Enhancing Safety and Efficiency in Firefighting Operations via Deep Learning and Temperature Forecasting Modeling in...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Predictive Modeling

Background:

• Firefighters face significant risks from extreme temperatures, smoke, and structural instability in burning buildings.
• Accurate real-time data is crucial for informed decision-making, improving safety and reducing casualties.

Purpose of the Study:

• To develop and evaluate unsupervised deep learning (DL) for classifying fire site danger levels.
• To implement autoregressive integrated moving average (ARIMA) models for predicting temperature changes at various heights in burning structures.

Main Methods:

• Unsupervised deep learning (DL) autoencoder artificial neural network (AE-ANN) for danger classification.
• Autoregressive integrated moving average (ARIMA) combined with random forest regressor for temperature prediction.
• Analysis of temperature changes from 0.6m to 2.6m height and over time.

Main Results:

• The AE-ANN model achieved an accuracy of 0.869 for classification.
• The ARIMA model demonstrated remarkable predictive performance for temperature change trends.
• The random forest regressor and ARIMA models were evaluated using an open-source dataset.

Conclusions:

• Deep learning and predictive modeling show potential for enhancing firefighter safety and decision-making.
• ARIMA models are effective for predicting temperature progression in fire incidents.
• This research contributes novel methods for fire site analysis using available data.