Absolute and relative importance trend detection

1. A method comprising:

acquiring a current condition indicator of a condition indicator set associated with an operating condition of a machine, the condition indicator set indicating sensor readings associated with an operating element of the machine under the operating condition;

determining, by a data server, a relative trend significance over a trend window of the condition indicator set using data related to an evaluation of the trend window in relation to a historical window of the condition indicator set, the historical window spanning a period prior to a start of the trend window;

determining, by the data server, an absolute trend significance over the trend window using data related to an evaluation of the trend window in relation to an indicator threshold, wherein the determining the absolute trend significance comprises determining how many of a plurality of predicted data points of a future window of the condition indicator set cross the indicator threshold;

determining, by the data server, whether trend criteria associated with the operating element is satisfied, wherein the trend criteria comprises criteria related to the relative trend significance and criteria related to the absolute trend significance, and wherein the criteria related to the absolute trend significance is satisfied in response to at least a specified minimum portion of the plurality of predicted data points crossing the indicator threshold; and

executing, by the data server, an alerting process in response to the determining that the trend criteria is satisfied.

2. The method of claim 1 comprising:

generating a model of the trend window; and

wherein the determining the relative trend significance comprises comparing one or more values from the model to the historical window.

3. The method of claim 2 comprising:

analyzing a quality of the model, the analyzing yielding a quality metric; and

causing alert suppression responsive to the quality metric not satisfying quality criteria.

4. The method of claim 1 comprising:

generating a linear regression model of the trend window;

analyzing a quality of the linear regression model using a coefficient of determination; and

causing alert suppression responsive to the coefficient of determination not satisfying quality criteria.

5. The method of claim 1 , wherein the determining the relative trend significance comprises fitting the historical window to a probability distribution.

6. The method of claim 5 , wherein the criteria related to the relative trend significance is satisfied in response to at least one of the following:

at least one value in a model of the trend window is greater than a specified percentage of the historical window according to the probability distribution; and

a most recent value in a model of the trend window is greater than a specified percentage of the historical window according to the probability distribution.

7. The method of claim 5 ; wherein the criteria related to the relative trend significance is satisfied in response to at least one of the following:

at least one sample in the trend window is greater than a specified percentage of the historical window according to the probability distribution;

at least a minimum number of samples in the trend window are greater than a specified percentage of the historical window according to the probability distribution; and

a most recent sample in the trend window is greater than a specified percentage of the historical window according to the probability distribution.

8. The method of claim 5 , wherein the criteria related to the relative trend significance is satisfied in response to at least one of the following:

an aggregate value for individual samples of the trend window is greater than a specified percentage of the historical window according to the probability distribution, wherein the aggregate value is selected from the group consisting of minimum, maximum, mean, median, and mode; and

an aggregate value for a model of the trend window is greater than a specified percentage of the historical window according to the probability distribution, wherein the aggregate value is selected from the group consisting of minimum, maximum, mean, median, and mode.

9. The method of claim 5 , wherein the probability distribution comprises a normal probability distribution.

10. The method of claim 1 , wherein:

the determining the relative trend significance comprises:

determining a first aggregate value for the trend window and a second aggregate value for the historical window; and

comparing the first aggregate value to the second aggregate value, the comparing yielding a comparison metric; and

the criteria related to the relative trend significance is satisfied in response to the comparison metric meeting a comparison-metric threshold.

11. The method of claim 1 comprising:

generating a first linear regression model for the trend window;

generating a second linear regression model for the historical window;

wherein the determining the relative trend significance comprise comparing a slope of the first linear regression model to a slope of the second linear regression model, the comparing yielding a comparison metric; and

wherein the criteria related to the relative trend significance is satisfied in response to the comparison metric meeting a comparison-metric threshold.

12. The method of claim 1 , wherein the trend window and the historical window are non-overlapping windows.

13. The method of claim 1 , wherein the plurality of predicted data points of the future window of the condition indicator set are predicted using a model of the trend window.

14. The method of claim 1 , comprising:

determining a trend indication in the trend window responsive to the determining how many of the plurality of predicted data points cross the indicator threshold;

determining, based on a probability distribution of the historical window, a probability of crossing the indicator threshold; and

disregarding the trend indication in response to the probability satisfying specified criteria.

15. The method of claim 1 , wherein:

the determining the absolute trend significance comprises determining an aggregate value for the trend window, wherein the aggregate value is selected from the group consisting of mean, median, mode, maximum and minimum; and

the criteria related to the absolute trend significance is satisfied in response to the aggregate value falling within a specified range of the indicator threshold.

16. The method of claim 1 , wherein the executing the alerting process comprises generating an alert signal.

17. A computer system comprising a processor and memory, wherein the processor and the memory in combination are operable to perform a method comprising:

acquiring a current condition indicator of a condition indicator set associated with an operating condition of a machine, the condition indicator set indicating sensor readings associated with an operating element of the machine under the operating condition;

determining a relative trend significance over a trend window of the condition indicator set using data related to an evaluation of the trend window in relation to a historical window of the condition indicator set, the historical window spanning a period prior to a start of the trend window;

determining an absolute trend significance over the trend window using data related to an evaluation of the trend window in relation to an indicator threshold, wherein the determining the absolute trend significance comprises determining how many of a plurality of predicted data points of a future window of the condition indicator set cross the indicator threshold;

determining whether trend criteria associated with the operating element is satisfied, wherein the trend criteria comprises criteria related to the relative trend significance and criteria related to the absolute trend significance, and wherein the criteria related to the absolute trend significance is satisfied in response to at least a specified minimum portion of the plurality of predicted data points crossing the indicator threshold; and

executing an alerting process in response to the determining that the trend criteria is satisfied.

18. A computer-program product comprising a non-transitory computer-usable medium having computer-readable program code embodied therein, the computer-readable program code adapted to be executed to implement a method comprising:

acquiring a current condition indicator of a condition indicator set associated with an operating condition of a machine, the condition indicator set indicating sensor readings associated with an operating element of the machine under the operating condition;

determining a relative trend significance over a trend window of the condition indicator set using data related to an evaluation of the trend window in relation to a historical window of the condition indicator set, the historical window spanning a period prior to a start of the trend window;

determining an absolute trend significance over the trend window using data related to an evaluation of the trend window in relation to an indicator threshold, wherein the determining the absolute trend significance comprises determining how many of a plurality of predicted data points of a future window of the condition indicator set cross the indicator threshold;

determining whether trend criteria associated with the operating element is satisfied, wherein the trend criteria comprises criteria related to the relative trend significance and criteria related to the absolute trend significance, and wherein the criteria related to the absolute trend significance is satisfied in response to at least a specified minimum portion of the plurality of predicted data points crossing the indicator threshold; and

executing an alerting process in response to the determining that the trend criteria is satisfied.

19. The computer-program product of claim 18 , the method comprising:

generating a model of the trend window; and

wherein the determining the relative trend significance comprises comparing one or more values from the model to the historical window.

20. The computer-program product of claim 19 , the method comprising:

analyzing a quality of the model, the analyzing yielding a quality metric; and

causing alert suppression responsive to the quality metric not satisfying quality criteria.