System and method for detecting a shift in real data trend using the configurable adaptive threshold

Patent metadata

Field	Value
Publication number	US-2020004619-A1
Application number	US-201816022059-A
Country	US
Kind code	A1
Filing date	Jun 28, 2018
Priority date	Jun 28, 2018
Publication date	Jan 2, 2020
Grant date	—

Abstract

A computer-implemented system for detecting shifts in data is provided. The system is configured to: calculate, based on the value of a plurality of user-selectable baseline configuration parameters, baseline values for a series of data items in a data structure, wherein the baseline values include an average value and a standard deviation value; calculate, based on the value of a plurality of user-selectable weighted threshold parameters, a weighted threshold level for the series of data items; detect, based on the value of a plurality of user-selectable shift detection parameters, a shift in the series of data items, wherein the shift comprises an abrupt shift, a rapid drift, or a gradual drift; convert, based on the value of a plurality of user-selectable normalization parameters, the value of each data item in the series of data items to a normalized value; and determine whether the normalized values indicate a data shift.

First claim

What is claimed is:

1 . A computer-implemented system for detecting shifts in data indicating a fault condition, the system comprising:
a baseline calculation module configured to calculate baseline values for a series of data items in a data structure, the baseline values including an average value and a standard deviation value for the series of data items, the baseline calculation module configured to calculate the average value and the standard deviation value based on the value of a plurality of user-selectable baseline configuration parameters; a weighted threshold calculation module configured to calculate one or more weighted threshold levels for the series of data items based on the value of a plurality of user-selectable weighted threshold parameters; a shift detection module configured to detect whether a shift in the series of data items exists, the shift comprising an abrupt shift, a rapid drift, or a gradual drift, the shift detection module configured by the value of a plurality of user-selectable shift detection parameters to detect the shift; and a normalization module configured to convert the value of each data item in the series of data items to a normalized value, the normalization module configured by the value of a plurality of user-selectable normalization parameters to convert each data item to its normalized value, the normalization module configured to determine whether the normalized values indicate a data shift.

2 . The system of claim 1 , wherein the baseline configuration parameters include a plurality of: an average calculation method parameter, which indicates a method for calculating an average value; a domain parameter, which indicates whether the baseline values are calculated based on a certain number of points or over a time period; a maximum baseline size parameter, which indicates a maximum number of points used for the calculation of baseline parameters; a buffer size parameter, which indicates a number of the most recent points that are included in a buffer and not included in baseline calculations; an outlier usage parameter, which indicates whether outliers should be included or excluded from baseline calculations; and an outlier sigma parameter, which indicates a sigma level beyond which outliers are filtered out when calculating baseline values.

3 . The system of claim 1 , wherein the user-selectable weighted threshold parameters include a plurality of: a threshold type parameter, which indicates a type of weighted threshold; a lower absolute threshold parameter, which indicates the value of a lower absolute threshold; a higher absolute threshold parameter, which indicates the value of a higher absolute threshold; a delta threshold from mean parameter, which indicates the value of a delta threshold from mean; a transition shape parameter, which indicates the shape of a transition from a fixed threshold to a weighted threshold; a data before transition parameter, which indicates an amount of data before starting a transition from a fixed to a weighted threshold; a transition length parameter, which indicates an amount of data for a transition from a fixed to a weighted threshold; and a weighting factor parameter, which indicates a weight of a fixed threshold in the weighted threshold.

4 . The system of claim 1 , wherein the user-selectable shift detection parameters include: a shift detection method selection parameter, which indicates a method for shift detection; and a shift detection direction selection parameter, which indicates a specific shift direction for which to detect.

5 . The system of claim 4 , wherein the user-selectable shift detection parameters further include a plurality of: a Western Electric (WE) rule selection parameter, which indicates the specific WE rules to be applied; a WE window size parameter, which indicates the size of window for WE rule evaluation; a WE number of exceptional points parameter, which indicates the number of points that has to exceed limits in WE rules to trigger the finding of a fault; a boundary sigma level parameter, which indicates a boundary sigma level that if exceeded triggers the finding of a fault; and a cumulative sigma level parameter, which indicates the sum of sigma levels that if exceeded during a window triggers the finding of a fault.

6 . The system of claim 4 , wherein the user-selectable shift detection parameters further include a smoothed window parameter, which indicates the number of points averaged when calculating the smoothed value.

7 . The system of claim 1 , wherein the user-selectable normalization parameters include a plurality of: a normalization select parameter, which indicates whether the normalization function is chosen for operation; a normalization floor parameter, which indicates the sigma level for a base; a normalization normal ceiling parameter, which indicates, sigma level for a normal upper boundary; a normalization abnormal cap parameter, which indicates the sigma level for an abnormal boundary; and a normalization ceiling for normal variation parameter, which indicates a value of a normal ceiling.

8 . The system of claim 1 , wherein the baseline calculation module is configured to calculate the baseline values by using a moving window of data items wherein the size of the moving window is determined by user-selectable parameter values.

9 . The system of claim 1 , wherein the baseline calculation module is configured to calculate the baseline values by using a buffer to exclude the most recent data items from the calculation wherein the size of the buffer is determined by user-selectable parameter values.

10 . The system of claim 1 , wherein the weighted threshold calculation module is configured to calculate weighted threshold levels by applying a first weighting factor to a fixed threshold and a second weighting factor to an adaptive threshold, wherein the fixed threshold is determined by user-selectable parameter values and wherein the adaptive threshold is calculated based on the calculated average and standard deviation of the data series.

11 . The system of claim 1 , wherein the weighted threshold calculation module is configured to transition from a fixed threshold to the weighted threshold using a ramp transition when user-selectable parameters indicate that the ramp transition be used and using a step transition when user-selectable parameters indicate that the step transition be used.

12 . The system of claim 1 , wherein the shift detection module is configured to apply Western Electric rules to detect a shift when a user-selectable parameter value indicates application of the Western Electric rules.

13 . The system of claim 1 , wherein the shift detection module is configured to apply a smoothing window to detect a shift when a user-selectable parameter value indicates application of the smoothing window.

14 . The system of claim 1 , wherein to calculate the normalized value, the normalization module is configured to:
compute a normal floor threshold level, a normal ceiling threshold level, and an abnormal cap threshold level; and map raw values of input data items to normalized values, wherein to map the raw values to normalized values, the normalization module is further configured to:
map the raw value to a minimum value near zero when the raw value is less than the normal floor threshold level;
linearly scale the raw value between the minimum value and a ceiling for normal variation value when the raw value is between the normal floor threshold level and the normal ceiling threshold level;
linearly scale the raw value between the ceiling for normal variation value and the value

Assignees

• Honeywell Int Inc

Inventors

• Kim Kyusung
• Hickenbottom Christopher
• Uluyol Onder
• Ertl Lukas
• Rudolecky Tomas
• Hrncir Zdenek

Classifications

• G05B23/024 Primary

  Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks · CPC title

• G06F11/0754 Primary

  by exceeding limits · CPC title

• G06F17/18

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title

• G06F11/0769

  Readable error formats, e.g. cross-platform generic formats, human understandable formats · CPC title