A company wants to get full value from expensive software. So, the newest (and least experienced) engineer often works with the program because that person has the freshest and best computer skills. Management forgets that proficiency with a computer doesn't equal ability to understand and get the right answer. No one wants to hear that results are worthless — or, worse yet, misleading. However, frequently no answer is better than a wrong answer.
In the past two weeks alone, four examples of seriously wrong calculated results crossed my desk. Each came from standard, commercially available process simulators. The engineers didn't realize the built-in data methods only apply over the ranges of data used to generate physical and thermodynamic properties of the systems.
Solvent removal system. This uses batch vacuum flashing to pull out solvent from a solvent/water/polymer mixture. The simulator calculated an absolute pressure of 76 torr would reduce solvent composition to the allowed value of 1,000 ppm. Always be wary of composition values predicted in the parts per million range from process simulators. The data set used often doesn't extend down that far.
Henry's Law prediction for equilibrium conditions for this system provided the necessary reality check. It showed a vacuum pressure of 3 torr would be required. At 76 torr, the composition actually would be 22,000 ppm. A completely different vacuum system and much longer batch times were needed to get the solvent concentration down.
Water vaporization in mixed hydrocarbon/water systems. A consultant predicted an oil/water stream could be heated an extra 26°F. The operational limit was how hot the stream could get before water vaporized, which would create massive hydraulic problems in the unit. Several different methods and plant data showed the "best" prediction of water flash point allowed only a 16°F increase. This moved the project from one with an adequate return to one on life support.
Water dew point prediction. The temperature, and hence location, of water dew point in systems significantly impacts which materials are selected to prevent corrosion. Prediction often is based on water partial pressure and the steam table value of condensation temperature at that pressure. Many data methods handle water as a special case and use these partial pressure approaches. However, in systems with significant concentrations of ionic species actual water dew point temperature can be 40°F or 50°F higher. Problem systems include ones with high concentrations of ammonia, hydrogen sulfide, carbon dioxide and other acid or base gases. The standard approach is to set operating conditions or select materials with an experience-based dew point margin. In some cases, the standard margin isn't large enough.
Electrolyte thermodynamic methods can provide accurate prediction, but they take a lot more data and are rarely used. As a result, units suffer corrosion in unexpected locations.
Amine treating for acid gas removal. Alkanolamines absorb acid gases such as hydrogen sulfide and carbon dioxide by weak chemical bonding with contaminants. Careful selection of operating conditions, equipment parameters and solvent will even allow selective absorption of one component and "slip" of a second through the system. A question arose on reboiler design for a plant intended to absorb hydrogen sulfide but allow carbon dioxide to slip into the gas product. One method to predict performance is to start with data from a very similar system. Lacking field data, complex alkanolamine systems require analysis with reaction kinetics, equilibrium and enthalpies, not conventional equilibrium approaches. However, the reboiler was being designed with a simulator with a standard — and inaccurate — amine prediction package. It gave dramatically optimistic values for duties, equipment sizes and performance. Calculating slip is difficult in the first place; using inaccurate basic data makes accurate design impossible.
Modern simulators can provide extraordinarily useful answers. They can also give you incredible garbage. Just because a simulator can calculate a value doesn't mean it's the right answer. Understand the basic data behind your prediction methods. If the software company can't, or won't, give you some basis for its range of applicability, don't use the data. In special cases, confirm at least a few operating points against reputable field or literature data before accepting results. Specific cases more prone to problems include critical components at low concentrations, multiple liquid phase systems, highly ionic systems, and operations near the critical point.
Andrew Sloley is a Chemical Processing Contributing Editor. You can e-mail him at ASloley@putman.net.