i worked on a model to describe a desalination reverse osmosis process. I have all the equations needed and the model is ready. the model depends on five parameters.
how many experiments do I need to validate the model ? if one experiment gives the same results as the model, can I say that the model is validated ? should it be on lab scale or on a pilot scale ?
thank you.
Dear Abdeljalil Chougradi you can check links below:
Article A nonlinear model of a reverse osmosis desalination system: ...
http://www.msc-les.org/proceedings/emss/2010/EMSS2010_135.pdf
https://www.mdpi.com/2227-9717/7/5/271/pdf
Regards
Hello Abdeljalil,
How does your model look like? Is this an empirical (statistically driven) model or a mechanistic model? In the latter case, it is common practice to collect some data under various process conditions (e.g. study the influence of different parameters on the process performance).
If your model contain one or more parameters, you need to calibrate the model to estimate the value(s) parameters(s). In this case it can be of interest to split up the data set in two parts. The first part can then be used for calibration while the second (independent) dataset serves for validation of the model.
For the evaluation of the goodness-of-fit, you can rely on different approaches, such as the calculation of the Theill’s inequality coefficient (TIC) or the akaike information criteria (AIC).
You should validate the model at several operarational conditions.
You need to adjust the parameters to one condition, and predict with the same parameters to the other conditions. You must calculate the determination coefficient, comparing the experimental data with the model prediction. Values higher than 0.9 will validate the model.
Best regards
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling