How can software simulation be integrated with actual production?
Thread Content
Software simulation primarily relies on mathematical solutions derived from theoretical calculations using certain thermodynamic methods, based on literature or experimental data; it also requires experimental verification or integration with actual production processes. How can we improve the consistency between the two and make the simulation more accurate to reality? For example, when simulating tower separation in PRO/II, how can one take into account setting the efficiency of the tower based on its size and the material being processed during simulation? How can one ensure the composition of the overhead product while also making the number of tray levels sowie the load on the condenser and reboiler as close as possible to those in reality? This facilitates the calculation of energy consumption ； Another example is simulated heat exchangers: are there considerations regarding heat transfer efficiency to make the amount of condensate used more accurate to reality? This would provide better guidance for selecting energy-consuming devices, as well as pipes and pumps. I have limited knowledge and experience in simulation and manufacturing processes, so what I’ve said above may not be entirely accurate. Please consider it just an attempt to initiate discussion; I hope experts on this forum can offer their guidance, and I welcome everyone’s participation in the conversation.
When it comes to combining simulation with reality, I immediately think of regression of binary interaction parameters. In fact, the thermodynamic methods used in simulation software are mostly empirical or semi-empirical formulas; therefore, simulation and reality are inherently inseparable. PROII is a computational tool, but what really matters are the experience of the person operating it and the quality of the data used. If you have the correct thermodynamic methods, provide accurate binary interaction parameters, set up the flowchart based on actual conditions, and use board efficiencies that reflect reality, then the reflux ratios, condensation and reboiling loads, etc., that you obtain will certainly be accurate. Conversely, to quote an expert, if you input junk data, then the results you get will also be junk
I also conducted simulations for two years, mainly using PRO 2. In my opinion, it’s necessary to have some practical experience or on-site data; practical experience can be gained through accumulation and learning, while on-site data needs to be collected carefully. Otherwise, it’s impossible to determine whether the results of the simulations are correct or not. Practical experience or on-site data is needed to adjust the model, so that it can be useful for guiding production and design.
It makes a lot of sense; this is actually a characteristic of the chemical industry – any calculations should be adjusted based on actual data. Yet many companies don’t pay enough attention to this. Our company specializes in chemical processing equipment, but we fail to collect data on how those devices are used, especially usage statistics, which can be quite frustrating at times.
What was said upstairs is also a very practical situation – people rarely go out of their way to collect such data. This means that when conducting simulations, we need to compare and calculate more with the data obtained on-site, in order to adjust the binary parameters; only in this way can the errors in the simulation results be reduced, and the results can be accurate and reliable
Recently, I have been working on the regression of vapor-liquid equilibrium data for a ternary system. Different thermodynamic methods all have their drawbacks: some can only effectively relate data for one or two binary systems, while others can only work well for vapor-liquid equilibrium data at high or low pressures. It is necessary to compare the vapor-liquid equilibrium data calculated using the binary interaction coefficients derived from various thermodynamic methods with experimental data, in order to identify the method that results in the smallest discrepancies. While this method can provide a good fit within the experimental range, as the temperature increases, the calculated values for the bubble point line match the experimental values well, but the calculated values for the dew point line begin to deviate significantly from the experimental values. By applying these regression-based binary interaction coefficients, the vapor phase composition and equilibrium temperature in a ternary system were calculated given the liquid phase composition and equilibrium pressure; the results showed that the temperature deviation was within 3%, and the absolute deviation in vapor phase composition was within 1.3%. Is it certain that these interaction coefficients can be used to simulate actual production? There are some concerns: 1. These experimental data were obtained using self-built equipment; although careful efforts were made, the accuracy is uncertain, and it’s not clear whether the consistency among the data implies reliability ； 2. Compared with mature binary interaction coefficients, the deviation between simulation and production can be adjusted by empirically setting efficiencies. However, when fitting binary interaction coefficients using these experimental data to simulate distillation separation, and in the absence of actual engineering data, can boundary conditions such as efficiencies be set using the aforementioned empirical methods? I think this must be combined with actual production. In other words, relying solely on laboratory data to simulate real-world processes is not sufficient.