| Author | Message | ||
     alun evans (Alun_Evans) |
We are currently reviewing the software we use with our NIR hardware and I am interested in what the general feeling is regarding available software. I am led to believe that GRAMS/AI is in use pretty much everywhere. We currently use the NIR for monitoring gasoline production and gasoline feedstock streams. The eventual aim is to use NIR to control octanes during blending of gasoline. | ||
| Kari Aaljoki |
Depending on the target system I'm using Unscrambler or Pirouette for PLS modeling. I do the data collection and preprocessing with the software developed myself. Unscrambler (V8.0) is pretty stable and easy to use. Pirouette has some nice features, but the interface is a bit old fashioned and it can crash in certain circumstances. Saving your project frequently is the easiest way to avoid problems. | ||
| hlmark |
Alun - aside from the instrument manufacturer's software packages, there are three third-party suppliers of software that are commonly used to analyze NIR data. One is GRAMS-AI from Thermo/Galactic Industries, as you already know about. One is The Unscrambler, from CAMO (www.camo.com). One is Pirouette, from Infometrix, Inc. (www.infometrix.com) Historically they started from conpletely different approaches and philosophies about how to analyze data, but over the years there has been considerable convergence, and to some extent they can do several of the same things. Nevertheless there still exist significant differences among them, and you should check out all three before deciding which one is best for you. Your question comes exactly one day too late, since the exhibition at Pittcon closed yesterday, and had you been there you could have spoken with representatives of all three companies (Infometrix didn't have a booth, but the founder and president was there and would have been pleased to tell you about their package) Howard \o/ /_\ | ||
| Tony Davies (Td) |
Alun, I think that this is going to be repetitive. I am an Unscrambler fan! Harald Martens gave me instruction on operating version 1. (Quite a few years ago!) I am interested that you are asking the question as I think this is the way instruments should be used. I do not think that instrument manufacturers should be trying to write general purpose software. They need to concentrate on getting good data from their instruments (which probably includes instrument standardisation) but the rest should be left to specialist software companies. I think you have to try out the top three as identified by Howard and choose the one you feel most comfortable with. Best wishes, Tony | ||
| MPDC |
Unscrambler: very well thought out user interface, can be learned by lab-technicians in 2-3 days. Very bad at importing - exporting data and "playing nice" with other programs. They have a run-time predictor module which makes things a little easier, but it is strictly Windows-only. It is for example impossible to save spectra directly in the Unscrabler format, unless you cut a deal with Camo directly. Pirouette: long time since I used it, I think it is only popular in the US. The crashing reported by Kari seems to indicate not much changed since I used it. GRAMS is more complex to use, but is a lot more programmers-friendly, in case you wish to adapt/write software for your application. Don't forget MATLAB, which has probably the most complete toolkit, but is more of a programming language then a stand-alone program. This one takes some serious training, plus good understanding of the matrix operations behind multivariate statistics. Umetrics also has a package, and they are specialised in batch-operations. You can follow a batch and "stretch" your X axis (time) if e.g. your batch runs more slowly. Definitely worth a look if your proces is batch-based, like fermentation. Probably less interesting for your application (oil). Last but not least some manufacturors bundle their own software. For example Gabi (active on this forum) from Brimrose has a pretty neat Locally Weighed Regression package, but it will obviously only work on Brimrose hardware. | ||
| hlmark |
Alun - yes, as MPDC indicates, MATLAB can be (and is) used for analysis of NIR data. I didn't include it before because, as MPDC said, it's not a data analyis package as such. However, there is some assistance in the form of two chemometric data analysis packages that are available that will run under MATLAB: the Chemometrics Toolbox written by Rich Kramer (www.chemometrics.com) and the PLS toolbox written by Barry Wise (www.eigenvector.com). As MPDC indicates, you will in any case have a moderately steep learning curve if you're a novice at MATLAB. In exchange for that, it offers the most power and flexibility in adapting the available software to your particular needs. If you want to pursue this approach, feel free to contact me off the discussion group. Howard \o/ /_\ | ||
| Gabi Levin |
Hi, While a fan of Unscrambler myself, because of the many practical aspects, I think the package by Umetrics shall not be overlooked, while it maybe less user friendly in some practical aspects, and contains mabe many features for statiticians, it has a strong tool called orthogonal signal correction, that many time improves the prediction capability of models. The key thing to using this featues in on line situations is having the spectrun collection and treatment software compatible with the Umetrics package, and have the "hooks" to use the model in real time. I don't recall their web site of hand, but any google search will find Umetrics. Gabi Levin Brimrose | ||
| hlmark |
Gabi - the usual direct approach works, it's: http://www.umetrics.com Howard \o/ /_\ | ||
| alun evans (Alun_Evans) |
Thanks everyone for getting back to me on this topic. One thing I forgot to mention in my initial message was that we are currently using LT Industry software from the early 90's which is still DOS based so any move towards windows would be a good move!! Our main aim once we have developed good models will be to control the blending of gasolines using the field NIR and which ever software we purchase. Thanks again Alun |