Using Machine Learning to Classify Danger Levels and Prediction Techniques to Forecast Temperature Changes with Height in Burning Sites

Firefighters go into burning structures to rescue trapped victims and extinguish the fire as soon as possible. Factors such as extreme temperatures, smoke, toxic gases, explosions, and falling objects inhibit their efficiency and risk their safety. These factors could change within a twinkle of an eye. Firefighters must be provided with accurate information and data about the burning site. They can make informed decisions about their duties and know when it is safe to enter and evacuate to reduce casualties. This research presents Machine Learning (ML) algorithms for classifying the danger levels and prediction models to forecast temperature changes with height and time in burning sites using environmental factors such as temperature, smoke, and carbon monoxide, CO. The classifier algorithms bring the chief firefighter’s awareness of the danger levels in the burning compartment. The predictor models forecast the rise in temperature from 0.6m to 2.6m and the changes in temperature with time at 2.6m. Knowing the temperature at 2.6m is essential because the temperature rises faster with height and the elevated temperature at this height weakens the building's structural members. We investigated four techniques for classification and used data analysis prediction methods to predict temperature parameters. Classification methods covered the support vector machine (SVM), logistic regression (LR), k-nearest neighbors (k-NN), and autoencoderartificial neural network (AE-ANN). The prediction data analytical approach covered ARIMA with random forest regression implementation. Analysis showed that AE-ANN performed poorly in classifying the latent representation of our custom environmental data into dangerous levels, and SVM outperformed the other ML models. The ARIMA model excellently predicts the trend of temperature changes in the burning site. Still, the errors are a bit high due to significant variations in the decay phase of fire development stages exhibited by different fire grounds.

Name thesis:degree_name

Master of Science

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Engineering

Grantor

Texas State University

Year dc:date.issued

2022

Author dc:creator

• Ishola, Adenrele

Advisor dc:contributor.advisor

• Valles, Damian

Committee members dc:contributor.committeemember

• Stapleton, William
• Aslan, Semih

• firefighting
• machine learning
• classifications
• predictions
• SVM
• logistic regression
• KNN
• ARIMA

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10877/20003

OAI identifier oai:identifier

oai:digital.library.txst.edu:10877/20003