Hello everyone. I am recently working NL PSO fitting of my data in Originlab.
NLPSO is the user-defined equation: qt = qe^2*k2*t/1+qt*k2*t
I'm quite new to the adsorption study, therefore, I only use the equation wiz initial parameters of qe=1 and k2 =1, the boundary of qe>0 and k2>0.
i know my experimental data look poor, cuz I have only three time intervals of 2, 24 and 168 hours.
Time (min) 0,0,120,120,1440,1440,10080, 10080
qe (mg/g) 0,0, 6.376 6.556 6.558 6.519 6.544 6.250
well, R2 is 0,99 and everything looks fantastic, except for the results of K2.
K2 is 4.15E33 with SE of 6,16E44.
I don't think it is right...but Idk how and where to debug.
I do appreciate it if someone could give even a small help\!!!!!
thanks
Hi,
I haven't used OriginLab before, but I have experience in modelling adsorption kinetics and used the PSO equation in a past paper. I used MATLAB but it should make little difference. Not sure if parentheses around the denominator in your equation would help (if modelling looks ok that should be fine)? What I can see is you are missing most of where the actual curve is occurring in your adsorption, looks as though adsorption of your solute is basically done after 1 hour. This might have some impact on your study as the rate might appear almost instantaneous in the curve-fitting algorithm (again, I'm of limited experience there, as it has been a while since I've worked on that). A quick calc gives me k=188.7 but with huge error bounds, I think you need more data points to help boost your fitting, otherwise you are likely giving your fitting algorithm too much room for estimation (and fairly so at just 4 time points, including origin). If you look at my 2022 paper on copper adsorption, we ended up using 6 points not including origin, but where they were in all the right locations on the curve our model turned out looking fairly good.
David
An exciting question, but it is not about origin. If you look at your "curve", it consists of 2 areas. A jump from 0 to 6.46717 (=qe) in the first time interval and a constant course of 6.46717 (=qe) over the rest of the time. This means that k2 cannot be determined and can take on more or less any (very high) value and you will always get the same R2. qe is not influenced by time. k2 is influenced by time, but there is only the constant final value of qe. If k2 is very large, the equation can be simplified to qt = qe. So k2 = 100000000000 is the same that k2 = 100000000000000000 or k2 = 1234567891011
This is what you see.
To get k2 you need to measure a "curve", with 0 < qt < qe