Prediction model for chemical explosion consequences via multimodal feature fusion

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Risk Assessment

Background:

• Chemical explosions pose significant risks to safety and the environment.
• Accurate prediction of these incidents is crucial for risk mitigation in the chemical industry.

Purpose of the Study:

• To develop an innovative model for predicting chemical explosion consequences.
• To integrate Quantitative Structure-Property Relationship (QSPR) and Quantitative Property-Consequence Relationship (QPCR) principles for enhanced risk assessment.

Main Methods:

• Utilized a Bayes-Transformer-SVM model with multimodal feature fusion.
• Input features included molecular descriptors from SMILES and Gaussian16, plus leakage condition parameters.
• Employed a multimodal fusion theoretical framework and an optimized Transformer-SVM architecture.

Main Results:

• The optimized Bayes-Transformer-SVM model demonstrated superior performance.
• Achieved high accuracy with R² of 0.9475 and RMSE of 0.1139 on the test set.
• Outperformed alternative prediction models in assessing explosion consequences.

Conclusions:

• The developed model offers a novel and effective approach for explosion risk assessment.
• Enables rapid assessment for chemical storage and transport, supporting safety-by-design.
• Facilitates prospective risk assessment for new and existing chemical substances.