Some questions regarding Aspen Plus data regression
Thread Content
This post was last edited by Yi Ran 0517 on 2017-5-5 at 11:33. I conducted a parameter regression for the equation used to calculate saturated vapor pressure (the Antoine equation) and discovered some interesting issues; let’s discuss them together. For the specific process, see the simulation file. First, after entering the experimental data on saturated vapor pressure versus T (temperature), regression of the parameters for the Antoine equation is performed. By checking the help file, it can be seen that the extended Antoine equation consists of 9 parameters in total; parameter 1 is related to temperature and pressure, while C8 and C9 represent the temperature range. During the regression process, by looking at some training materials, it can be seen that the parameters obtained through regression vary from one person to another; there are regressions for C1, C2, C4, and C5, as well as regressions for other parameters. Of course, different parameters yield different parameter values. My understanding is that the parameter values are determined based on experimental data, through linear fitting; therefore, since the parameters used for regression differ, the resulting parameter values will naturally vary as well. So here comes a question: which parameters are appropriate to regress in the Antoine equation? Below are the regression results for C1, C2, C3, C4, as well as for C1, C2, C4, and C5. Before performing the regression, it is necessary to remove the Plxant results; otherwise, the regression will be conducted based on the existing regression parameters, resulting in different outcomes. The experimental results are generally consistent with those obtained from the various regression simulations.
Okay, thank you. There is one more question: during the simulation, it was found that the results obtained through regression using different property calculation methods varied as well. Is this due to different calculation formulas for plexant in these various methods, or is there some other reason?
It doesn’t seem right; you’re simply using temperature and pressure values to derive a function relating temperature to pressure. Even without the Aspen environment, it’s possible to do this regression using other software. It should have nothing to do with physical property methods, right?
Attach two images for an intuitive understanding. I don’t quite understand this either, because it’s possible to view the corresponding parameters of Plxant directly without using property-based methods. However, it can be seen from the graph that the ideal regression results should meet the requirements, while the SRK standard deviation is too high and thus fails to meet those requirements
Upload the file to take a look: dizzy:
It’s included in the theme – an ideal compressed file
In the simulation process, you need to first remove the initial parameters, as shown in the figure above; simultaneously, the plxant term also needs to be removed. Otherwise, the results will remain the same. Please refer to the attachment for details. The results, as shown on the floor above, are different
It seems there’s really no reply. . .
Actually, the numbers calculated by SRK are correct as well; when these parameters were entered into an Excel sheet for further calculation, the accuracy was satisfactory. What’s unclear is why different methods yield different results. As for the error that occurs when using SRK with Aspen, I don’t know why either – the values are correct yet an error still appears; it’s possible this is a bug