Method and electronic system for detecting rail switch degradation and failures

Patent metadata

Field	Value
Publication number	US-11186304-B2
Application number	US-201816027630-A
Country	US
Kind code	B2
Filing date	Jul 5, 2018
Priority date	Jul 5, 2018
Publication date	Nov 30, 2021
Grant date	Nov 30, 2021

Abstract

Method for detecting rail switches degradation and failures, the method having the steps of applying a Discrete Wavelet Transform to stored measurement data relative to a switch move to be analysed and obtaining a feature vector associated with said switch move and including Discrete Wavelet Transform coefficients delivered by the applied Discrete Wavelet Transform; comparing the obtained feature vector with feature vectors associated with rail switch moves previously obtained and associating each to a respective health class among several given health classes; and determining a health class for the said switch move to be analyzed by selecting one of the health classes associated with the previously obtained feature vectors based upon the comparison step.

First claim

The invention claimed is:

1. Method for detecting rail switches degradation and failures, the method comprising the following steps:
step (i): collecting and storing, in a database, rail switch measurement data measured during the rail switch moves to be analyzed;
step (ii): applying a Discrete Wavelet Transform on the stored measurement data relative to one of the switch moves to be analyzed and obtaining a feature vector associated with said switch move and including Discrete Wavelet Transform coefficients delivered by the applied Discrete Wavelet Transform;
comparing the obtained feature vector with each of a plurality of feature vectors associated to rail switch moves previously obtained by steps (i) and (ii) and associating each to a respective health class among n given health classes, n being greater than or equal to 2;
selecting k previously obtained feature vectors among the plurality of previously obtained feature vectors based upon the comparison step and,
determining a health class for the said switch move to be analyzed by selecting, based on a majority vote, one of the health classes associated with k selected feature vectors, k being greater than n.

2. Method for detecting rail switches degradation and failures according to claim 1 , further comprising a step of triggering a maintenance action relative to the rail switch having performed said switch move, depending on the determined health class.

3. Method for detecting rail switches degradation and failures according to claim 1 , wherein the comparison and health determination steps includes the following steps:
determining the Euclidean distances between the obtained feature vector and the previously obtained feature vectors associated with a respective health class;
comparing the Euclidean distances between the obtained feature vector and the previously obtained feature vectors;
determining the k smallest Euclidean distances;
selecting the health class for the said switch move to be analyzed as one of the health classes most associated with the previously obtained feature vectors corresponding to the determined k smallest Euclidean distances.

4. Method for detecting rail switches degradation and failures according to claim 1 , wherein the measurement data relative to a switch move comprises samples and the step (ii) further comprising the following steps of normalizing the measurement data relative to a switch move:
calculating the mean of the samples of the measurement data relative to the switch move and subtracting the calculated mean of each sample relative to the switch move; and
calculating the standard deviation of the samples of the measurement data relative to the switch move and dividing each sample relative to the switch move with the standard deviation; and the Discrete Wavelet Transform is applied to the normalized measurement data.

5. Method for detecting rail switches degradation and failures according to claim 1 , wherein the step (ii) further comprises:
performing time filtering of the measurement data relative to the switch move based on a sliding window of duration T w ,
wherein the Discrete Wavelet Transform is applied to the measurement data relative to a move time filtered by the sliding window.

6. Electronic system for detecting rail switches degradation and failures, the system comprising:
a data acquisition unit configured for collecting and storing, in a database, rail switch measurement data measured during rail switch moves to be analysed;
a first module configured for applying a Discrete Wavelet Transform to the stored measurement data relative to one of the switch moves to be analysed and for obtaining a feature vector associated to said switch move and including Discrete Wavelet Transform coefficients delivered by the applied Discrete Wavelet Transform;
a second module configured for comparing the obtained feature vector with each of a plurality of feature vectors associated to rail switch moves previously obtained by the first module and associating each to a respective health class among n given health classes; wherein the second module is configured for selecting k previously obtained feature vectors among the plurality of previously obtained feature vectors based upon the said comparison and determining a health class for the said switch move to be analyzed by selecting, based on a majority vote, one of the health classes associated with k selected feature vectors; wherein n is greater than or equal to 2 and k is greater than n.

7. The electronic system for detecting rail switches degradation and failures according to claim 6 , configured for triggering a maintenance action relative to the rail switch having performed said switch move, depending on the determined health class.

8. The electronic system for detecting rail switches degradation and failures according to claim 6 , wherein the second module is configured for determining the Euclidean distances between the obtained feature vector and the previously obtained feature vectors associated with each respective health class; for comparing the Euclidean distances between the obtained feature vector and the previously obtained feature vectors; for determining the k smallest Euclidean distances; and for selecting the health class for the said switch move to be analyzed as one of the health classes mostly associated to the previously obtained feature vectors corresponding to the determined k smallest Euclidean distances.

9. The electronic system for detecting rail switches degradation and failures according to claim 6 , wherein the measurement data relative to a switch move comprises samples and the first module is configured for normalizing the measurement data relative to a switch move by calculating the mean of the samples of the measurement data relative to the switch move and by subtracting the calculated mean of each sample relative to the switch move; and by calculating the standard deviation of the samples of the measurement data relative to the switch move and by dividing each sample relative to the switch move with the standard deviation;
wherein the first module is further configured for applying the Discrete Wavelet Transform to the normalized measurement data.

10. The electronic system for detecting rail switches degradation and failures according to claim 6 , wherein the first module is further configured for performing time filtering of the measurement data relative to a move based on a sliding window of duration T w and for applying the Discrete Wavelet Transform to the measurement data relative to a move time filtered by the sliding window.

Assignees

• Alstom Transp Tech
• Florida Institute Of Tech

Inventors

• Mijatovic Nenad
• Peter Adrian Manoj
• Bryan Kaylen Jamal
• Tian Charlie Jinshaw
• Coley Christina Mariah

Classifications

• B61L27/53 Primary

  for trackside elements or systems, e.g. trackside supervision of trackside control system conditions · CPC title

• B61L23/04

  for monitoring the mechanical state of the route · CPC title

• E01B7/00 Primary

  Switches; Crossings (operating mechanisms B61L) · CPC title

• G01R31/343

  in operation · CPC title

• G01R31/2846

  using hard- or software simulation or using knowledge-based systems, e.g. expert systems, artificial intelligence or interactive algorithms · CPC title