# US 11,079,423 B2 · System and method for detecting an electric arc

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,079,423 B2
- Document ID: US11079423B2
- Status: Granted
- Kind: B2
- Application number: 15/484,291
- Filed: 2017-04-11
- Published: 2017-10-19
- Granted: 2021-08-03
- Earliest claimed: 2016-04-14
- Assignees: Commissariat A L'energie Atomique Et Aux Energies Alternatives
- Applicants: Commissariat A L'energie Atomique Et Aux Energies Alternatives
- Inventors: Ibrahima Djigo; Mikael Carmona; Vincent Heiries
- Web page: https://patentgrep.com/patent/US11079423B2

## CPC

- G01R 31/1209: {using acoustic measurements (acoustic measurements G01H3/00)}
- G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12)

## Abstract

The invention concerns a method and a system for detecting an electric arc in an electrical installation (7) equipped with at least one sensor (5) capable of detecting temporal acoustic signals produced by the electrical installation (7), said method being characterised in that it comprises a device for analysis (3) configured to analyse any temporal acoustic signal (5) coming from said one or more sensors (5) by associating with said acoustic signal a corresponding parsimony measurement (m) whose value indicates whether the acoustic signal is an electrical arc signature or not.

## Claims

### Claim 1

A system for detecting an electric arc in an electrical installation equipped with at least one sensor configured to detect temporal acoustic signals produced by the electrical installation, said system comprising:
an analysis device configured to
analyze a temporal acoustic signal (S) coming from said at least one sensor by associating with said temporal acoustic signal a corresponding parsimony measurement (m), a value of the parsimony measurement indicating whether or not the temporal acoustic signal is a signature of an electric arc,
determine said parsimony measurement by sampling said temporal acoustic signal to form a sampled signal f, compressing the sampled signal to form a compressed signal and reconstructing the sampled signal from said compressed signal, and
trigger an alarm when the value of the parsimony measurement indicates that the temporal signal is a signature of an electric arc,
wherein the compressed signal is generated by iteratively multiplying the sampled signal at each current iteration with an acquisition matrix which differs at each iteration.

### Claim 2 (depends on 1)

The detection system according to claim 1, wherein the analysis device comprises:
a pre-conditioning circuit configured to transform said temporal acoustic signal (S) into the sampled signal f, the sampled signal f of predetermined dimension N and which is expressed in a temporal base,
a processor configured to
determine the parsimony measurement of said temporal acoustic signal by calculating a parsimony probability of the sampled signal f, and
detect whether said parsimony probability exceeds a predetermined threshold value, said exceeding of the threshold value being representative of a signature of an electric arc.

### Claim 3 (depends on 2)

The detection system according to claim 2, wherein the acquisition matrix is a random acquisition matrix ϕ which differs at each iteration, and at each iteration multiplication of the sampled signal with the acquisition matrix forms a current sub-sampled signal y of dimension M where M<N, and said processor is further configured to:
calculate, from the current sub-sampled signal y, a current reconstruction sampled signal f by seeking a solution which minimizes a predetermined norm l_1 which measures the parsimony of the sampled signal,
identify a predetermined number k of current estimated temporal positions {circumflex over (p)}1, . . . , {circumflex over (p)}k of a greatest amplitudes of the sampled reconstruction signal {circumflex over (f)};
determine the parsimony probability P_L of said sampled signal f by calculating a proportion of a number of iterations during which the estimated temporal positions {circumflex over (p)}1, . . . , {circumflex over (p)}k of the greatest amplitudes of the sampled reconstruction signal f are concordant with a true temporal positions {circumflex over (p)}1, . . . , {circumflex over (p)}k of the greatest amplitudes of the sampled signal f,
compare said parsimony probability P_L with said predetermined threshold value, said temporal acoustic signal being a signature of an electric arc if the parsimony probability P_L is greater than said predetermined threshold value.

### Claim 4 (depends on 3)

The detection system according to claim 3, wherein said predetermined number k is equal to one and said processor is configured to:
identify a current estimated temporal position of the greatest amplitude of the sampled reconstruction signal {circumflex over (f)}; and
determine the parsimony probability P_L of said sampled signal f by calculating the proportion of the number of iterations during which each estimated temporal position of the greatest amplitude of the sampled reconstruction signal f is concordant with the true temporal position of the greatest amplitude of the sampled signal f.

### Claim 5 (depends on 3)

The detection system according claim 3, wherein said processor is configured to express the sampled signal f before the sampled signal f is compressed in another orthonormal base in which the sampled signal f is parsimonious, by using a transfer matrix.

### Claim 6 (depends on 2)

The detection system according to claim 2, wherein the pre-conditioning circuit comprises:
a modulator configured to form a modulated acoustic signal by modulating said temporal acoustic signal according to a predetermined central frequency and a predetermined frequency band,
a low-pass filter configured to form a filtered acoustic signal by performing low-pass filtering according to a predetermined cut-off frequency on said modulated acoustic signal, and
a sampler configured to sample the filtered acoustic signal at a sampling frequency equal to or greater than double the cut-off frequency thus forming said sampled signal f.

### Claim 7 (depends on 6)

The detection system according to claim 6, wherein said at least one sensor is selected such that a gain diagram of the at least one sensor exhibits a maximum gain within the predetermined frequency band.

### Claim 8 (depends on 6)

The detection system according to claim 6, wherein said modulator is configured to form the modulated acoustic signal by performing band-pass filtering on the temporal acoustic signal according to the predetermined frequency band and by multiplying the temporal acoustic signal by a cosine function of the predetermined central frequency and zero phase-shift.

### Claim 9 (depends on 6)

The detection system according to claim 6, wherein the determined central frequency has a value between 30 kHz and 200 kHz and preferably equal to about 100 kHz, where the predetermined frequency band has a value between 10 kHz and 50 kHz and preferably equal to about 20 kHz, where the cut-off frequency is about 5 kHz, and where the sampling frequency is about 10 kHz.

### Claim 10 (depends on 1)

The detection system according to claim 1, wherein the electrical installation comprises a set of electrical elements, each equipped with a specific sensor to allow the electrical element at the location at which the electric arc is produced to be identified.

### Claim 11

A method for detecting an electric arc in an electrical installation equipped with at least one sensor configured to detect temporal acoustic signals produced by the electrical installation, said method comprising:
analyzing a temporal acoustic signal coming from said at least one sensor by associating with said temporal acoustic signal a corresponding parsimony measurement, a value of the parsimony measurement indicating whether or not the temporal acoustic signal is a signature of an electric arc; and
triggering an alarm when the value of the parsimony measurement indicates that the temporal signal is a signature of an electric arc,
wherein said analyzing comprises forming a sampled signal by sampling of said temporal acoustic signal, forming a compressed signal by compressing the sampled signal, and reconstructing the sampled signal from said compressed signal, and
the compressed signal is generated by iteratively multiplying the sampled signal at each current iteration with an acquisition matrix which differs at each iteration.