Predicting The Detection Delay Of An ESP32-Based Fire Detection System Using Flame Sensor, MQ-2 Sensor, And Multiple Linear Regression.
TL;DR
Findings indicate that the proposed Multiple Linear Regression model is effective for predicting the detection delay of an IoT-based fire detection system and accurately estimates the system response time.
Abstract
Fire is one of the most common non-natural disasters that can cause significant material losses and threaten human safety. Therefore, an effective fire detection system is required to provide early warnings and minimize potential risks. This study proposes an Internet of Things (IoT)-based fire detection system integrated with a Flame Sensor, an MQ-2 gas sensor, an ESP32 microcontroller, the Blynk platform, and Telegram notifications for real-time monitoring and early warning. A Multiple Linear Regression model was developed to predict the system detection delay. The independent variables used in this study were Fire Detection Intensity (X₁) and MQ-2 Sensor Value (X₂), while the dependent variable was Detection Delay (Y). A total of 500 experimental data records were collected through the Blynk platform, consisting of 350 training data and 150 testing data. The proposed model achieved an MAE of 0.050535 s, an MSE of 0.003343, an RMSE of 0.057822 s, a MAPE of 3.16%, and an R² value of 0.988871, indicating excellent prediction performance. The comparison between the actual and predicted detection delays also demonstrates that the proposed model accurately estimates the system response time. These findings indicate that the proposed Multiple Linear Regression model is effective for predicting the detection delay of an IoT-based fire detection system.