jomit/anomaly-detection

Serverless Anomaly Detection using Cognitive Services

Serverless Anomaly Detection using Azure

High Level Components

• Simulated PLC

  • Generates sensor reading data and sends it to IoT Hub

• Function App

  • Triggers when IoT Hub receives a new message
  • Sends the message to Anomaly Detector API (Cognitive Services)
  • If Anomaly is detected, calls a Logic App

• Logic App

  • Fetches the email content from query string
  • Sends the email using Send Grid

• (Optional) Real time Dashboard

High Level Deployment Steps

IoT Hub

• Create a new IoT Hub Device and copy the connection string.

Logic App

• Create a new Send Grid Account.
• Create a logic app using logicapp\template.json as guidance

Anomaly Detector API

• Create the Anomaly Detector cognitive service resource in Azure.
• Copy the key and endpoint from the Quick Start section.

Function App

• Create a NodeJS Function App.
• Add following Application Settings:
  • Name: anomalydetector_endpoint | Value: <Anomaly Detector API key>
  • Name: anomalydetector_key | Value: <Anomaly Detector API endpoint>
• Create a new Function with IoT Hub as trigger and connect it to the IoT Hub created above.
• Use the function Console to install following npm packages in D:\home\site\wwwroot directory:
  • npm install @azure/cognitiveservices-anomalydetector
  • npm install @azure/ms-rest-js
• Update the index.js file code with function\index.js
• Update the logicAppWorkflowPath variable in sendEmail method with the logic app workflow path

Simulated PLC

• Clone the repo
• Install Nodejs
• Goto simulated-plc folder in command line
• Run npm install
• Create .env file and add following variables:
  • deviceConnectionString='<Your IoT Hub Device Connection String>'
• Run node .