Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Chemical analysis
Reexamination Certificate
1999-07-26
2002-07-16
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Chemical analysis
C702S030000, C702S023000, C356S073000
Reexamination Certificate
active
06421614
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a novel and useful system for obtaining reliable data from a sample under variable environmental conditions.
Analytical measurements are typically obtained with respect to on-line chemical processes to maintain proper control of the same. Such measurements are also extremely important to ascertain the reliability of such measurements. Multivariate models are routinely used to calibrate on-line analyzers, particularly near infrared (NIR) process spectrometers, that acquire a spectrum of contiguous wavelength. If a chemical sample falls outside the range of conditions of a multivariate model or when an unmodeled interference enters into the system, analytical measurements become biased. In certain cases such biased measurements may constitute an error, commonly referred to as an “outlier”.
All analytical instruments and methods are classified according to the type of data provided. Instruments that provide single data points, such as pH meters or single wavelength filter photometers are classified as zero-order instruments. A single value provided by these instruments is considered a zero-order tensor. Generally, it is not possible to detect the occurrence of an interference on a zero-order instrument. On the other hand, first order instruments, which obtain multiple input data points can detect occurrences of an interference or “upset”. Such first order instruments include spectrometers, chromatographs and arrays of zero-order sensors. It should be noted that process NIR spectrometers employ entire or partial continuous spectra rather than single analytical wavelengths. Such instruments, although successful in obtaining data are elaborate and expensive when compared to photometers which analyze a set number of discrete or narrow width wavelengths. “Upsets” have been detected in continuous spectra spectrometers through a variety of descriptive statistics that can be calculated from a continuous spectrum. Such statistics provide the user of the instrument with a measure of confidence in the result.
Two important descriptive statistics employed with full or continuous spectrum instruments are the sum of squares, Q, and Mahalanobis distance, M. Generally, low values for Q or M indicates a good analytical result. To indicate when there are problems in the analysis, it is necessary to determine limits that indicate when a spectrum no longer fits the chemometric model, e.g. an outlier.
U.S. Pat. No. 5,715,058 shows a method and device for the optical determination of a physical quantity by deriving a two phase-shifted signal from a common optical channel.
U.S. Pat. No. 5,532,823 teaches a method of measuring optical characteristics of a liquid crystal cell using polarized light.
U.S. Pat. No. 5,706,092 teaches a differential spectrometry system which detects very narrow band spectral features by the use of optical interference filters. The detector outputs are differenced by an operational amplifier to cancel detector signals resulting from spectral features, to both detectors.
U.S. Pat. No. 5,606,164 shows a method and apparatus for measuring the concentration of a fluid in a chemical process by the application NIR to a flow cell. Light may be directed through a chopper wheel to make both light and dark measurements. In addition, modulated light could be passed through a filter wheel prior to transmission to through the flow cell. In addition, light may be directed from a plurality of narrow band width NIR sources which is a plurality of diodes. The data obtained may be processed to determine outliers using the Mahalanobis distance or Robust distance. Calibration data is obtained to compensate for spectral artifacts.
U.S. Pat. No. 5,592,402 describes a method for interpreting complex data from detecting instrumentation such as a chromatogram or a spectrum. The determination of any outlier data includes the first step of calculating the average calibration residual spectrum through an analysis technique the residual spectrum is characterized by a single value such that data may be classified as an outlier.
A system which is capable of producing outlier determination of data from a fixed wavelength filter photometer would be a remarkable advance in the field of chemical analyses.
SUMMARY OF THE INVENTION
In accordance with the present invention a novel and useful system for obtaining reliable chemometric data from sample is herein provided.
The system of the present invention utilizes a photometer, including a fixed filter type, a moving filter type, a dispersive type, as those possessing a prism or grating, and the like. In any case, a photometer employing a plurality of independent analytical narrow band wavelengths to analyze a sample, such as one which would be detected in an on-line chemical process, may be used herein. Reference is made to U.S. Pat. No. 5,825,478 which shows a multifunctional fixed filter photometer apparatus which would be suitable in the system of the present invention. It has been found that such a fixed wavelength filter photometer performs on-line analyses which are considered to be a difficult task for analytical instruments. The particular photometer shown in U.S. Pat. No. 5,825,478 uses a plurality of independent analytical wavelengths derived from a broad band source of light by the use of narrow bandwidth interference filters. Thus, this instrument is considered to be a first order instrument and may be used to obtain data to ascertain a sample characteristics, such as absorbance, fluorescence, turbidity, optical density, and others. It is well known that absorbance is directly related to concentration and may be employed to that end. Thus, through at least first and second wavelengths, each possessing a narrow bandwidth, photometric data may be obtained from a chemical sample.
Also, in the present invention, means is provided for determining the reliability of the data obtained by the fixed filter photometer. Such means may take the form of a utilizing such fixed filter photometer or utilizing other chemical analyses to produce a training set of data representing a sample characteristic, such as color, specific chemical component, and the like, by using multiple wavelengths to analyze the sample. Also, certain sample characteristics, such as temperature, pressure, contaminant concentration, and the like are varied. At this point, a calibration set is established in which a multivariate method such as principal components analysis (PCA), principal components regression (PCR), partial least squares (PLS), or multiple linear regression (MLR) is employed. The variation in the data is explained by factors or principal components (PC). Mathematically, these techniques rely on the eigenvector decomposition of the covariance or correlation matrix of the data set. The eigenvector may be calculated mathematically or through PCA on the mean centered data set. Essentially, the eigenvector is comprised variables measured on a data set which can be made to lie in a plane.
The identification of outliers is obtained by calculating either the sum of squares (Q) or the Mahalanobis distance (M) on the training set of data. Other types of analyses may be used to obtain outliers, such as F-test, F-ratio, and the like. Either of these types of statistics provide a measure of confidence in subsequent measurements by the fixed filter photometer.
Finally, upper confidence limits are established for both Q and M in order to obtain a number which represents the determination as to whether a particular piece of data is reliable or not reliable, e.g. an outlier is found permitting a geometric interpretation of Q and M, illustrated to visualize the same as a measure of distance from the multivariate mean, which is the intersection of the principal components of the data set. Q may be illustrated as a measure of distance off the plane of an ellipse encompassing the PC data, while M may be visualized as a point within the plane of such ellipse but outside the confines of such ellipse. Thus, the value of Q or M may be compa
Goldman Donald S.
Lytle Nelson Wayne
Bielen, Jr. Theodore J.
Hoff Marc S.
Kim Paul
LandOfFree
Photometer system for obtaining reliable data does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Photometer system for obtaining reliable data, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Photometer system for obtaining reliable data will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2821844