Categorical inference for training a machine learning model
US-2021390424-A1 · Dec 16, 2021 · US
US11568330B2 · US · B2
| Field | Value |
|---|---|
| Publication number | US-11568330-B2 |
| Application number | US-201916717666-A |
| Country | US |
| Kind code | B2 |
| Filing date | Dec 17, 2019 |
| Priority date | Dec 21, 2018 |
| Publication date | Jan 31, 2023 |
| Grant date | Jan 31, 2023 |
A practical reading order for non-experts. Skip the full description unless you need deep technical detail.
What the patent document calls the invention.
A short plain-language summary of the technical disclosure.
Who owns or filed the patent and who is credited as inventor.
Filing, priority, publication, and grant dates set the timeline.
The legal scope of protection — read this for what is actually claimed.
Technology tags used to group this patent with similar filings.
Prior art links and similar publications in this corpus.
Official abstract text for this publication.
The invention relates to a method and a device for recording digital data representative of an operating variable of an observed system, the digital data representative of said variable being obtained in the form of samples, the method comprising a data recording over time. The method receiving successive samples representative of said observed variable, and for a current sample, for at least two observation windows of different sizes, each observation window including a number, equal to the size of said window, of successive samples received before the moment in time corresponding to the current sample,calculating an average value per observation window,calculating a difference between the current sample and each of said average values,comparing each difference, in absolute value, to a predetermined threshold value associated with said observation window, and in case of excess, triggering a recording of the current sample in a non-volatile memory.
Opening claim text (preview).
The invention claimed is: 1. A method for recording digital data representative of an operating variable of an observed system, the digital data representative of said variable being obtained in the form of samples, the method comprising a data recording over time, the method comprising the following steps: receiving successive samples representative of said observed variable, and for a current sample, for at least two observation windows of different sizes, each observation window including a number, equal to the size of said window, of successive samples received before the moment in time corresponding to the current sample, calculating an average value per observation window, calculating a difference between the current sample and each of said average values, comparing each difference, in absolute value, to a predetermined threshold value associated with each of said at least two observation windows, and if the threshold value is exceeded for the threshold value associated with one of the observation windows, triggering a recording of the current sample in a non-volatile memory, the recording of the current sample being triggered only when the threshold value is exceeded for a threshold value associated with one of the observation windows. 2. The recording method according to claim 1 , implementing a plurality of observation windows of increasing sizes. 3. The recording method according to claim 1 , wherein the threshold values are all equal to a same threshold value. 4. The method according to claim 1 , wherein each threshold value is calculated as a function of the average value calculated for the corresponding observation window and a standard deviation of the samples of the corresponding observation window. 5. A method for monitoring and predictive maintenance of a physical system, including a capture at a predetermined temporal frequency of digital data representative of an operating variable of the system, implementing a method for recording data according to claim 1 and implementing an algorithm for monitoring and predictive maintenance from recorded data. 6. A device for recording digital data representative of an operating variable of an observed system, the digital data representative of said variable being obtained in the form of samples, the device being configured to perform a digital data recording over time, the device comprising modules configured to: receive successive samples representative of said observed variable, and for a current sample, for at least two observation windows of different sizes, each observation window including a number, equal to the size of said window, of successive samples received before the moment in time corresponding to the current sample, calculate an average value per observation window, calculate a difference between the current sample and each of said average values, compare each difference, in absolute value, to a predetermined threshold value associated with each of said at least two observation windows, and if the threshold value is exceeded for the threshold value associated with one of the observation windows, trigger a recording of the received sample in a non-volatile memory, the recording of the current sample being triggered only when the threshold value is exceeded for a threshold value associated with one of the observation windows. 7. The recording device according to claim 6 , wherein the modules are made by analog components. 8. The recording device according to claim 6 , including a computing processor, a non-volatile electronic memory unit, and a random-access memory unit, wherein said modules are made in software form including software instructions implemented by the computing processor. 9. The recording device according to claim 6 , implementing a plurality of observation windows of increasing sizes. 10. A system for monitoring and predictive maintenance of a physical system, including at least one sensor suitable for detecting, at a predetermined temporal frequency of digital data representative of an operating variable of said physical system, implementing a device for recording data according to claim 6 , and including a processor configured to implement an algorithm for monitoring and predictive maintenance from recorded data.
Solid-state data loggers · CPC title
Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202) · CPC title
Inference or reasoning models · CPC title
Administration of product repair or maintenance · CPC title
Related publications grouped by family.
Answers are generated from the same data shown on this page.