Autoignition temperature: comprehensive data analysis and predictive models.

Area of Science:

• Computational chemistry and materials science.
• Chemical engineering and process safety.

Background:

• Autoignition temperature (AIT) is crucial for assessing fire and explosion hazards of combustible materials.
• Accurate AIT prediction is essential for safe handling, storage, and processing of chemicals.

Purpose of the Study:

• To develop a reliable predictive model for autoignition temperature (AIT).
• To identify a robust dataset for training and validating the AIT predictive model.
• To enhance the safety assessment of combustible materials through accurate AIT prediction.

Main Methods:

• Critical analysis of available structure-AIT data from diverse sources.
• Development and evaluation of Support Vector Regression (SVR) models on selected data subsets.
• Selection of a final dataset comprising 875 compounds with annotated AIT values.

Main Results:

• The SVR model achieved a determination coefficient (r²) of 0.77 and a Mean Absolute Error (MAE) of 37.8°C in cross-validation.
• External validation on 20 industrial compounds demonstrated good predictive power with an MAE of 28.7°C.
• The developed model shows reliable performance in predicting AIT for a wide range of compounds.

Conclusions:

• The developed SVR model provides a reliable method for predicting autoignition temperature.
• The model's accuracy in both cross-validation and external validation supports its utility in safety hazard assessments.
• This predictive capability contributes to improved safety protocols in the chemical industry.