Electricity: measuring and testing – Particle precession resonance – Determine fluid flow rate
Reexamination Certificate
1998-12-31
2001-03-20
Patidar, Jay (Department: 2862)
Electricity: measuring and testing
Particle precession resonance
Determine fluid flow rate
C324S307000, C324S300000
Reexamination Certificate
active
06204664
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to the field of measurement of polymer properties with a process NMR system, and more particularly to the use of chemometrics to find correlations for NMR data to predict the concentration of styrene in butadiene-styrene resin.
BACKGROUND OF THE INVENTION
In the chemical industry, it is important to effectuate precise prediction of properties of compounds. Several methods have been employed, including wet methods such as extraction. However, such methods have proven to be time consuming and yield low precision results. A typical method of determining properties has therefore been through instrumental analysis. In particular, the use of quantized energy states of matter through spectroscopy solves many of the efficiency problems involved in methods such as extraction.
Nuclear magnetic resonance (NMR) is a powerful spectroscopic technique for structural analysis which utilizes commonly found elements such as hydrogen and carbon as “chromophores.” With the aid of NMR, it is possible to define the environment of practically all commonly occurring functional groups, as well as fragments (e.g., hydrogen atoms attached to carbon) that are not otherwise accessible to spectroscopic or analytical techniques.
The single most important application of NMR has been in the qualitative identification of organic compounds and the elucidation of their structure. However, NMR can also be used for quantitative determination of compounds in mixtures and hence for following the progress of chemical reactions. More sophisticated applications often yield kinetic and thermodynamic parameters for certain types of chemical processes; and others, in particular spin-spin coupling, often give accurate information about the relative positions of groups of magnetic nuclei within molecules.
One means of obtaining an NMR spectrum involves the application of a strong radio frequency (RF) pulse of energy over the whole range of frequencies while the magnetic field is kept constant. As a result, nuclei are flipped to their upper state from which, over time, they will return (decay) to the lower state. Collecting the thus-induced current as a function of time through a computer creates a time-domain signal, which is a generally complex pattern called the free-induction decay (FID). Interpretation of an FID is often difficult; however, a Fourier transformation of an FID performed on the same dedicated computer yields a spectrum virtually identical to the regular absorption spectrum. This type of spectroscopy is called Fourier transform (FT) spectroscopy and is mostly applied on “high resolution” instruments with high magnetic fields (i.e. 2-14 Tesla).
NMR instruments that are used for process purposes typically have very low magnetic fields (i.e. ~0.5 Tesla). However, at such low magnetic fields, there are not enough energy differences between different types of nuclei to resolve them by FT, especially in analyzing solid samples. Therefore, the time domain signal, or FID, is the main source of information for low magnetic field instruments. However, interpretation of the FID data when using the NMR for industrial process analysis and control can be quite difficult. Prior art methods suggest solutions that involve various iterative techniques for interpreting and thereafter utilizing the FID curve to interpret properties of compounds. Such methods involve the use of large tables of data with a single equation for interpreting the FID curve or alternatively, construction of a mathematic model where the results of experiments are expressed as a mathematical function of the experimental conditions. The mathematical function method provides a means of predicting and estimating the results at levels that were not directly studied. The mathematical equation that expresses the results (e.g., solubility of inorganic salts) in terms of the experimental factors (e.g., temperature and ionic strength) is referred to as the model. The experimental results are referred to as the responses. For optimization purposes, such a model can be very crucial.
To construct the model, instrument responses from samples with known concentration levels are measured and a mathematical relationship is estimated which relates the instrument response to the concentration of the chemical component(s). This model may be used to predict the concentration of a chemical component in future samples using the measured instrument response(s) from those samples. Prior art applications such as U.S. Pat. No. 5,675,253 issued to Smith et al. (“Smith ′253”) discloses developing such a mathematical model. In particular, the patent discloses using a Marquardt-Levenberg (M-L) curve-fitting approximation technique to determine the magnitude of all the parameters that best fit the FID curves. Smith ′253 further teaches a calibration procedure which compares known samples and curve-fit points using time function equations including Gaussians, exponentials, Abragrams (defined herein as Gaussian multiplied by the quantity sin (∞t) divided by ∞t), modified exponentials (defined herein as Ce
−z
where C is a constant, z=(kt)
∝
, and ∝ lies between 0 and 1 or 1 and 2), modified Gaussians (defined herein as Gaussian multiplied by the cosine of the square root of t), and trigonometrics. However, such a curve-fitting procedure decreases the accuracy of the model and thus the accuracy of the resulting prediction.
There is a continuing need in the industry for an improved on-line system of relating multiple responses from an instrument to a property or properties of a polymer to enhance the accuracy, precision, and efficiency of prediction.
SUMMARY OF INVENTION
The present invention provides a process of using an on-line procedure to predict properties of a polymer sample using chemometric techniques.
In a preferred embodiment, extraction is used to create a first sample set of known concentrations of styrene in samples of homo- and co-polymers of butadiene-styrene (“K-Resin”). A process low-resolution nuclear magnetic resonance spectrometer (NMR) is used to obtain the free induction decay of each of said sample to create a free induction decay data set. Each data point of the free induction decay data sets are then analyzed using principle component analysis to create latent variables, thereby creating a principle component data set. The first sample set, the free induction decay data set, and the principle component data set are analyzed using partial least squares analysis to create a predictive calibration model having a predictive data set. The predictive data set is then validated using both internal cross-validation and external validation with a second smaller data set.
An unknown sample set containing compositions of homo- and co-polymers of K-Resin are placed in the NMR instrument to obtain the free induction decay of each of said compositions of K-Resin to create a second free induction decay data set. Using partial-least squares analysis, the predictive calibration model containing said predictive data set is compared with said second free induction data to predict the concentration of styrene in each of said compositions of K-Resin.
The objects, advantages, and features of the present invention will be apparent from the following description when read in conjunction with the drawings and appended claims.
REFERENCES:
patent: 5302897 (1994-04-01), Tache et al.
patent: 5596275 (1997-01-01), Dechene et al.
patent: 5675253 (1997-10-01), Smith et al.
Sharaf, et al.,Chemometrics, 1986, pp. 281-292.
Beebe, et al.,Chemometrics A Practical Guide, 1998, pp. 245-338.
Lai Xiaonian
Sardashti Maziar
Patidar Jay
Phillips Petroleum Company
Richmond, Hitchcock, Fish & Dollar
Shrivastav Brij B.
LandOfFree
Chemometric technique for predicting styrene content in... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Chemometric technique for predicting styrene content in..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Chemometric technique for predicting styrene content in... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2520433