Smart IoT-Based Predictive Maintenance with Azure IoT & Azure Machine Learning: A Mini Project

January 27, 2025

In today's fast-paced world, minimizing downtime and maintaining operational efficiency are critical for industries. Predictive maintenance has emerged as a game-changing solution, enabling organizations to proactively address equipment issues before they lead to costly failures.

This guide takes you through building a Smart IoT-Based Predictive Maintenance System using Azure IoT Hub and Azure Machine Learning. The application collects real-time data from IoT-enabled devices, processes it for insights, and uses predictive models to foresee potential failures.

By leveraging Azure's robust ecosystem—including IoT Hub, Stream Analytics, Blob Storage, Cosmos DB, and Power BI—this project demonstrates a scalable and efficient solution for predictive maintenance.

What You'll Learn

Setting up IoT devices to send real-time data to Azure IoT Hub.
Processing and storing data using Azure Stream Analytics and storage solutions.
Training and deploying a machine learning model in Azure Machine Learning Studio for failure prediction.
Creating a seamless integration of IoT, ML, and visualization tools like Power BI.

Flow of Data

IoT devices send sensor data to Azure IoT Hub.
Data is processed in real-time by Azure Stream Analytics and stored in Blob Storage and Cosmos DB.
Azure ML Endpoint is used for failure prediction.
Azure Function orchestrates predictions and logs results.
Power BI and Azure Monitor visualize and notify based on predictions.

This guide is designed for developers, data engineers, and cloud enthusiasts who want to explore Azure’s capabilities in IoT and AI, while solving real-world industrial challenges. By the end of this guide, you'll have a working application ready to deploy and scale in production environments.

Here’s a detailed end-to-end implementation example for a Smart IoT-Based Predictive Maintenance System using Azure IoT Hub and Azure Machine Learning.

Project: IoT-Based Predictive Maintenance

1️⃣ Set Up IoT Devices to Send Data to Azure IoT Hub

1.1 Hardware Requirements

Raspberry Pi or ESP32 (or any IoT device).
Sensors: Temperature, Vibration, and Pressure.

1.2 Configure Azure IoT Hub

Create IoT Hub:
- Go to Azure Portal > Create a resource > IoT Hub > Fill in the details (e.g., Region, Pricing Tier).
Register a Device:
- In the IoT Hub, go to IoT devices > Add Device > Provide a device ID.
- Note down the Connection String for the device.

1.3 Code for Sending Data

Use Python to send data from Raspberry Pi to IoT Hub:

import random
import time
from azure.iot.device import IoTHubDeviceClient, Message

CONNECTION_STRING = "Your IoT Hub Device Connection String"
client = IoTHubDeviceClient.create_from_connection_string(CONNECTION_STRING)

while True:
    temperature = random.uniform(20, 100)
    vibration = random.uniform(0.1, 5.0)
    pressure = random.uniform(1, 10)
    message = {
        "temperature": temperature,
        "vibration": vibration,
        "pressure": pressure
    }
    client.send_message(str(message))
    print(f"Sent message: {message}")
    time.sleep(5)

2️⃣ Stream Data for Real-Time Processing

2.1 Configure Azure Stream Analytics

Create a Stream Analytics Job:
- Go to Azure Portal > Stream Analytics > Add Job.
- Configure Input as IoT Hub and Output as Azure Blob Storage or Cosmos DB.

Query to Filter Data:

SELECT
    temperature,
    vibration,
    pressure,
    System.Timestamp AS event_time
INTO
    [YourOutput]
FROM
    [YourInput]
WHERE
    temperature > 80 OR vibration > 4.0

2.2 Store Data

Blob Storage: Store raw data for further analysis.
Cosmos DB: Store structured data for querying.

3️⃣ Train a Predictive Maintenance Model

3.1 Prepare Training Data

Use historical sensor data stored in Azure Blob Storage.
Example columns: temperature, vibration, pressure, failure (1/0).

3.2 Create an Azure ML Workspace

Go to Azure Portal > Azure Machine Learning > Create Workspace.
Use Azure ML Studio to manage datasets, experiments, and models.

3.3 Build a Model in Azure ML Studio

Upload the dataset and preprocess it (e.g., normalize, handle missing values).
Use an algorithm like Random Forest or Logistic Regression.
Train the model to predict failures (failure column as target).
Deploy the trained model as an Azure ML Endpoint.

4️⃣ Deploy the ML Model and Integrate with IoT

4.1 Deploy Model as a REST Endpoint

Once trained, deploy the model in Azure ML Studio.
Copy the REST endpoint and authentication key.

4.2 Use Azure Functions for Real-Time Predictions

Create an Azure Function:
- Go to Azure Portal > Azure Functions > Create Function App.
- Choose "HTTP Trigger" template.

Azure Function Code:

import json
import requests

def main(req: func.HttpRequest) -> func.HttpResponse:
    data = req.get_json()
    temperature = data['temperature']
    vibration = data['vibration']
    pressure = data['pressure']

    # Call ML Model Endpoint
    ml_endpoint = "Your Azure ML REST Endpoint"
    headers = {"Authorization": "Bearer YourAPIKey"}
    payload = {"data": [[temperature, vibration, pressure]]}

    response = requests.post(ml_endpoint, json=payload, headers=headers)
    prediction = response.json()['prediction']

    return func.HttpResponse(f"Prediction: {prediction}")

4.3 Connect IoT Hub to Azure Function

Use Azure Event Grid to route IoT Hub messages to Azure Functions.

5️⃣ Visualize Data in Power BI or Azure Dashboard

5.1 Export Data to Power BI

Connect Power BI to Azure Blob Storage or Cosmos DB.
Create real-time dashboards to monitor temperature, vibration, and failure predictions.

5.2 Create Alerts

Set up alerts in Power BI or Azure Monitor to notify maintenance teams about predicted failures.

6️⃣ Test and Run

Simulate sensor data with high temperature or vibration values.
Verify predictions and alerts in the dashboard.
Ensure the ML model provides accurate results and adjust it as needed.

This project covers the complete lifecycle from IoT data ingestion to ML-based predictions and real-time monitoring.

Search This Blog

DevSyncra