Sensor Fall Curve Identification

Patent metadata

Field	Value
Publication number	US-2020150640-A1
Application number	US-201816189736-A
Country	US
Kind code	A1
Filing date	Nov 13, 2018
Priority date	Nov 13, 2018
Publication date	May 14, 2020
Grant date	—

Abstract

A computer implemented method includes turning off a sensor, receiving fall curve data from the sensor, and comparing the received fall curve data to a set of fall curve signatures to identify the sensor or a sensor fault.

First claim

1 . A computer implemented method comprising:
turning off a sensor; receiving fall curve data from the sensor; and comparing the received fall curve data to a set of fall curve signatures to identify the sensor or a sensor fault.

2 . The method of claim 1 wherein the fall curve data corresponds to an analog signal output from the sensor in response to turning off the sensor.

3 . The method of claim 2 wherein the fall curve data comprises a time series sampling of the analog signal that is sampled periodically or in response to anomalous data received from the sensor prior to turning off the sensor.

4 . The method of claim 3 wherein the set of fall curve signatures include fall curve signatures from properly operating sensors that are different for different types of sensors and fall curve signatures from faulty sensors that are different for different types of faults.

5 . The method of claim 1 wherein the fall curve data comprises a time series sampled analog output from the sensor in response to turning off the sensor and wherein comparing the received fall curve data comprises:
generating a feature vector for the time series sampled analog output; and
comparing the feature vector with feature vectors corresponding to the set of fall curve signatures to match the generated feature vector with one of the feature vectors of the set of fall curve signatures.

6 . The method of claim 5 wherein the feature vector is generated by:
fitting a polynomial curve to the time series sampled analog output;
using polynomial coefficients of the polynomial curve as the feature vector;
finding a nearest neighbor to the polynomial coefficients; and
identifying the sensor or sensor fault in response to the nearest neighbor being within a certain threshold.

7 . The method of claim 6 wherein the hyper-parameters comprising fall curve width and polynomial degree are tuned to trade off accuracy based on resource and power constraints.

8 . The method of claim 5 wherein the feature vectors corresponding to the set of fall curve signatures are generated by:
recording multiple fall curve signatures for multiple sensors and an ID for each of the multiple sensors;
fitting polynomial curves to the recorded signatures;
using polynomial coefficients of the polynomial curve as the feature vector; and
performing clustering on the feature vectors for each sensor to identify a smaller set of unique features for each sensor.

9 . The method of claim 8 and further comprising optimizing a set of hyper-parameters comprising fall curve width, degree of polynomial, and number of clusters for the feature vectors.

10 . The method of claim 1 wherein the sensor comprises a wireless sensor.

11 . The method of claim 1 wherein the set of fall curve signatures comprises fall curve signatures from sensors having one or more of the faults comprising short fault, spike fault, stuck-at fault.

12 . The method of claim 1 wherein comparing the received fall curve data to a set of fall curve signatures to identify the sensor or a sensor fault is performed. via cloud-based computing resources or network edge based resources.

13 . The method of claim 1 wherein comparing the received fall curve data to a set of fall curve signatures to identify the sensor or a sensor fault is performed by the sensor comprising a digital sensor.

14 . A machine-readable storage device having instructions for execution by a processor of a machine to cause the processor to perform operations to perform a method of managing communication accounts, the operations comprising:
turning off a sensor; receiving fall curve data from the sensor; and comparing the received fall curve data to a set of fall curve signatures to identify the sensor or a sensor fault.

15 . The device of claim 14 wherein the set of fall curve signatures include fall curve signatures from properly operating sensors that are different for different types of sensors and fall curve signatures from faulty sensors that are different for different types of faults.

16 . The device of claim 14 wherein the fall curve data comprises a time series sampled analog output from the sensor in response to turning off the sensor and wherein comparing the received fall curve data comprises:
generating a feature vector for the time series sampled analog output; and
comparing the feature vector with feature vectors corresponding to the set of fall curve signatures to match the generated feature vector with one of the feature vectors of the set of fall curve signatures.

17 . The device of claim 16 wherein the feature vector is generated by:
fitting a polynomial curve to the timer series sampled analog output;
using polynomial coefficients of the polynomial curve as the feature vector;
finding a nearest neighbor to the polynomial coefficients; and
identifying the sensor or sensor fault in response to the nearest neighbor being within a certain threshold.

18 . The device of claim 17 wherein the feature vectors corresponding to the set of fall curve signatures are generated by:
recording multiple fall curve signatures for multiple sensors and an ID for each of the multiple sensors;
fitting polynomial curves to the recorded signatures;
using polynomial coefficients of the polynomial curve as the feature vector; and
performing clustering on the feature vectors for each sensor to identify a smaller set of unique features for each sensor.

19 . A device comprising:
a processor; an analog sensor coupled to the processor; and a memory device coupled to the processor and having a program stored thereon for execution by the processor to perform operations comprising:
turning off a sensor;
receiving fall curve data from the sensor; and
comparing the received fall curve data to a set of fall curve signatures to identify the sensor or a sensor fault.

20 . The device of claim 19 wherein the fall curve data comprises a time series sampled analog output from the sensor in response to turning off the sensor and wherein comparing the received fall curve data comprises:
generating a feature vector for the time series sampled analog output; and
comparing the feature vector with feature vectors corresponding to the set of fall curve signatures to match the generated feature vector with one of the feature vectors of the set of fall curve signatures.

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Chakraborty Tusher
• Srirangam Narashiman Akshay Uttama Nambi
• Chandra Ranveer
• Sharma Rahul Anand
• Swaminathan Manohar
• Kapetanovic Zerina
• Appavoo Jonathan

Classifications

• G01D18/00

  Testing or calibrating apparatus or arrangements provided for in groups G01D1/00 - G01D15/00 · CPC title

• G05B23/0232 Primary

  based on qualitative trend analysis, e.g. system evolution · CPC title

• G01N27/223

  for determining moisture content, e.g. humidity (rain detectors on vehicle windows B60S1/0825) · 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

• G01R31/2829

  Testing of circuits in sensor or actuator systems (testing of apparatus for measuring electric or magnetic variables G01R35/00; testing of indicating or recording apparatus G01D; in airbag systems B60R21/0173; checking gas analysers G01N33/007; monitoring or fail-safe circuits for electromagnets H01F7/1844) · CPC title