Testing the Model’s Predictions Locally with Workbench

The ScaleOut Digital Twins Workbench can be used to validate your model locally before deploying it.

First, create a new project UnitTests in your digital twin model solution (e.g. a console app). You will use this project to run some basic tests.

Second, add references to 2 Nuget packages:

• Scaleout.DigitalTwin.Workbench

• Scaleout.DigitalTwin.Workbench.MachineLearning

Note

The minimum version for these 2 packages is 4.4.0.

Finally, write a simple program to initialize the Workbench with your model, initialize the Anomaly Detection Provider and send a message.

The Scaleout.DigitalTwin.Workbench.MachineLearning package introduces a new API that allows you to initalize the anomaly detection provider and make it available in the Workbench. Call AddAnomalyDetectionProvider and provide:

• the name of the model,

• the reference name of the trained algorithm as chosen in the Training Tool,

• and the path to the zip file produced by the Training Tool.

static void Main(string[] args)
{
  using RealTimeWorkbench wb = new RealTimeWorkbench();

  // Add our trained algorithm for SensorsRT called Overheating
  wb.AddAnomalyDetectionProvider("SensorsRT", "Overheating", "Overheating.zip", null);

  // Add our SensorsRT model
  var endpoint = wb.AddRealTimeModel("SensorsRT", new SensorsRTMessageProcessor());

  // Send a message to an instance
  var msg = new SensorsMessage()
  {
      RPM = 42500f,
      Temperature = 228f,
      Friction = 2.31f
  };

  endpoint.Send("InstanceA", msg);
}

You can test your anomaly detection through logging or through a debugger and set a breakpoint in the ProcessMessages method.