https://www.euro-online.org/enog/inoc2007/Papers/mac-slots.html https://www.euro-online.org/enog/inoc2007/Papers/m https://www.euro-online.org/enog/inoc2007/Papers/mac-slots.html

aging of instrument

InfraRed's picture
Forums: 

Dear all,

to what extent does aging of an instrument influence the measurement?

Is it necessary to consider aging in calculating a calibration or do i have to recalibrate an instrument after a certain time?

Thank you and best wishes,

Michael

hlmark's picture

Michael - I'm going to guess that by "influence the measurement" you mean "is there error induced in quantitative spectorscopic analysis (probably using chemometrics) as a result of the aging" and answer accordingly: To the extent that aging will affect the spectrum produced by the instrument, of course there will be some effect on the readings. The magnitude of the effect will depend on several factors, of which one of the more important ones is the nature of the chemometric calibration model being used. Generally, the first indication of such effects is a "bias change" as it is called: a consistent change in the readings (i.e., an error that is the same, in magnitude and direction, for every sample) regardless of the sample composition. For this reason, post-modeling actions for a new calibration model should include implementation of a quality-control procedure for the instrument readings. Instrument manufacturers differ somewhat in the details of what they specify for this QC procedure, but ususally incorporate two types of readings:

1) A daily check of the hardware using one or more samples with known aborabance properties

2) A check procedure run at somewhat longer intervals, using a set of 10-20 samples with known composition. This will detect any situation where the readings start to depart from their known values in a consistent manner. If this happens, then one common action is to use a "bias correction", whereby the magnitude of the error induced in the readings (as determined from the known samples) is simply subtracted from all future readings of the samples for all future analyses.

Howard

 

\o/

/_\

 

InfraRed's picture

Thank you very much, Howard. 

Your response helped me a lot.

Best Regards

Michael

 

hlmark's picture

Michael - no problem, what I described is standard practice in the technology. I recommend you obtain one (or more) of the several excellent books that are available so you can read up on what is considered good practice. There is also a free on-line tutorial available at www.nearinfrared.com; click on the "tutorial" link to download a copy.

\o/

/_\

 

ianm's picture

Modern NIRS instruments can be expected to work effectively for many years. We have former NIRSystems (now Foss) Model 6500 scanning spectrophotometers that have been working, and continue to pass diagnostic tests, for over 20 years, and I know of a 3-filter former Neotec (now Foss) GQA Model 31EL that worked for over 30 years. But, yes, they do age, mainly to increasing fatigue of some components, other than the lamps, which have a limited life.

There is a bit of philosophy involved with this question. There are two main approaches to the use of a NIRS instrument. The first is the feasibility study. Most of the literally hundreds of papers that have been spawned by NIRS are of this type, which lends itself to the production of such papers. For this approach the longevity of an instrument is important, but it is only necessary to ensure that the instrument is functioning properly for the duration of the project, which is usually only a few weeks. There may be a gap between projects, and when a new project is about to commence the instrument performance has to be re-verified. The instrument diagnostics are adequate for such performance verification. To estimate the instrument spectral precision a check sample should be scanned periodically during the scanning of samples for these types of project.

The second approach is the day-to-day use of the instrument as an analytical tool in an industrial application. For this approach a useful tactic is to monitor the instrument performance by means of a check sample, which should be interpolated at least 2 or 3 times during each working day. An NIRS instrument uses a calibration as a constant. The calibration itself cannot change, and any changes in the analytical output of the instrument, as detected by the results of check sample analysis, can be assigned to the instrument (assuming that operator error has been established). This becomes more complicated with on-line systems, because it is more difficult to introduce a check sample under normal conditions of operation. Instrument performance can be monitored by withdrawing samples from time to time and having the composition verified by laboratory analysis. Persistent changes in bias, even small changes, will indicate that the instrument is undergoing progressive change. This can be addressed by periodical up-dating of the calibrations. The modest expenses in carrying out such a monitoring process should be assessed in terms of the considerable savings in operating costs effected by use of NIRS.            

Posted by Ian Michael on behalf of
Phil Williams,
PDK Projects, Inc. Nanaimo, B.C. Canada

Sus's picture

with all due respect - Former FOSS NIRSystems is now Metrohm NIRSystems. Therefore, NIRSystems did not stop existing.

KR,

Sus

Central Queensland University's picture

I have similar thinking like Micheal and initiated a study with handheld NIR-Vis spectrometers. These spectrometers are extensively used in field for at least 4-6 months each year and some are more than 6 years old. We checked the performance of each machine taking spectra of PTEF-whilte tile (same tile) each yere in terms of spectral absorbance and SD. Further, we choose 20-25 samples and spectra are taken with all machine and develop PLS model on one attribute of interest and tally the performace of machine over several years. The result so far are some machines are fairly stable, while few tend to break-down and need software and hardware maintenance. The approach we follow here is kind of QC every spectrometer has to pass before sending to field operature.

Umesh