AI Predictive Maintenance for Smart Factories

Unexpected equipment stoppages can interrupt production, increase maintenance costs and reduce manufacturing consistency. This paper presents a low-cost Internet of Things (IoT) prototype for machine-condition monitoring as a foundation for predictive maintenance in smart factories. The prototype uses an ESP32 microcontroller connected to a DHT22 temperature-and-humidity sensor, an SW-420 digital vibration detector and an ACS712 current-sensor module. A low-voltage DC motor is driven through an L298N module, while LEDs and a buzzer provide local threshold-based indications. The ESP32 periodically sends a JSON payload over Wi-Fi to a FastAPI service hosted on a Mac computer. The backend exposes endpoints for receiving the latest sensor data and retrieving historical records, which are stored in an SQLite database. A Streamlit dashboard displays the latest measurements and supports historical visualization. During implementation, sensor readings were visible through the serial monitor, the HTTP POST request returned status code 200, and the history endpoint returned stored records. These observations demonstrate the operation of the core data pipeline. The current prototype performs monitoring and rule-based alerts; a trained and validated artificial-intelligence model has not yet been implemented. In addition, the current field from the ACS712 represents a raw analogue-to-digital converter (ADC) reading rather than calibrated amperes. Future work includes sensor calibration, longer-term data collection, fault-labelled experiments, anomaly detection, and validation of machine-learning methods against documented operating conditions.