Feature Description¶
The Counter Guard feature is based on the predefined signal Counter, see Proemion Standard Metrics, that is used by the DataPortal.
In the DataPortal, the Counter Guard detects inconsistencies events in the counter signal data.
As an example, this is relevant for the following features listed below (see DataPortal User Manual):
- Cluster Heat Map, Counter, Gauge widgets, see Widget Catalog.
-
Maintenance Tasks
-
Machine Actions
-
Assignments reporting tool
The Counter Guard feature is an included DataPortal feature.
Use Cases¶
Signals defined as counter can be calculated in order to analyze their usage/consumption/operation over time.
The Counter Guard can be also used by OEMs to check for inconsistencies in the data sent by machines, i.e.to have a continuous counter e.g.in case of erroneous resets by the machines.
By applying patches to the machine controllers, they will then get clean data which is a precondition to use it for other DataScience applications and features.
Thereby, the Counter Guard can be a basis for other DataScience features.
The Counter Guard consists of two sub-features for detecting issues with counters, these are:
Counter Violation Detection¶
The Counter Violation Detection sub feature detects counter violation events in machine counters.
A counter violation is a period in the time series in which the signal behaved suspicious.
This is done by checking the monotonicity of the time series.
The monotonicity is interrupted by peaks (change of counter direction), therefore Counter Guard checks if peaks occurred in the given time series.
For each peak group a counter violation event is created, it includes a snippet of the time series in which the suspicious behavior was observed.
Note
The approach detects peaks independent of the counters context and direction.
This excludes the solution from detecting contextual outliers.
Warning
Counter resets (intended or not) will be considered as counter violation.
Note that signals can be ignored, see Configuration.
Invalid Increase Detection¶
The Invalid Increase Detection sub feature detects invalid increase events on increasing hour based counters (e.g.operating hours counter).
An invalid increase is a period in the time series in which more hours were logged than time has passed.
This is done by dividing the time series into sessions of work and comparing the sum of the deltas of time passed in hours to the sum of the deltas of hours logged within a session.
For each session, where the sum of hours logged exceeds the sum of time passed by more than the hours time exceedance threshold, an event is created.
A session is defined by the time passed between logs in hours.
If this delta between two data points is greater than the minimum session distance the machine is considered to have a break in operation and therefore the session has ended.
A session needs to have the minimum session data points to be eligible for invalid increase detection.
By using the sum of deltas of the hours logged, the solution is resistant against resets and spikes (e.g.caused by unintended machine start/stop behavior) as long as the value returns to its pre-error value.
Those cases are already covered by the Counter Violation Detection.
Activation¶
In the DataPortal, you can activate the Counter Guard feature via the Feature Switch, see Feature Switch.
You can then view the events/notifications in the DataPortal via Event Functionality.
Configuration¶
The Counter Guard is a built-in system function, see Counter Guard Signals.
No special metrics need to be transferred to use this function.
Signals to be checked for monotonicity must be defined as category xsi:type="Counter" in the PDC Value Definition.
Note
By default, SIM traffic is ignored (for more information on the signal key, see Proemion Standard Metrics → "Proemion System Metrics").
Note that other signals can be ignored, too.
To implement this, contact your Proemion representative.