Processing sensor logs

Patent metadata

Field	Value
Publication number	US-10885456-B2
Application number	US-202016853093-A
Country	US
Kind code	B2
Filing date	Apr 20, 2020
Priority date	Dec 16, 2016
Publication date	Jan 5, 2021
Grant date	Jan 5, 2021

Abstract

A method of processing sensor logs is described. The method includes accessing a first sensor log and a corresponding first reference log. Each of the first sensor log and the first reference log includes a series of measured values of a parameter according to a first time series. The method also includes accessing a second sensor log and a corresponding second reference log. Each of the second sensor log and the second reference log includes a series of measured values of a parameter according to a second time series. The method also includes dynamically time warping the first reference log and/or and second reference log by a first transformation between the first time series and a common time-frame and/or a second transformation between the second time series and the common time-frame. The method also includes generating first and second warped sensor logs by applying the or each transformation to the corresponding ones of the first and second sensor logs.

First claim

The invention claimed is:

1. A computer-implemented method for aligning time series data of sensor logs to enable identification of anomalous sensor data, the method comprising:
accessing a first sensor log and a first reference log corresponding to the first sensor log, wherein the first sensor log comprises a first series of measured values of a first parameter according to a first time series and the first reference log comprises a first series of measured values of a second parameter according to the first time series;
accessing a second sensor log and a second reference log corresponding to the second sensor log, wherein the second sensor log comprises a second series of measured values of the first parameter according to a second time series and the second reference log comprises a second series of measured values of the second parameter according to the second time series;
dynamically time warping the first reference log and the second reference log, wherein dynamically time warping the first and second reference logs comprises:
dividing the first time series into a first plurality of sub-periods and the second time series into a second plurality of sub-periods;
determining a type of at least one sub-period; and
executing a time warping algorithm based at least in part on the type of the at least one sub-period to transform each of the first and the second reference logs to a common time-frame;

determining, based at least in part on the dynamically time warping the first and second reference logs, a first transformation between the first time series and the common time-frame and a second transformation between the second time series and the common time-frame;
applying the first transformation to the first sensor log to generate a first warped sensor log and the second transformation to the second sensor log to generate a second warped sensor log; and
causing a presentation of a user interface, the user interface including a representation of at least a first section of the first warped sensor log and at least a second section of the second warped sensor log that corresponds to the first section.

2. The method of claim 1 , wherein determining the type of the at least one sub-period comprises determining a respective type of each of the first plurality of sub-periods and each of the second plurality of sub-periods, wherein dynamically time warping the first reference log and the second reference log further comprises determining, based at least in part on each respective sub-period type, a first mapping of the first plurality of sub-periods to a warped time series comprising a set of intervals of the common time-frame and a second mapping of the second plurality of sub-periods to the warped time series, and wherein the first mapping defines the first transformation and the second mapping defines the second transformation.

3. The method of claim 2 , wherein the first mapping is a one-to-one mapping function that maps each point in the first time series to a respective corresponding one point in the warped time series, wherein no two points in the first time series are mapped to a same point in the warped time series.

4. The method of claim 2 , wherein the first mapping is a many-to-one mapping function that maps two or more points in the first time series to a respective corresponding one point in the warped time series.

5. The method of claim 2 , wherein dynamically time warping the first reference log further comprises identifying, in the first plurality of sub-periods, a first group of consecutive sub-periods corresponding to a first type and a second group of consecutive sub-periods corresponding to a second type.

6. The method of claim 5 , wherein the first transformation is a non-linear transformation according to which at least one of: i) the first group of sub-periods is expanded in time to span a first specific interval of the set of intervals of the common time-frame or ii) the second group of sub-periods is contracted in time to span a second specific interval of the set of intervals.

7. The method of claim 2 , wherein dynamically time warping the first reference log further comprises:
aggregating the first plurality of sub-periods into at least a first group of sub-periods and a second group of sub-periods, wherein each sub-period in the first group of sub-periods corresponds to a first sub-period type and each sub-period in the second group of sub-periods corresponds to a second sub-period type;
ordering the first group of sub-periods and the second group of sub-periods based at least in part on the first sub-period type and the second sub-period type; and
mapping the first group of sub-periods to a first interval of the set of intervals of the common time-frame and mapping the second group of sub-periods to a second interval of the set of intervals.

8. The method of claim 7 , wherein the first type of sub-period and the second type of sub-period are dependent at least in part on the second parameter.

9. The method of claim 7 , wherein aggregating the first plurality of sub-periods into at least the first group of sub-periods and the second group of sub-periods comprises:
identifying a first sub-period corresponding to the first sub-period type;
identifying a second sub-period corresponding to the first sub-period type, wherein the first sub-period and the second sub-period are non-consecutive;
arranging the first sub-period and the second sub-period consecutively.

10. The method of claim 7 , further comprising:
calculating, based at least in part on all sub-periods in the first group of sub-periods corresponding to the same first sub-period type, a single value for each time within the first interval of the set of intervals.

11. The method of claim 1 , further comprising:
accessing a plurality of additional sensor logs and a corresponding plurality of additional reference logs, each additional sensor log and each additional reference log comprising a respective series of measured values of a third parameter according to a third time series, each sensor log corresponding to a first sensor type;
executing the time warping algorithm to transform the plurality of sensor logs to the common time-frame; and
generating a baseline sensor log for the first sensor type at least in part by calculating, using each respective series of measured values of the third parameter, an average or expected value of the third parameter for a common time series spanning the common time-frame.

12. A system for aligning time series data of sensor logs to enable identification of anomalous sensor data, the system comprising:
at least one processor; and
at least one memory storing computer-executable instructions, wherein the at least one processor is configured to access the at least one memory and execute the computer-executable instructions to:
access a first sensor log and a first reference log corresponding to the first sensor log, wherein the first sensor log comprises a first series of measured values of a first parameter according to a first time series and the first reference log comprises a first series of measured values of a second parameter according to the first time series;
access a second sensor log and a second reference log corresponding to the second sensor log, wherein the second sensor log comprises a second series of measured values of the first parameter according to a second time series and the second reference log comprises a second series of measured values of the second parameter according to the second time series;
dynamically time warp the first reference log and the second reference log, wherein the at least one processor is configured to

Assignees

• Palantir Technologies Inc

Inventors

• Spiro Ezra
• Menck Andre Frederico Cavalheiro
• Maag Peter
• Powell Thomas

Classifications

• G05B23/0221 Primary

  Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods · CPC title

• G06F2218/16

  by matching signal segments · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06F18/40

  Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor · CPC title

• G05B23/0237

  based on parallel systems, e.g. comparing signals produced at the same time by same type systems and detect faulty ones by noticing differences among their responses · CPC title