Composition analysis of waste solvent Log Out | Topics | Search
Moderators | Register | Edit Profile

NIR Discussion Forum » Bruce Campbell's List » I need help » Composition analysis of waste solvent « Previous Next »

Author Message
Top of pagePrevious messageNext messageBottom of page Link to this message

Michael Hodge (Mhodge)
Posted on Thursday, August 12, 2004 - 2:59 am:   

Hello,
I am relatively new to NIR spectroscopy and I am working on a project to determine the composition of five different waste solvent tanks all with varying compositions of seven solvents. Initially, I am making seperate models for each waste tank and the ultimate goal is to build a model capable of analysing waste from any waste tank. I have been using pls2 mainly for modelling and having been using MSC because of scattering caused by dissolved solids in the waste. I get good correlation, but the predictive ability of the models are bad. I can't figure out why this is. Does anyone know of any literature/published papers dealing with a similar type of problem that I could look up? I will also be establishing an on-line NIR method for analysing streams leaving and entering distillation columns for the solvent recovery process, so any information on this would also be greatly appreciated.

Thanking you with regards,
Michael Hodge
Top of pagePrevious messageNext messageBottom of page Link to this message

Gabi Levin
Posted on Thursday, August 12, 2004 - 5:26 am:   

Michael,

I don't know that I can help with such little information.

Here are my questions:

1. Are all seven solvents present all the time?
2. What is the range of concentrations for each?
3. What is the reference method?
4. is the prediction capability poor for all? or for some?
5. If for some, what is the range of concentrations at which the poor predictability occurs.
6. What chemometric package do you use?
When you say dissolved solids do you mean suspended solid particulate material?
7. How large a concentration of the particles, assuming there are such suspended solids?
8. Is it possible that the composition of the solids vary or that their concnetration varies or both?
9. Are you using flow cell or pobe, for probes - do you use a probe for transmission only, or a probe that can collect the spectra in transmission and diffuse reflectance form the particles?

Knowledge of some of these paramters may help to understand.

One more question - if you use the Unscrambler, and if you do full cross validation, you get a plot that shows the calibrtaion set and the cross validation set. Usually the calibrtaion data points are blue circle, and the cross validation points are solid red dots. If you see a situation where the red points don't overlap nicely the blue circles, you can expect poor predictability.

One more thing, you can use PLS1 also.

Thanks,

Gabi Levin

Add Your Message Here
Posting is currently disabled in this topic. Contact your discussion moderator for more information.