How to build an ML Model.

Tutorials → Tutorial | Concept | How-to Guide | Reference

Step 1:

1. Navigate to Historian → Tables
2. Create table: ProductionData
3. Add tags:
   • Tag.ProductionRate
   • Tag.Efficiency
   • Tag.QualityScore
   • Tag.DowntimeMinutes
4. Set scan rate: 60 seconds
5. Enable storage

Step 2: Create KPI Calculator Class

1. Go to Scripts → Classes
2. Create class: KPICalculator
3. Configure as standard Methods class

csharp

public double CalculateOEE(double availability, double performance, double quality)
{
    return availability * performance * quality * 100;
}

public double GetAverageProduction(DateTime startTime, DateTime endTime)
{
    // Calculate average production rate over period
    double totalProduction = @Tag.TotalUnits;
    double hours = (endTime - startTime).TotalHours;
    return hours > 0 ? totalProduction / hours : 0;
}

public string GetProductionStatus()
{
    if (@Tag.ProductionRate > 100)
        return "High Performance";
    else if (@Tag.ProductionRate > 80)
        return "Normal";
    else
        return "Below Target";
}

Step 3: Create MCP Tool Class

1. Create new class: ProductionMCPTool
2. Select MCP Tool type
3. Add decorated methods:

csharp

[MCPMethod(Description = "Get current production KPIs")]
public object GetCurrentKPIs()
{
    return new {
        ProductionRate = @Tag.ProductionRate,
        Efficiency = @Tag.Efficiency,
        OEE = @Script.Class.KPICalculator.CalculateOEE(
            @Tag.Availability, @Tag.Performance, @Tag.Quality),
        Status = @Script.Class.KPICalculator.GetProductionStatus(),
        Timestamp = DateTime.Now
    };
}

[MCPMethod(Description = "Get production history for specified hours")]
public object GetProductionHistory(
    [MCPParameter(Description = "Hours to look back")] int hours)
{
    var endTime = DateTime.Now;
    var startTime = endTime.AddHours(-hours);
    
    // Query historian
    var data = @Historian.Table.ProductionData.GetData(startTime, endTime);
    
    return new {
        Period = $"Last {hours} hours",
        AverageRate = @Script.Class.KPICalculator.GetAverageProduction(startTime, endTime),
        TotalUnits = @Tag.TotalUnits,
        DataPoints = data.Rows.Count
    };
}

[MCPMethod(Description = "Analyze production trend")]
public string AnalyzeProductionTrend(
    [MCPParameter(Description = "Time period in hours")] int periodHours)
{
    var current = @Tag.ProductionRate;
    var average = @Script.Class.KPICalculator.GetAverageProduction(
        DateTime.Now.AddHours(-periodHours), DateTime.Now);
    
    if (current > average * 1.1)
        return "Trending Up - Production improving";
    else if (current < average * 0.9)
        return "Trending Down - Requires attention";
    else
        return "Stable - Within normal range";
}

Step 4: Test MCP Tool

1. Start runtime
2. Verify KPIs calculating
3. Check historian data collection
4. Test with AI assistant:
   • "What are the current production KPIs?"
   • "Show me production history for last 8 hours"
   • "Analyze the production trend"

Edge AI with ML.NET (Tutorial)

This tutorial demonstrates using ML.NET 4.0 for real-time anomaly detection on sensor data using Script Tasks.

Prerequisites:

• Complete Scripts & Data Enrichment (Tutorial)
• Tags for sensor monitoring
• ML.NET 4.0 references (included in FrameworX 10.1)

Step 1: Create Monitoring Tags

1. Navigate to Unified Namespace → Tags
2. Create tags:
   • Tag.SensorValue (Double) - Current reading
   • Tag.AnomalyScore (Double) - Detection score
   • Tag.IsAnomaly (Boolean) - Alert flag
   • Tag.Threshold (Double) - Detection threshold (default: 0.3)

Step 2: Create ML Detection Task

1. Go to Scripts → Tasks
2. Create task: AnomalyDetector
3. Set trigger: Period = 1000ms
4. Add code:

csharp

// Simple spike detection using ML.NET
using Microsoft.ML;
using Microsoft.ML.Data;

// Static ML context (initialized once)
if (@Tag.MLContext == null)
{
    @Tag.MLContext = new MLContext(seed: 0);
    @Tag.DetectionEngine = InitializeDetector();
}

// Data class for ML model
public class SensorData
{
    public float Value { get; set; }
}

public class AnomalyPrediction
{
    [VectorType(3)]
    public double[] Prediction { get; set; }
}

// Initialize detector (runs once)
private ITransformer InitializeDetector()
{
    var dataView = @Tag.MLContext.Data.LoadFromEnumerable(new List<SensorData>());
    
    var pipeline = @Tag.MLContext.Transforms
        .DetectSpikeBySsa(
            outputColumnName: "Prediction",
            inputColumnName: "Value",
            confidence: 95,
            pvalueHistoryLength: 30,
            trainingWindowSize: 90,
            seasonalityWindowSize: 30);
    
    return pipeline.Fit(dataView);
}

// Detection logic (runs every second)
var currentValue = (float)@Tag.SensorValue;

var data = new SensorData { Value = currentValue };
var prediction = @Tag.DetectionEngine.Transform(
    @Tag.MLContext.Data.LoadFromEnumerable(new[] { data }));

var result = @Tag.MLContext.Data
    .CreateEnumerable<AnomalyPrediction>(prediction, false)
    .First();

// Update tags with results
@Tag.AnomalyScore = result.Prediction[0];  // Spike score
@Tag.IsAnomaly = result.Prediction[0] > @Tag.Threshold;

// Log anomalies
if (@Tag.IsAnomaly)
{
    @Alarm.GlobalSettings.AuditTrail.AddCustomMessage(
        $"Anomaly detected: Sensor={currentValue:F2}, Score={result.Prediction[0]:F3}");
}

Step 3: Create Simple Simulator

1. Create task: SensorSimulator
2. Set trigger: Period = 500ms
3. Add simulation code:

csharp

// Simulate normal sensor data with occasional spikes
Random rand = new Random();
double baseValue = 50.0;
double noise = rand.NextDouble() * 5 - 2.5;

// Inject anomaly occasionally (5% chance)
if (rand.NextDouble() < 0.05)
{
    @Tag.SensorValue = baseValue + (rand.NextDouble() * 30 + 20);  // Spike
}
else
{
    @Tag.SensorValue = baseValue + noise;  // Normal variation
}

Step 4: Create Monitoring Display

1. Create display with:
   • Trend chart for SensorValue
   • Indicator for IsAnomaly
   • Text display for AnomalyScore
   • Threshold adjustment slider

Step 5: Test ML Detection

1. Start runtime
2. Observe sensor simulation
3. Watch for anomaly detection
4. Adjust threshold as needed
5. Check audit trail for logged anomalies

Next Steps

• Advanced MCP Tools → Complex multi-tool scenarios
• ML.NET Models → Regression and classification
• Edge Computing → Deploy to field devices

Create Value Simulator

1. With these tags created: Pressure (Integer) and AnomalyBuffer (Text Array 9 position)

2. In Devices → Protocols, select the Value Simulator and click the "New Channel..." button.

3. In Devices → Points, create points to generate simulated data.

For more information about the Value Simulator, see: Value Simulator Connector

Step 2: Create ML Anomaly Detection Script Class

1. Navigate to Scripts → Classes

2. Click the "Create a New Class" button

3. In "Import code from Library:", select AnomalyML

4. Open the script and uncomment the line that returns the detection to the AnomalyBuffer tag in Check() method.

Step 3: Create an Expression

This expression will check for anomalies each time the tag value changes.

1. Go to Scripts → Expressions

2. Create the following expression:

Where:

• <DesiredTag> is the tag you want to monitor for anomalies

Example:

Step 4: Test the System

1. Go in Runtime → “Run Startup”

2. Wait a couple minutes to have some data in the model.

3. Open the PropertyWatch

4. See the values in the AnomalyBuffer, to see the predictions.

In this section...