Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Chemical analysis
Reexamination Certificate
2002-06-17
2003-12-30
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Chemical analysis
C702S027000, C702S032000, C702S069000
Reexamination Certificate
active
06671629
ABSTRACT:
FIELD OF AND BACKGROUND OF THE INVENTION
The invention relates to a method for measuring the characteristics of a sample by spectral analysis. The spectral data thus obtained are used to calculate the characteristics by way of a calibration model that has been established on the basis of reference samples.
The invention further relates to a device for measuring the characteristics of a sample using a spectrometer for spectral analysis of the sample and a calibration model that has been established on the basis of reference samples and that calculates the characteristics from the spectral data delivered by the spectrometer.
One example of spectral analysis of samples is NIR spectroscopy, in which the molecular structure in the sample material is excited by photons in the near infrared range (NIR). The molecules reach vibration states corresponding to their structure and exhibit typical energy absorptions. In the resulting spectrum, the energy absorption values are recorded relative to discrete wavelengths by means of a detector. The characteristics of the sample to be determined are then calculated from the spectral data thus obtained using a calibration model.
These characteristics can in principle comprise all the sample parameters that correlate with the information content of the spectrum. Thus, the characteristics include, in particular, the molecular structure of the sample and the physical and chemical properties resulting therefrom. The calibration model is established by chemometric methods, such as MLR (Multiple Linear Regression) or PLS (Partial Least Squares) on the basis of the spectral data and the characteristics of selected or prepared reference samples. The characteristics of the reference samples are already known and/or are determined by reference analysis, for example in the laboratory. The reference samples must correspond as far as possible with the respective samples to be analyzed and cover, representatively, the range in variations of the characteristics of the sample to be determined.
In practice, it is difficult to ensure that the measurements of certain characteristics of samples performed by means of the calibration model, once established, remain accurate as long as possible. Over time, there may be changes in the measuring device or in the sample composition that go unnoticed. Furthermore, there may always be influences that failed to be taken into account when the calibration model was established, e.g., due to the selection of the reference samples or due to the external circumstances of calibration. This can lead to increased measuring errors. The calibration model must therefore be checked within the context of control measurements at certain time intervals or by means of random calibration checks whenever the external conditions change. If the measurement deviation is too large, the existing calibration model must be recalibrated or corrected.
OBJECTS OF THE INVENTION
An object of the invention is automatically to detect with the least possible effort any measurement deviations even before a check measurement, and to identify the sample affected.
SUMMARY OF THE INVENTION
This and other objects are attained by the invention in its various formulations. According to one formulation, the invention is directed to a method for measuring characteristics of a sample by spectral analysis, which includes: calculating the characteristics from spectral data obtained from the spectral analysis using a calibration model that has been established on the basis of reference samples; performing an additional calculation of the characteristics of the sample based on the same spectral data using at least one additional calibration model that has been established on the basis of additional reference samples; and determining and outputting deviations between the characteristics calculated by the respective calibration models, to permit an evaluation of the quality of the measurement.
According to another formulation, the invention is directed to a device for measuring characteristics of a sample, which includes: a spectrometer for spectral analysis of the sample; a calibration model that has been established on the basis of reference samples and that calculates the characteristics from spectral data supplied by the spectrometer; at least one additional calibration model that has been established on the basis of additional reference samples and that performs an additional calculation of the characteristics; and a comparator arranged downstream from the calibration models, which determines deviations between the characteristics calculated by the respective calibration models.
As such, in the above-summarized method, at least one additional calibration model is established on the basis of additional reference samples. This additional calibration model is then used for an additional calculation of the characteristics of the sample. Any deviations between the characteristics calculated by the respective calibration models are preferably determined and output, e.g., for further processing.
Analogously, the device summarized above thus includes at least one additional calibration model which is established on the basis of additional reference samples and which performs an additional calculation of the characteristics. A comparator arranged downstream from the calibration models determines and outputs the deviations between the characteristics calculated by the respective calibration models.
Double or multiple calculation of the characteristics using independent calibration models and the determination of the deviations between the calculated characteristics improve the reliability of the measurements and their robustness against the influences of unnoticed errors. Since the established deviations are available together with the calculated characteristics, the user is able to evaluate the quality of the measurement for each sample. In particular, by monitoring whether the determined deviations exceed a predefined threshold, it becomes possible to determine when the measuring error becomes too large. The measuring device according to the invention preferably can do this automatically during routine operation, so that it can also be used in otherwise unmonitored on-line operation. The need for regular check measurements is thus eliminated, and recalibration is necessary only if there is an automatically detected and reported exceeding of the threshold.
The additional calibration and computation effort required due to the at least one additional calibration model is not significant and corresponds to the number and selection of the samples used for calibration. The possible objection that by combining the reference samples used for the independent calibration models one could establish a single, more comprehensive calibration model and achieve a comparable improvement in the measuring behavior is accurate only in especially favorable cases, since the distribution of the reference samples must be carefully selected from the standpoint of uniform coverage of the measuring range. Furthermore, the possibility afforded by the invention of detecting simply and automatically any change relative to the calibration conditions would be lost.
There are different options to select the reference samples for establishing the calibration model and the additional reference samples for establishing the additional calibration model. It is useful if the reference samples and the additional reference samples, respectively, cover unequally large variation ranges of the characteristics to be determined in the samples. For instance, the one calibration model can be configured to cover a relatively large variation range with naturally selected reference samples, while the additional calibration model is established over a relatively narrow variation range using specially prepared samples. The one calibration model then calculates the characteristics in the larger variation range with relatively low resolution, while the additional calibration model calculates the characteristics in the narrower r
Hoff Marc S.
Siemens Aktiengesellschaft
Suarez Felix
Sughrue & Mion, PLLC
LandOfFree
Method and device for measuring characteristics of a sample does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and device for measuring characteristics of a sample, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and device for measuring characteristics of a sample will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3104450