0 ×

04_​Creating_​a_​ControlChart_​of_​a_​Time_​Series

Workflow

Anomaly Detection. Control Chart
anomaly detectiontime series analysisauto-regressive modelstime alignmentIoTInternet of Things
- Read 313 spectral amplitudes time series produced by the Pre-processing/Time Alignment & Visualization workflow - missing value replaced by last value in time series - loop on all columns to calculate 1st level alarm Training - calculate cumulative average and cumulative standard deviation Deployment - throws first level alarm if signal wanders outdside boundaries (=average +- 2* stddev) - throws second level alarm if alarm 1 averaged across all frequency bands and all signals > 0.25 Take Action - if second level alarm fires => email for mechanical check up - plot of alarm level 2 for the whole time series to see when it was the earliest moment alarm level 2 would fire. Anomaly Detection: Control ChartThis workflow detects anomalies just by checking the wandering off of the signal from a band centered around the time series "normal conditions" averageand large as 4 times the corresponding standard deviation. remove datefrom loop on columnsline plotswith date on x-axisaverage alarmon y-axisread AlignedData.csvproduced by Pre-processing/Time Alignment & Visualizationlearning boundaries andthrow 1 if signal isoutside of boundariesmean(alarm 1) across signals at same timeif mean(alarm 1) > 0.25 => alarm 2resolve missingswith last value RowID Image Port Writer Generate LinePlot File Reader 1st level alarm 2nd level alarm Fire if level2 Alarm Missing Value - Read 313 spectral amplitudes time series produced by the Pre-processing/Time Alignment & Visualization workflow - missing value replaced by last value in time series - loop on all columns to calculate 1st level alarm Training - calculate cumulative average and cumulative standard deviation Deployment - throws first level alarm if signal wanders outdside boundaries (=average +- 2* stddev) - throws second level alarm if alarm 1 averaged across all frequency bands and all signals > 0.25 Take Action - if second level alarm fires => email for mechanical check up - plot of alarm level 2 for the whole time series to see when it was the earliest moment alarm level 2 would fire. Anomaly Detection: Control ChartThis workflow detects anomalies just by checking the wandering off of the signal from a band centered around the time series "normal conditions" averageand large as 4 times the corresponding standard deviation. remove datefrom loop on columnsline plotswith date on x-axisaverage alarmon y-axisread AlignedData.csvproduced by Pre-processing/Time Alignment & Visualizationlearning boundaries andthrow 1 if signal isoutside of boundariesmean(alarm 1) across signals at same timeif mean(alarm 1) > 0.25 => alarm 2resolve missingswith last value RowID Image Port Writer Generate LinePlot File Reader 1st level alarm 2nd level alarm Fire if level2 Alarm Missing Value

Download

Get this workflow from the following link: Download

Nodes

04_​Creating_​a_​ControlChart_​of_​a_​Time_​Series consists of the following 26 nodes(s):

Plugins

04_​Creating_​a_​ControlChart_​of_​a_​Time_​Series contains nodes provided by the following 6 plugin(s):