Eviction of weakly correlated signals from collections
Systems, methods, and other embodiments associated with eviction of weakly correlated signals from collections are described. In one embodiment, a mock signal that has random signal properties is generated. A mock correlation coefficient between the mock signal and a measured time series signal from a collection of measured time series signals is then generated. A discrimination value that indicates a weak signal correlation is then selected, based at least in part on the mock correlation coefficient. A first measured signal is then identified from the collection of measured time series signals that has the weak signal correlation by determining that a first correlation coefficient between the first measured signal and a second measured signal is weak based on the discrimination value. The first measured signal is then evicted from the collection of signals in response to the determination that the first measured signal has the weak signal correlation.
1. A computer-implemented method, comprising:
generating a mock signal that has at least random signal properties;
generating a mock correlation coefficient between the mock signal and at least one measured time series signal from a collection of measured time series signals;
selecting a discrimination value that indicates a weak signal correlation based at least in part on the mock correlation coefficient;
identifying a first measured signal from the collection of measured time series signals that has the weak signal correlation based on comparing a first correlation coefficient between the first measured signal and a second measured signal to the discrimination value; and
evicting the first measured signal from the collection of measured time series signals in response to identifying that the first measured signal has the weak signal correlation.
2. The computer-implemented method of claim 1 , further comprising:
generating one or more additional mock signals that have at least random signal properties, wherein the additional mock signals are generated to have a sequence of signal values unique among the mock signals;
for each of the additional mock signals,
replacing the mock signal with the additional mock signal, and
generating an additional mock correlation coefficient between the additional mock signal and at least one measured time series signal from the collection of measured time series signals;
for each of the mock signals, identifying a maximum magnitude mock correlation coefficient between the mock signal and the measured time series signals; and
finding an average based on the maximum magnitude mock correlation coefficients;
wherein the discrimination value selected is the average of the maximum magnitude mock correlation coefficients.
3. The computer-implemented method of claim 2 , wherein a total number of mock signals for which a maximum magnitude mock correlation coefficient is generated either (i) is at least twenty or (ii) causes the average of the maximum magnitude mock correlation coefficients to not vary in a fourth decimal place.
4. The computer-implemented method of claim 2 , further comprising:
generating a histogram of the maximum magnitude mock correlation coefficients; and
finding a mean of the histogram;
wherein the average based on the maximum magnitude mock correlation coefficients is the mean of the histogram.
5. The computer-implemented method of claim 1 , further comprising generating a Gaussian random number sequence of a given number of observations, wherein signal values of the mock signal are the Gaussian random number sequence, and wherein the given number of observations is a number of observations for the at least one measured time series signal.
6. The computer-implemented method of claim 1 , further comprising:
evicting signals in the collection of measured time series signals that have a maximum correlation coefficient with an other signal in the collection of measured time signals that indicates weaker correlation than the discrimination value;
training a multivariate anomaly detection model using signals selected from remaining, non-evicted signals in the collection of measured time series signals, and not using evicted signals; and
monitoring the selected, non-evicted signals with the trained multivariate anomaly detection model to detect anomalies.
7. The computer-implemented method of claim 1 , further comprising including the second measured signal in a univariate anomaly detection model in response to identifying that the first measured signal has the weak signal correlation.
8. A non-transitory computer-readable medium that includes stored thereon computer-executable instructions that when executed by at least a processor of a computer cause the computer to:
generate a mock signal that has at least random signal properties;
generate a mock correlation coefficient between the mock signal and at least one measured time series signal from a collection of measured time series signals;
select a discrimination value that indicates a weak signal correlation based at least in part on the mock correlation coefficient;
identify a first measured signal from the collection of measured time series signals that has the weak signal correlation based on comparing a first correlation coefficient between the first measured signal and a second measured signal to the discrimination value; and
evict the first measured signal from the collection of measured time series signals in response to identifying that the first measured signal has the weak signal correlation.
9. The non-transitory computer-readable medium of claim 8 , wherein the instructions further cause the computer to:
generate one or more additional mock signals that have at least random signal properties, wherein the additional mock signals are generated to have a sequence of signal values unique among the mock signals;
for each of the additional mock signals,
replace the mock signal with the additional mock signal, and
generate an additional mock correlation coefficient between the additional mock signal and at least one measured time series signal from the collection of measured time series signals;
for each of the mock signals, identify a maximum magnitude mock correlation coefficient between the mock signal and the measured time series signals; and
find an average based on the maximum magnitude mock correlation coefficients;
wherein the discrimination value selected is the average of the maximum magnitude mock correlation coefficients.
10. The non-transitory computer-readable medium of claim 9 , wherein a total number of mock signals for which a maximum magnitude mock correlation coefficient is generated either (i) is at least twenty or (ii) causes the average of the maximum magnitude mock correlation coefficients to not vary in a fourth decimal place.
11. The non-transitory computer-readable medium of claim 9 , wherein the instructions further cause the computer to:
generate a histogram of the maximum magnitude mock correlation coefficients; and
find a mean of the histogram;
wherein the average based on the maximum magnitude mock correlation coefficients is the mean of the histogram.
12. The non-transitory computer-readable medium of claim 8 , wherein the instructions further cause the computer to generate a Gaussian random number sequence of a given number of observations, wherein signal values of the mock signal are the Gaussian random number sequence, and wherein the given number of observations is a number of observations for the at least one measured time series signal.
13. The non-transitory computer-readable medium of claim 8 , wherein the instructions further cause the computer to:
evict signals in the collection of measured time series signals that have a maximum correlation coefficient with an other signal in the collection of measured time signals that indicates a weaker correlation than the discrimination value;
train a multivariate anomaly detection model using signals selected from remaining, non-evicted signals in the collection of measured time series signals, and not using evicted signals; and
monitor the selected, non-evicted signals with the trained multivariate anomaly detection model to detect anomalies.
14. The non-transitory computer-readable medium of claim 8 , wherein the instructions further cause the computer to include the second measured signal in a univariate anomaly detection model in response to identifying that the first measured signal has the weak signal correlation.
15. A computing system comprising:
a processor;
a memory operably connected to the processor;
a non-transitory computer-readable medium operably connected to the processor and memory and storing computer-executable instructions for autonomous discrimination of operation vibration signals that when executed by at least a processor of the computing system cause the computing system to:
generate a mock signal that has at least random signal properties;
generate a mock correlation coefficient between the mock signal and at least one measured time series signal from a collection of measured time series signals;
selecting a discrimination value that indicates a weak signal correlation based at least in part on the mock correlation coefficient;
identify a first measured signal from the collection of measured time series signals that has the weak signal correlation based on comparing a first correlation coefficient between the first measured signal and a second measured signal is smaller than the discrimination value; and
evict the first measured signal from the collection of measured time series signals in response to identifying that the first measured signal has the weak signal correlation.
16. The computing system of claim 15 , wherein the instructions further cause the computing system to:
generate one or more additional mock signals that have at least random signal properties, wherein the additional mock signals are generated to have a sequence of signal values unique among the mock signals;
for each of the additional mock signals,
replace the mock signal with the additional mock signal, and
generate an additional mock correlation coefficients between the additional mock signal and at least one measured time series signal from the collection of measured time series signals;
for each of the mock signals, identify a maximum magnitude mock correlation coefficient between the mock signal and the measured time series signals; and
find an average based on the maximum magnitude mock correlation coefficients;
wherein the discrimination value selected is the average of the maximum magnitude mock correlation coefficients.
17. The computing system of claim 16 , wherein a total number of mock signals for which a maximum magnitude mock correlation coefficient is generated either (i) is at least twenty or (ii) causes the average of the maximum magnitude mock correlation coefficients to not vary in a fourth decimal place.
18. The computing system of claim 16 , wherein the instructions further cause the computing system to:
generate a histogram of the maximum magnitude mock correlation coefficients; and
find a mean of the histogram;
wherein the average based on the maximum magnitude mock correlation coefficients is the mean of the histogram.
19. The computing system of claim 15 , wherein the instructions further cause the computing system to generate a Gaussian random number sequence of a given number of observations, wherein signal values of the mock signal are the Gaussian random number sequence, and wherein the given number of observations is a number of observations for the at least one measured time series signal.
20. The computing system of claim 15 , wherein the instructions further cause the computing system to:
evict signals in the collection of measured time series signals that have a maximum correlation coefficient with an other signal in the collection of measured time signals that indicates a weaker correlation than the discrimination value;
train a multivariate anomaly detection model using signals selected from remaining, non-evicted signals in the collection of measured time series signals, and not using evicted signals; and
monitor the selected, non-evicted signals with the trained multivariate anomaly detection model to detect anomalies.