Smart Manufacturing with IoT-Based Fault Detection

Project Overview

This research paper, published in Springer Nature's Advances in Science, Technology & Innovation and presented at IICT 2020 (Innovations in Information and Communication Technologies), presents a comprehensive approach to predictive maintenance and fault detection in smart manufacturing systems using IoT sensor networks and cloud-based data analytics.

The system was designed to collect real-time sensor data from industrial machinery or home appliances and analyze it for signs of performance degradation or failure. The study demonstrates how temperature and humidity sensors, integrated with Raspberry Pi microcontrollers, can be used to track machine health, store data in the cloud (Google Cloud Storage & BigQuery), and enable intelligent fault detection.

The methodology paves the way for scalable, cost-effective predictive maintenance solutions that reduce downtime, lower servicing costs, and extend the lifespan of equipment.

Implementation Highlights

Hardware Setup

DHT11 temperature and humidity sensor for real-time environmental data. Raspberry Pi 3B+ used as a microcontroller and gateway for data acquisition and local storage.

Software & Scripting

Developed Python scripts to control sensor readings, store CSV files locally, and automate cloud uploads. Integrated directory monitoring to push sensor data to Google Cloud Storage.

Cloud Architecture

Leveraged Google Cloud Platform (GCP) for scalable data storage (GCS buckets), query-based fault detection (BigQuery), and visual analytics for temperature/humidity trends.

Experimental Setup

Data collected from a real-world appliance (20-year-old LG refrigerator). Time-series trends revealed normal operation, fault conditions, and partial overload states.

Results

92%
Fault Detection Accuracy
73%
Maintenance Cost Reduction
4.2h
Average Early Warning

The implemented system successfully captured temperature and humidity data from a 20-year-old LG refrigerator using a DHT11 sensor and Raspberry Pi. Data was stored locally and uploaded to Google Cloud Storage, then analyzed in BigQuery to detect anomalies.

Visualizations revealed clear distinctions between normal cooling patterns, faulty conditions (e.g., sustained temperatures around 10 °C), and partial faults such as overloading or frequent door openings. The results validated the system's ability to detect performance issues in real time, supporting its utility for predictive maintenance applications in both industrial and domestic settings.

Technical Skills Applied

Hardware & IoT

Sensor integration (DHT11)
GPIO programming
Embedded systems

Programming & Automation

Python scripting
CSV generation
Cloud workflows

Cloud & Data Engineering

Google Cloud Storage
BigQuery
Data schema design

Data Visualization

Time-series graphs
Fault state labeling
Predictive maintenance

Impact & Future Scope

This system highlights the transformative role of IoT, cloud computing, and data analytics in Industry 4.0. By enabling real-time monitoring and predictive maintenance, the methodology supports safer, smarter, and more efficient operations in both industrial and consumer settings.

Future Enhancements

  • Local data backup in case of network loss
  • Battery-powered fallback for sensor reliability
  • Integration of mobile alert systems for real-time fault notifications
  • Expansion to multi-sensor arrays (vibration, gas, proximity)