Perform Learning on Maintenance Data
This process is used to create recommendations for maintenance programs adjustments based on data analysis using the Oracle IoT Analytics platform.
Once complete, the recommendations are viewable in an infolet in the Maintenance Management landing page, as well as in the Recommendations task flow page.
You can access this report from the Scheduled Process page.
When to Use
It is suggested to run the process to create recommendations only after enough historical data has been accumulated in the IoT Analytics platform. The historical maintenance data must be routinely extracted, transformed, and loaded into the IoT Analytics platform using the process entitled Extract, Transform, and Load (ETL) Maintenance Data. Only once a representative sample has populated the data lake would running the process to perform learning produce any meaningful recommendations.
An example would be to run the ETL on a weekly or monthly basis, depending on the volume of maintenance work orders that have been completed. Then running the Perform Leaning on Maintenance Data process could be performed after 6 to 12 months, depending on the number of maintenance work orders being completed. Users will view the recommendations and determine if they want to implement the changes for maintenance programs adjustments.
It is then recommended to only rerun the process again after enough additional maintenance data is updated in the IoT Analytics platform to make additional suggestions.
Privileges Required
- Privilege: MNT_PERFORM_MACHINE_LEARNING_ON_MNT_DATA_PRIV
Specifications
Use these specifications when you run the scheduled process.
| Specification | Description |
|---|---|
| Job Type | On an adhoc basis, you can run at any time. |
| Frequency | This should only be run after sufficient data is accumulated in the IoT Analytics platform data lake. Then, it should only be run again after additional maintenance history is accumulated in order to provide meaningful recommendations. |
| Time of Day | Any time. |
| Duration | It could take a while depending on the volume of data being analyzed. |
| Compatibility | There should be only one instance of the job running at any one time. |
Parameters
You must define at least a range of Work Order numbers or dates to process the report.
| Parameter | Optional or Required | Description | Parameter Value | Special Combinations Required |
|---|---|---|---|---|
| Reliability rate | Optional | Parameter that is considered by the learning algorithms in making maintenance program adjustments. | Positive numerical value. | None |
| Reliability tolerance | Optional | Parameter that is considered by the learning algorithms in making maintenance program adjustments. | Positive numerical value. | None |
Troubleshooting Information
- Once submitted, you can view the status of the process on the Scheduled Processes UI. A successfully completed process, as well as any child processes, will end in Succeeded status.
- If issues are encountered, the process or any of its subprocesses may be Warning or Error. The specific validation errors and warning messages that prevented the process from completing successfully are displayed in the error log(s).
- Any interactive warning validations are NOT performed.
- When the process is submitted, you can Resubmit, Put on Hold, Cancel Process, Release Process as provided by the Scheduled Processes UI.