Patent US5745382A successfully refreshed from Google Patents.
US5745382A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Neural network based system for equipment surveillance
Assignee: Arch Development Corp
•
Filing: 1995-08-31
•
Granted: 1998-04-28
•
Expires: 2015-08-31
Abstract
Abstract A method and system for performing surveillance of transient signals of an industrial device to ascertain the operating state. The method and system involves the steps of reading into a memory training data, determining neural network weighting values until achieving target outputs close to the neural network output. If the target outputs are inadequate, wavelet parameters are determined to yield neural network outputs close to the desired set of target outputs and then providing signals characteristic of an industrial process and comparing the neural network output to the industrial process signals to evaluate the operating state of the industrial process.
Inventors
- 👤 Richard B. Vilim
- 👤 Kenneth C. Gross
- 👤 Stephan W. Wegerich
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-29 02:26
CPC Classifications
G; G05; G05B; G05B23/00; G05B23/02; G05B23/0205; G05B23/0218; G05B23/0224; G05B23/024; G06; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/0499; G06N3/08; G06N3/09; G21; G21D; G21D3/00; G21D3/001; G06F; G06F2218/00; G06F2218/12; Y; Y02; Y02E; Y02E30/00; Y02E30/30; Y10; Y10S; Y10S706/00; Y10S706/902; Y10S706/903; Y10S706/904; Y10S706/906
Claims Analysis
Total Claims: 22
Claim 1
Independent Claim
171 words
A method of performing surveillance of transient signals of an industrial device to determine an operating state thereof, comprising the steps of: (a) reading into a memory training data; (b) determining neural network weighting values by the steps comprising: (1) solving a set of linear equations for obtaining the neural network weighting values; (2) computing a neural network output; (3) evaluating the neural network output to determine whether the output is close to a set of target outputs; (4) continuing steps (1)-(3) until achieving the target outputs; and (5) providing a neural network output; (c) providing signals characteristic of an industrial process; (d) comprising the neural network output to said industrial process signals to ascertain the operating state of the industrial process comprising the steps of: (1) hypothesizing said industrial process signals belong to a particular class; (2) at least one of translating and scaling stored neural network training signals: and (3) determining error between the neural network output for said industrial process signals and said stored neural network training signals.
Claim 2
Independent Claim
171 words
A method of performing surveillance of transient signals of an industrial device to determine an operating state thereof, comprising the steps of: (a) reading into a memory training data; (b) determining neural network weighting values by the steps comprising: (1) solving a set of linear equations for obtaining the neural network weighting values; (2) computing a neural network output; (3) evaluating the neural network output to determine whether the output is close to a set of target outputs; (4) continuing steps (1)-(3) until achieving the target outputs; and (5) providing a neural network output; (c) providing signals characteristic of an industrial process; (d) comprising the neural network output to said industrial process signals to ascertain the operating state of the industrial process comprising the steps of: (1) hypothesizing said industrial process signals belong to a particular class; (2) at least one of translating and scaling stored neural network training signals: and (3) determining error between the neural network output for said industrial process signals and said stored neural network training signals.
Claim 3
Dependent Claim
68 words
The method as defined in claim 1 further includes additional steps after said step (b) when the target outputs are not sufficiently close to the target outputs, said steps comprised of: commencing iterating on wavelet parameters; and calculating the neural network output for the wavelet parameters until achieving a desired set of the wavelet parameters such that the network output converges to the desired set of target outputs.
Claim 4
Dependent Claim
49 words
The method as defined in claim 1 wherein the network weighting values are adjusted to enable meeting the target outputs by minimizing an error equation given by: ##EQU22## where t p .sbsb.1 =target value at p th output mode when input signal 1 is presented to the neural network.
Claim 5
Dependent Claim
30 words
The method as defined in claim 1 wherein n potential classes exist and the industrial process signals are tested n times to determine the proper class for the process signals.
Claim 6
Dependent Claim
53 words
The method as defined in claim 1 where output values u p1 of the neural network are given in terms of the weighting values w kp , and coefficients thereof including a k , b k and c k as follows: ##EQU23## where i 1 (t)=value of input signal 1 at time t.
Claim 7
Dependent Claim
43 words
The method as defined in claim 5 wherein a defined best set of w kp values is found by least squares followed by applying a gradient descent method to determine the values of said a k , b k and c k .
Claim 8
Dependent Claim
15 words
The method as defined in claim 1 wherein the industrial process signals comprise biological signals.
Claim 9
Dependent Claim
14 words
The method as defined in claim 7 wherein the biological signals comprise heart signals.
Claim 10
Dependent Claim
29 words
The method as defined in claim 1 wherein the industrial process signals comprise signals having substantially constant signals either side thereof characteristic of equilibrium signals for the industrial process.
Claim 11
Dependent Claim
28 words
The method as defined in claim 9 wherein further including the step of determining onset of equilibrium signals from the industrial process and implementation of a SPRT analysis.
Claim 12
Dependent Claim
22 words
The method as defined in claim 1 wherein said step (b) comprises training on translated and scaled versions of the training data.
Claim 13
Dependent Claim
21 words
The method as defined in claim 11 wherein the training process includes setting the target outputs lower, thereby establishing a gradient.
Claim 14
Dependent Claim
69 words
The method as defined in claim 11 wherein a cartesian coordinate system is established with an original signal at point (0,1), a first axis corresponding to a translate dimension and a second axis corresponding to a scale dimension and said training data establishing a well-shaped response surface, thereby enabling evaluation of a presented signal to determine whether it is a member of the same class as the training data.
Claim 15
Dependent Claim
34 words
The method as defined in claim 13 wherein the output of the network is input to a two layer perceptron network thereby providing additional degrees of freedom for enhanced shaping of the response surface.
Claim 16
Dependent Claim
26 words
The method as defined in claim 13 wherein error between the training data and the presented signal is least for the class to which it belongs.
Claim 17
Dependent Claim
23 words
The method as defined in claim 13 wherein a genetic search method is used to find a global extremum in said response surface.
Claim 18
Independent Claim
209 words
A method of performing surveillance of transient signals of an industrial device to determine an operating state thereof, comprising the steps of: (a) reading into a memory training data; (b) determining neural network weighting values by the steps comprising: (1) solving a set of linear equations for obtaining the neural network weighting values; (2) computing a neural network output; (3) evaluating the neural network output to determine whether the output is close to a set of target outputs; (4) continuing steps (1)-(3) until achieving the target outputs; and (5) providing a neural network output; (c) providing signals characteristic of an industrial process; (d) calculating the neural network output for the wavelet parameters until achieving a desired set of the wavelet parameters which yield the neural network output close to a desired set of target outputs by performing the following steps: (1) hypothesizing said industrial process signals belong to a particular class; (2) translating and scaling stored neural network training signals; and (3) determining error between the neural network output for said industrial process signals and said stored neural network training signals. (e) providing signals characteristic of an industrial process; (f) comparing the neural network output to said industrial process signals to ascertain the operating state of the industrial process.
Claim 19
Independent Claim
209 words
A method of performing surveillance of transient signals of an industrial device to determine an operating state thereof, comprising the steps of: (a) reading into a memory training data; (b) determining neural network weighting values by the steps comprising: (1) solving a set of linear equations for obtaining the neural network weighting values; (2) computing a neural network output; (3) evaluating the neural network output to determine whether the output is close to a set of target outputs; (4) continuing steps (1)-(3) until achieving the target outputs; and (5) providing a neural network output; (c) providing signals characteristic of an industrial process; (d) calculating the neural network output for the wavelet parameters until achieving a desired set of the wavelet parameters which yield the neural network output close to a desired set of target outputs by performing the following steps: (1) hypothesizing said industrial process signals belong to a particular class; (2) translating and scaling stored neural network training signals; and (3) determining error between the neural network output for said industrial process signals and said stored neural network training signals. (e) providing signals characteristic of an industrial process; (f) comparing the neural network output to said industrial process signals to ascertain the operating state of the industrial process.
Claim 20
Dependent Claim
15 words
The method as defined in claim 17 wherein the industrial process signals comprise transient signals.
Claim 21
Dependent Claim
33 words
The method as defined in claim 17 further including the step of annunciating an alarm condition upon detecting a deviation of the industrial process signals from a desired one of the particular class.
Claim 22
Dependent Claim
38 words
The method as defined in claim 17 wherein said industrial process signals are selected from the group consisting of engine startup signals, biological pulse signals, chemical process startup signals, power plant startup signals and electronic systems transient signals.
References Cited (Backward Citations)
US5402521A
Cited
US5473532A
Cited
US5214577A
Cited
US5285457A
Cited
US5459675A
Cited
US5410492A
Cited
US5223207A
Cited
US5419197A
Cited
US5677998A
Cited
US5668741A
Cited
US5623579A
Cited
US5463768A
Cited
US5524631A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2001-04-27 | 2001-10-27 | 2002-04-28 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2005-04-27 | 2005-10-27 | 2006-04-28 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2009-04-27 | 2009-10-27 | 2010-04-28 | $7700.00 | Window Closed / Lapsed |