Method for determining parameter of hydrocarbon

• US 5,572,030 A
• Filed: 01/11/1996
• Issued: 11/05/1996
• Est. Priority Date: 04/22/1994
• Status: Expired due to Term

• Current Assignee: Intevep, S.A. (PDVSA Gas SA)
• Original Assignee: Intevep, S.A. (PDVSA Gas SA)
• Inventors: Ranson, Aaron, Arroyo, Fernando, Parisi, Adriano, Prieto, Hernan
• Primary Examiner(s): Willis, Davis L.
• Assistant Examiner(s): HANIG, RICHARD E
• Application Number:
  US08/585,000
• Time in Patent Office: 299 Days
• Field of Search: 250/339.12, 250/339.09, 250/339.11, 250/339.07, 250/341.5, 395/932, 395/911, 395/22
• US Class Current: 250/339.12
• CPC Class Codes: G01N 21/3577 for analysing liquids, e.g. polluted water
• CPC Class Codes: G01N 21/359 using near infrared light

Abstract

A method for evaluating a hydrocarbon so as to determine a desired parameter of the hydrocarbon includes the steps of providing a hydrocarbon to be evaluated, obtaining a near-infrared signal from the hydrocarbon, codifying the near-infrared signal so as to reduce the signal to a number of points, providing a neural network trained for correlating the number of points to the desired parameter and processing the number of points with the neural network so as to determine the desired parameter.

Claims

1. A method for evaluating a hydrocarbon fuel from a family of hydrocarbon fuels selected from the group consisting of gasoline, diesel fuel, kerosene, naphtha and jet fuel to determine a desired parameter selected from the group consisting of Reid vapor pressure, simulated distillation values, research octane number, motor octane number, oxygen content, specific gravity octane number, bromine number, aniline point, smoke point, and combinations thereof, comprising the steps of:

(1) providing a computer configured as a neural network;

(2) training the neural network so as to evaluate a hydrocarbon fuel from the family of hydrocarbon fuels to determine the desired parameter, said training comprising the steps of;

(a) selecting a plurality of hydrocarbon fuels from the family of hydrocarbon fuels to be evaluated;

(b) obtaining an NIR spectra for each of said plurality of hydrocarbon fuels;

(c) codifying each of the NIR spectra obtained by providing a base line correction and thereafter reducing the base line corrected spectra to a desired number of points corresponding to the parameters being evaluated;

(d) developing a first matrix from the desired number of points, said first matrix to be subsequently inputted to the neural network;

(e) obtaining a second matrix of parameter values from an analytical evaluation of the plurality of hydrocarbon fuels;

(f) processing the first matrix and the second matrix in the neural network to obtain a functional relationship between the first matrix and the second matrix so as to develop a weighted matrix; and

(g) repeating steps (b) through (f) to obtain an optimal weighted matrix.

(3) comparing an NIR spectra from a hydrocarbon fuel selected from the family of hydrocarbon fuels with the optimal weight matrix in the neural network to obtain a predicted value of the desired parameters for the selected hydrocarbon fuel.

2. A method according to claim 1, wherein the codifying step includes reducing the signal to about 47 points.

3. A method according to claim 1, wherein the parameter to be determined is Reid vapor pressure and wherein the codifying step includes reducing the signal to between about 30 to about 100 points.

4. A method according to claim 1, wherein the parameter to be determined is research octane number and wherein the codifying step includes reducing the signal to between about 30 to about 100 points.

5. A method according to claim 1, further including selecting a parameter to be determined which is non-linearly related to the near-infrared signal.

6. A method according to claim 1, wherein the near-infrared signal is obtained at a wavelength of between about 800 nm to about 1600 nm.

7. A method according to claim 1, wherein the near-infrared signal is obtained at a wavelength of between about 800 nm to about 1050 nm.

8. A method according to claim 1, wherein the near-infrared signal is obtained at a wavelength of between about 1100 nm to about 1500 nm.

9. A method according to claim 1, wherein the near-infrared signal is obtained on line.

10. A method according to claim 1, wherein the step of obtaining the near-infrared signal comprises obtaining a transmissive near-infrared spectrum from the hydrocarbon.

11. A method according to claim 1, wherein the step of obtaining the near-infrared signal comprises obtaining a reflective near-infrared spectrum from the hydrocarbon.

12. A method according to claim 1, wherein the codifying step further includes the step of base line correcting the signal prior to reducing the signal to the number of points.

13. A method according to claim 1, wherein the step of providing the neural network comprises providing a neural network having an input layer, an intermediate layer, and an output layer, and wherein the input layer has between about 30 to about 100 nodes, the intermediate layer has between about 3 to about 15 nodes, and the output layer has a number of nodes equivalent to a number of desired parameters to be determined.

14. A method according to claim 1, wherein the step of providing the neural network includes providing a neural network having layers and having transfer functions between the layers which are sigmoid functions.

15. A method according to claim 14, wherein the sigmoid functions are as follows:

##EQU3## wherein i and j are nodes of the network;

I_j is an output value of node i;

W_ji is a weighted value between i and j; and

O_j is an output value of node j.

16. A method according to claim 15, wherein the sigmoid function includes a sigmoid transfer function as follows:

##EQU4## wherein ##EQU5##