Predictions from Several Differerent ... Log Out | Topics | Search
Moderators | Register | Edit Profile

NIR Discussion Forum » Bruce Campbell's List » Chemometrics » Predictions from Several Differerent Spectra « Previous Next »

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

Ian Michael (admin)
Board Administrator
Username: admin

Post Number: 39
Registered: 1-2006
Posted on Monday, April 23, 2012 - 9:59 am:   

The article is now published at http://www.spectroscopyeurope.com/columns/tony-davies-column/3102-examining-diffuse-reflection-and-transmission-spectra-more-thoroughly-part-2-sample-temperature-and-sample-composition
Top of pagePrevious messageNext messageBottom of page Link to this message

Ian Michael (admin)
Board Administrator
Username: admin

Post Number: 38
Registered: 1-2006
Posted on Tuesday, April 10, 2012 - 11:35 am:   

I should be putting Karl's contribution up on the Spectroscopy Europe website later this week. In the meantime, you can read his previous contribuiton (Part 1) at http://www.spectroscopyeurope.com/columns/tony-davies-column
Top of pagePrevious messageNext messageBottom of page Link to this message

venkatarman (venkynir)
Senior Member
Username: venkynir

Post Number: 155
Registered: 3-2004
Posted on Monday, April 09, 2012 - 11:02 pm:   

Hi Karl;
I have gone through your suggesstion .Why we could not do deconvolution of spectra with Discrete wavelet of Db1 or other .I found that this give good infotmation and we can play with DWT until our object is achived with minimum number of variables .
I am worried that the tool kills our thinking ability .
Top of pagePrevious messageNext messageBottom of page Link to this message

Karl Norris (knnirs)
Senior Member
Username: knnirs

Post Number: 65
Registered: 8-2009
Posted on Monday, April 09, 2012 - 5:49 pm:   

William,
Noting that you are using Unscrambler, I suggest you explore the deconvolution of your spectra with the fourth derivative. I have a paper in the last issue of Spectroscopy Europe, and by coincidence a paper will be in the April/May issue describing the use of the fourth derivative on transmission spectra of flour.
Karl
Top of pagePrevious messageNext messageBottom of page Link to this message

William Wade (williamwade)
New member
Username: williamwade

Post Number: 5
Registered: 6-2010
Posted on Monday, April 09, 2012 - 5:53 am:   

@Karl

We have a few different setups, but the one we would be using for this is reflectance. We are looking at two different machines right now for measuring the pieces. (1100-2400nm generally, but most of our information for these specific tests are in the 1500-2200nm range) Once we purchase one of them, I will be trying out, just in Unscrambler, the numbers to see what I can learn.

We do have spectra of whole grain and flour already and I have not yet tried to use those together.
Top of pagePrevious messageNext messageBottom of page Link to this message

Karl Norris (knnirs)
Senior Member
Username: knnirs

Post Number: 64
Registered: 8-2009
Posted on Friday, April 06, 2012 - 5:13 pm:   

William,
Are you sing reflection, transmission, or interactance, and what is your wavelength region?

Karl
Top of pagePrevious messageNext messageBottom of page Link to this message

William Wade (williamwade)
New member
Username: williamwade

Post Number: 3
Registered: 6-2010
Posted on Friday, April 06, 2012 - 6:31 am:   

"Is your idea is to "tease" the modelling process, by using different base solutions, to prioritize on different things?"

Basically. The idea is that we want to add another aspect into the model that does more than just one solution can give us or even the two solutions independently. It is the use of the data together that might let us know more.
Top of pagePrevious messageNext messageBottom of page Link to this message

Andrew McGlone (mcglone)
Senior Member
Username: mcglone

Post Number: 26
Registered: 2-2001
Posted on Thursday, April 05, 2012 - 1:44 pm:   

Interesting. Is your idea is to "tease" the modelling process, by using different base solutions, to prioritise on different things? Or is this a bit like using temperature variation to make models, for a constituent component say, that are independent of temperature?
Top of pagePrevious messageNext messageBottom of page Link to this message

Howard Mark (hlmark)
Senior Member
Username: hlmark

Post Number: 475
Registered: 9-2001
Posted on Thursday, April 05, 2012 - 10:37 am:   

William - NIR analysis started from work in the agricultural areas, including (or even primarily in the early days) wheat, and there was a lot of work done in the late 1970s through the 1990s, and probably is still ongoing.

You might want to do literature searches for some of the prominent agricultural scientists involved then: Franklin "Woody" Barton, David Wetzel, Karl Norris, Phil Williams, and others. Woody and Karl were with the USDA and although retired now, are still active in related areas. Dave Wetzel was and is at the Univ of Kansas in Manhattan, KS.

Most of their work was done measuring solid powders (e.g., grund wheat), so if you want to measure suspensions in liquids, you'll be )mostly) breaking new ground.

They all keep their scientific activities up, and are regular attendees at the biennial IDRC, the next one of which will be held next summer in Chambersburg, PA - see:

www.idrc-chambersburg.org for info.

Howard

\o/
/_\
Top of pagePrevious messageNext messageBottom of page Link to this message

William Wade (williamwade)
New member
Username: williamwade

Post Number: 2
Registered: 6-2010
Posted on Wednesday, March 14, 2012 - 3:23 pm:   

I was wondering a bit ago if any of the NIR systems out there have a way of giving predictions from several scans of a powder suspended in various liquids?

I know I can do it by hand in any MVA software, but what I am thinking is standard equipment software that can do it. Has anyone even done this by hand? I imagine a process like this below:

Say 20 flour samples.

Suspend a sub-sample of each of the 20 samples in a water mixture.
Scan each suspended solution via NIR inputting a unique sample ID each time, but receive no predictions.
Suspend a second sub-sample of each of the 20 samples in a water/sucrose mixture.
Scan each suspended solution via NIR (this is the second part of the data) inputting the same unique sample ID as the first time
Receive predictions for each sample based on the two spectra.

Obviously this could be done more than with just two solutions.

In our case we do some flour quality predictions using several different solutions. Each of these gives a number, but those numbers alone do not add up to a complete picture of the quality. I am wondering if there is additional predictive data in the spectra that could only be seen in conjunction with the other specta. Then if there is how to create an easy method.

Thoughts?

William Wade
Soft Wheat Quality Lab
Wooster, Ohio

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