Smart Temperature Control Using Neuro-Fuzzy Model
Kelvin N. Nnamani1, Ken Aghaegbunam Akpado2, Augustine C.O. Azubogu3
1Kelvin N. Nnamani, Department of Electronic and Computer Engineering, Nnamdi Azikiwe University, Awka (Anambra), Nigeria.
2Prof. Ken Aghaegbunam Akpado, Department of Electronic and Computer Engineering, Nnamdi Azikiwe University, Awka (Anambra), Nigeria.
3Prof. Augustine C.O. Azubogu, Department of Electronic and Computer Engineering, Nnamdi Azikiwe University, Awka (Anambra), Nigeria.
Manuscript received on 30 September 2025 | Revised Manuscript received on 11 October 2025 | Manuscript Accepted on 15 October 2025 | Manuscript published on 30 October 2025 | PP: 4-9 | Volume-5 Issue-6, October 2025 | Retrieval Number: 100.1/ijainn.F110705061025 | DOI: 10.54105/ijainn.F1107.05061025
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Published by Lattice Science Publication (LSP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The temperature control model employs a neuro-fuzzy approach with a defined universe of discourse encompassing temperature (20℃-50℃), humidity (30%-90%), and fan speed (20%-70%). Membership functions were established, utilizing generalized bell functions for temperature and humidity, along with trapezoidal functions for fan speed. A rule base comprising nine rules was developed, incorporating temperature and humidity as linguistic input variables and fan speed as the linguistic output variable. In the data preprocessing phase using Python, 60% of the dataset was designated for training, while 40% was set aside for testing with the scikit-learn model. A convolutional neural network (CNN) was created using TensorFlow’s Keras API, featuring 64 neurons, ReLU activation, and two input shape features. The model underwent training for 100 epochs with the Adam optimizer and a batch size of 16, achieving a training loss of 0.9951 and a test loss of 1.0239. The closely matched and relatively low values of both training and test loss indicate that the model is not overfitting and has successfully captured the underlying patterns. For instance, when the current temperature and humidity were set to 35℃ and 65%, the recommended fan speed was 48%. Moreover, predicted fan speeds were 20.14%, 35.21%, and 43.64% for temperature and humidity settings of (35℃, 45%), (45℃, 75%), and (55℃, 85%), respectively.
Keywords: Temperature, Humidity, Fan Speed, Gbell Membership Function, Trapezoidal Membership Function, TensorFlow’s Keras API, Scikit-learn, Convolutional Neural Network (CNN).
Scope of the Article: Neural Networks