding of process behaviour. However, without sufficient quality input data and a good chemical engineering background, simulation results can be misleading. This was clearly shown in this investigation where the same process model can give widely differing results. In large petrochemical processes where stream specifications are tightly defined, even differences of up to 5 % can be significant in economic terms. In order to obtain reliable simulation results, the author recommends the following: Always use plant data whenever available. Obtain kinetic data from more than one good source. If insufficient kinetic data is available, then use conversion data in conversion reactor models. This will make the simulation more meaningful and reduce computation time. Always run your process model with at least two “suitable” thermodynamics property method. You will find areas of your process where extra caution must be exercised. Whenever possible, use your own binary interaction parameters for liquid activity models otherwise cast doubt on built-in values which often have limited range of applicability. Sometimes values of zero are assigned when there is nothing in the literature for certain compounds (ie assumed to be ideal components).