| Author | Message | ||
     Tony Davies (td) Moderator Username: td Post Number: 222 Registered: 1-2001 |
Dear Carolina, Welcome to the group! I am going to disappoint you because I think Katherine is correct. What is the concentration range of your metals? I expect it is quite low ppm? You would not expect to get an NIR signal from metals and especially not at low concentrations. However, it is possible that the majority of the metal ions are complexed with some specific proteins and then you might find some secondary correlation with protein - possibly around 0.5! The other question is "What do you want to achieve"? IF you were wanting to screen samples for unusually high levels of metals then a correlation of 0.5 might be OK. Would you liketo give us some more detail. Instrument, pre-processing, calibration method? Best wishes, Tony | ||
| Katherine Bakeev (katherineb) New member Username: katherineb Post Number: 2 Registered: 12-2009 |
Carolina, I am not sure that the information you would like to measure is best done by NIR. Where is the signal change for the components of interest for you? What vibration is this related to? Before applying preprocessing, etc. it is important to go back to the fundamentals of the measurement. Regards, Katherine | ||
| Gustavo Figueira de Paula (gustavo) Junior Member Username: gustavo Post Number: 8 Registered: 6-2008 |
Carolina, What algorithm you used to calibrate your method? Have you performed any pre-processing? I think that with a so large database (>4,000 samples), low correlation will be likely due to bad experimental procedures or bad mathematical treatment. Probably, a better pre-treatment would improve it very much. Hugs, Gustavo. | ||
| carolina astudillo reyes (caroastu) New member Username: caroastu Post Number: 1 Registered: 2-2010 |
Hi Mike,,,,I try to develop an equation for cuantify iron and zinc in beans seed meal. There are more than four thousand samples scanned with references data. The best correlation was R=0.5. What is your opinion or what is your advice about that....Thank you very much | ||
| yangzengling (yangzengling) New member Username: yangzengling Post Number: 1 Registered: 9-2009 |
Michael C Mound (mike), I am very interested in your research about Mineral Slurries. In 2006 I did some work to detect the mineral contents in pig manure by NIRS (Zengling Yang��Lujia Han��Xia Fan. Rapidly estimating nutrient contents of fattening pig manure from floor scrapings by near infrared reflectance spectroscopy. Journal of near infrared spectroscopy��2006��14(4)��261��268). For TP, Ca, Mg, Fe, Cu and Zn, good results were obtained in that study. I am looking for the principle and LOD of NIRS detecting mineral in manure. However, I didn��t find breakthrough point. Could you give me some suggestions and your research results published? My e-mail is [email protected]. Thanks very much. Thanks again, Yang Zengling | ||
| Michael C Mound (mike) Senior Member Username: mike Post Number: 61 Registered: 7-2007 |
Hi, Tony, Of course, I realize it is not easy...they tell me that is what they pay me the big bucks for (the bucks are not that big...but it makes the bosses happy to make that old chestnut statement). On the serious side, here is the situation: Minerals identifications are not impossible, once you can get the materials in a somewhat homogeneous mass. However, slurries are quite different to work with, and a keen knowledge of fluid dynamics as well as preventing settling of the flow load and where to position the infrared source and detector are not simple at all. Particle size varies quite a bit, but most materials coming out of a SAG mill (semi-autogenous grinding) will move though cyclones and then to thickeners with a nominal particle size of about 40% < 45 or so microns. This is not dependable though, since the dynamics of the process are mostly in gross statistical measures. Further, the medium, though mostly aqueous, can be complicated by addtions of xanthates or amines, depending on where you want the analytics to take place. More than that, there is quite a difference in pH values, as well as factors affecting cohesive forces, resulting in agglomerates. Add to this the difficulty in controlling turbulence, and you have a really interesting brew. Minerals liberation is where the "rubber meets the road", so to speak, and identification is made more complicatef (I agreed this was difficult) because of the formation of ctenospheres, which are accretions of hollow mineralizations of, for example, silica or calcium locking within the desired mineral. Anyway, thanks for the suggestions. As I mentioned, the challenge is to do this in real-time for trend characterization, where one does not have the luxury of sample preparation of conditioning in the interests of time. I'll muse about this a bit more before surrendering to the laws of physics and fluid dynamics. By the way, getting the spectra resolved from the water absorption sometimes works with the right PLS or other pattern fitting programs, but the spectra don't always behave, even with powerful computer juggling of the reflections. Thanks, again, Mike | ||
| Tony Davies (td) Moderator Username: td Post Number: 199 Registered: 1-2001 |
Hello Mike, I think you will need to tell us quite a bit more: What answers do you want? Are you looking for identifications or amounts of specified substances (or both). Particle size? This will not be easy (the easy problems have be done!). On the +side NIR is used for prospecting for minerals so there are useful signals in the NIR region. The major -side is that the spectra will be dominated by water and the water spectrum will be quite variable depending on soluble material in the slurries; so you cannot easily subtract the water absorption. If you can get useful spectra, my CARNAC method might be a possible way of obtaining analytical information. Ana Garrido and her team have obtained good results for animal feeds using CARNAC. But first you have to get the spectra ... I have a gut feeling that Raman might be a better technique but this is not based on experience. Good luck, Best wishes Tony | ||
| Michael C Mound (mike) Senior Member Username: mike Post Number: 60 Registered: 7-2007 |
Recently, I have been asked as to the practicality of using NIR in characterizing the mineral content of mineral slurries. These materials are high volume, high velocity streams varying from 30-60% solids content. Separating the slurry to pass by an analyzer using probes or "thief" sample equiment would necessitate very complicated collection and transport to assure non-settling of the solids to provide representative analyses. Further, the amount of aqueous content would be another problem, considering the absorption of infrared radiation by the containing waters. Obtaining samples for constructing a model could be done, but my guess is that hundreds of such samples would be necessary, and I am unsure as to whether they would really provide the kind of model that would be robust enough. Complicating matters further is that some streams are hot brines with the possiblity of selective and unpredicable crystallization of cool zones in the streams that could result in the settling out of wanted analyte from the main flow. Also, some of these streams would be highly corrosive, making materials of construction difficult. My own experience is that mid-IR rather than NIR would allow thorough content of the inorganics, as has been used for pharmaceuticals as well as examples of remote sensing in geological assays. Thus, without introducing some means of full content recovery via drying equipment, even damp samples would be difficult to secure. Has anyone found a solution to the contrary? I know that there could be a possibiity for dried material to give some results, but I cannot see a practical resolution of the conversion of a point analyses, or even a batch hybrid of the above to meet the conditions of process analytics. Thanks for any suggestions. Mike |