Design a Temperature Control System in a Fuzzy Logic-Based Baby Incubator

Abstract

Premature infants require a stable temperature of 33–35°C and a humidity of 60–80% RH, but conventional manual incubators have low accuracy (±2–5°C) and are less adaptive to environmental changes. This study designed a temperature and humidity control system for a baby incubator based on an IoT-integrated Adaptive Neuro-Fuzzy Inference System (ANFIS), using an ESP32 microcontroller, DHT22 sensor, as well as a PTC heater actuator, Peltier module, and ultrasonic humidifier. Two ANFIS Sugeno order-1 models are designed for heating and cooling control, each with 2 inputs, 3 Gaussian membership functions, and 9 fuzzy rules of hybrid learning training. The system features manual mode as well as a Firebase-based web dashboard for real-time monitoring. The test results showed the accuracy of the DHT22 sensor with an average error of 2.39%, as well as a first-order thermal incubator model with a dead time of 80 seconds and a time constant of 473.4 seconds. In setpoint tests of 33–35°C, the ANFIS mode is more responsive to the target temperature than the manual mode which tends to be stable below the setpoint, although it results in greater temperature and humidity oscillations. The system is also able to recover the temperature after being given an external interference. These results show that ANFIS improves control responsiveness compared to manual control, but still requires refinement of training data and actuation strategies to suppress oscillations.