Before You Begin

This tutorial shows you how to use anomalies to detect deviations from normal asset behavior, and to flag and address device issues in time.

Background

Acme Corporation specializes in injection molding, assembly, and manufacturing services. They focus on custom molding of things like performance parts. Acme Corporation is project driven - their jobs run from a couple of days to a couple of weeks, from a hundred to a few thousand parts. Their customers range from large contractors to lone inventors.

Acme Corporation does not want to be a shoot-and-ship job shop. They are constantly moving towards automation. For the last one year, the company is using Robots. Robots cost only $3 to $5/hr, much less than the operator.

Acme needs to ensure optimum environmental conditions, such as temperature and humidity, for the proper functioning of their robots and plant. Too much humidity and the robots can rust. High temperature may lead to equipment burnout. On the other hand, too low a temperature can mean high air-conditioning power bills.

Acme also needs to ensure optimal performance of its robots and plant at all times with minimal downtimes. It needs automatic monitoring of any equipment problems and robot errors. Acme should ideally be able to predict, and proactively address potential problems in advance. The plant Operations Managers should not need to monitor dashboards 24x7. Automatic notifications should be triggered for potential issues and problems.

Oracle can help Acme Corporation using the Oracle Internet of Things Asset Monitoring Cloud Service. Oracle IoT Asset Monitoring Cloud Service creates a digital twin version of Acme Corporation's IOT-enabled organizational assets, and lets them monitor the location, condition, and utilization of their assets.

As an Acme Administrator, your responsibility is to set up the digital twin versions of Acme Corp's organization and assets. And as an Operations Manager, you will use Oracle IoT Asset Monitoring Cloud Service to monitor the location and health of Acme's IoT-enabled assets. To make sure you receive timely notification of issues, you'll create rules that generate alert and incident notifications when optimal environmental conditions are exceeded. To make sure the issues are addressed promptly, you'll create mobile notifications (SMS) for field technicians.

You will define anomalies to detect deviations from normal asset behavior, and to flag potential device issues in time.

What Do You Need?

Access to an instance of Oracle IoT Asset Monitoring Cloud Service

Define Automatic Anomalies to Detect Robot Arm Issues

During normal operations, the error status plot for the robot arm sensor stays flat. When an error occurs, the corresponding error code is reported.

We can create automatic anomalies to detect deviations from regular patterns.

1. Click Menu and then click Design Center.
2. Select Asset Types from the Design Center sub-menu.
3. Select the Robot asset type from the Asset Types list.
4. Click Anomalies.
5. On the Anomalies page, click Create Anomaly.
6. Under Name, enter Arm Anomaly.
7. Under Description, enter Anomalies in Robot Arm.
8. Under Detection Target, select the error_status sensor attribute for Attribute.
9. Under Training Data, select Automatic Anomaly.

   Use an automatic anomaly to automatically look for deviations in a sensor or metric (KPI) value.

10. Select a Specimen Asset that provides the data pattern for anomaly detection. This is one of your robot assets.
11. Specify a Training Window.

    The training window defines the normal operation data that you use to train the system.

    Select Static Window and specify a Training Start Time and Training End Time to define a one-time data window.

12. Choose the Deviation Percentage that will trigger the anomaly.

    If you believe a 10% deviation from normal value constitutes a hardware issue, select 10%.

    

13. Click Save.
14. Click Back to return to the Anomalies page.
    

    The anomaly training status shows if the internal anomaly training, based on the data you provided, is complete.

    

    The server performs the training and scoring of the automatic anomaly model at midnight. You can proceed with creating the pattern-based anomaly in the next exercise.

Define Anomaly Patterns to Detect Anomalous Temperature Changes

AcmeCorp needs to detect temperature spike anomalies that are being reported by the temperature sensors. We define anomalies using telltale patterns in existing sensor data to help detect and report future anomalous data.

1. Click Menu and then click Design Center.
2. Select Asset Types from the Design Center sub-menu.
3. Select the TemperatureDetector asset type from the Asset Types list.

   

4. Click Anomalies.
5. On the Anomalies page, click Create Anomaly.
6. Under Name, enter Temperature Spike Anomaly.
7. Under Description, enter To Detect Spikes in Temperature.
8. Under Detection Target, select the temperature sensor attribute for Attribute.
9. Under Training Data, select User Defined Anomaly.

   A user-defined anomaly lets you manually specify anomalous or normal data patterns for a sensor or metric. You can select the data pattern from existing sensor, or metric, data. Alternatively, you can manually plot an anomalous data pattern that the system uses to identify anomalies.

10. Select a Specimen Asset that provides the data pattern for anomaly detection. This is one of your temperature detector assets.

    Select the temperature detector that has been active for the longest time, and has the most data.

11. Under Selection Type, choose Anomalous Data.

    We will select anomalous temperature data pattern from existing sensor data.

12. Click Generate Chart to display the sensor data for temperature.

    You should be able to see many anomalous spikes in the normal sinusoidal temperature data pattern.

    

13. Zoom in to an anomalous area using the mouse scroll wheel. Use the mouse to select the anomaly pattern in the data plot.

    

14. Click Save.
15. Click Back to return to the Anomalies page.
    

    The anomaly training status shows if the internal anomaly training, based on the data you provided, is complete.

    

Inspect Anomalies in the Operations Center

Let us next return to the Operations Center to examine any anomalies that get reported, based on our definitions.

1. In the Operations Center, click the navigation icon next to your organization name in the breadcrumbs.

   

   If you are in one of the Design Configuration screens, then use the following steps to go back to the Operations Center:

   1. Click Menu, and then click Previous next to Design Center.

      

   2. Click Operations Center.

      

   3. Click the navigation icon next to your organization name in the breadcrumbs.

   Adjust the zoom settings as desired. You can also use the mouse scroll wheel for this.

   

2. Select the Robots Division group.

   

3. Click Anomalies in the Menu Bar to the left of the map.

   Any detected anomalies for the Robots Division appear. You can choose to expand an anomaly to see details, such as occurrence time.

   

   The Anomalies page includes anomalies for all robot assets and temperature detector sub-assets. You can use the breadcrumbs to filter the results for a particular asset. You can also choose the desired time period.

4. Click Map on the menu bar to go back to the map view.

Want to Learn More?

• Use Anomalies to Track Deviations in Asset Behavior
• The Operations Center