Intangible sensor and method for making same

• US 5,373,452 A
• Filed: 04/15/1991
• Issued: 12/13/1994
• Est. Priority Date: 09/02/1988
• Status: Expired due to Term

• Current Assignee: Scott Paper Co. (Kimberly-Clark Corporation)
• Original Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Inventors: Guha, Aloke
• Primary Examiner(s): ZANELLI, MICHAEL J
• Application Number:
  US07/683,505
• Time in Patent Office: 1,338 Days
• Field of Search: 395/22, 364/550, 364/552, 364/556, 364/551.01, 340/514, 340/515-517, 371/15, 371/17, 371/24, 371/25, 73/866.4, 73/432.1, 73/866, 73/865.7
• US Class Current: 702/33
• CPC Class Codes: G01N 33/0031 comprising two or more sensors, e.g. a sensor array electrochemical electrode arrays G01N27/27
• CPC Class Codes: G01N 33/14 Beverages

Abstract

An intangible sensor for measuring intangible properties of a substance and a method for making the sensor is described. The intangible sensor may be embodied in a mapping neural network model. The intangible sensor herein is a device that quantitatively measures complex intangible properties of a sample of a substance. The term intangible implies a subjective connotation such as in the taste, creaminess or softness of a substance or product and therefore can only be subjectively defined. Although an intangible property is known to be a function of certain measurable physical properties of a substance, there are no known definitions of this function. The intangible sensor herein can implement this function simply without having any detailed knowledge of or making any analysis of the function.

Claims

1. A method for making an intangible sensor for measuring an intangible property of a particular substance, said method comprising the steps of:

(A) selecting a grading system for a human subjective evaluation of said intangible property which comprises a range of discrete index values;

(B) selecting at least two directly measurable tangible physical properties of said substance which affect said intangible property and selecting respective tangible sensors for measuring said physical properties;

(C)(1) establishing a relationship between said physical properties and one of said index values by selecting a sample of said substance and making relative thereto (1) physical measurements of said physical properties, and (2) subjective human evaluation of said intangible property comprising human selection of one of said index values corresponding to human sensation of said intangible property;

(C)(2) repeating step (C)(1) until a set of samples has been processed pursuant thereto which has said range of index values for said intangible property as defined in step (A);

(D) utilizing a mapping neural network model in a neural network having the capability to map a relationship between said measurable physical properties and said index values wherein said neural network adjusts internal weights such that inputs to said neural network which correspond to said tangible physical properties produce outputs from said neural network which correspond to said discrete index values in the same relation as exists in the set of samples processed in steps (C)(1) and (C)(2) above;

(E) mapping in said neural network for each of said samples via a supervised learning algorithm said physical measurements to a corresponding related one of said index values to thereby teach said mapping neural network model the relationship for said substance between said measurable physical properties thereof and said index values of said intangible property thereof.

2. The method according to claim 1 wherein said model is multilayered.

3. The method according to claim 2 wherein in order to realize a general function between real inputs and outputs said model has at least two layers including a hidden layer in addition to an input layer.

4. The method according to claim 1 wherein said model is a back propagation type.

5. A method for making an intangible sensor for measuring at least two intangible properties of a particular substance, said method comprising the steps of:

(A) selecting a corresponding number of grading systems for a human subjective evaluation of said intangible properties each of which comprises a range of discrete index values;

(B) selecting at least two directly measurable tangible physical properties of said substance which affect each of said intangible properties and selecting respective tangible sensors for measuring said physical properties;

(C)(1) establishing a relationship between said physical properties and one of said index values by selecting a sample of said substance and making relative thereto (1) physical measurements of said physical properties, and (2) subjective human evaluation of said intangible properties comprising human selection of one of said index values corresponding to human sensation of said intangible properties;

(C)(2) repeating step (C)(1) until a set of samples has been processed pursuant thereto which has said range of index values for said intangible properties as defined in step (A);

(D) utilizing a mapping neural network model in a neural network having the capability to map a relationship between said measurable physical properties and said index values wherein said neural network adjusts internal weights such that inputs to said neural network which correspond to said tangible physical properties produce outputs from said neural network which correspond to said discrete index values in substantially the same relation as exists in the set of samples processed in steps (C)(1) and (C)(2) above;

(E) respectively mapping in said neural network for each of said samples via a supervised learning algorithm said groups of physical measurements to corresponding related ones of said index values to thereby teach said mapping neural network model the relationship for said substance between said measurable physical properties thereof and said index values of the corresponding one of said intangible properties thereof.

6. The method according to claim 5 wherein said model is multi-layered.

7. The method according to claim 6 wherein in order to realize a general function between real inputs and outputs said model has at least two layers including a hidden layer in addition to an input layer.

8. The method according to claim 5 wherein said model is a back propagation type.

9. A sensor for identifying discrete values of an intangible property of a substance comprising:

a set of parallel input sensors for respectively measuring tangible physical properties of said substance,

an output means for indicating degree of an intangible quality corresponding to a range of said discrete values chosen by subjective human evaluation of a sample set of said substance, and

neural network means between said set of input sensors and said output means trained on subjective relationships between said discrete values and said measurements of said set of input sensors for determining said discrete values based on said measurements of said set of input sensors.