Related Experiment Videos
A Bayesian Neural Network approach to estimating the Energy Equivalent Speed
C Riviere1, P Lauret, J F Manicom Ramsamy
1Université de La Réunion, Laboratoire de Génie Industriel, Equipe Génie Civil et Thermique de l'Habitat, 15 avenue René Cassin, BP 7151, 97705 Saint-Denis Cedex, Ile de la Réunion, France. carine.riviere@uni-reunion.fr
Accident; Analysis and Prevention
|November 1, 2005
Summary
Accurate vehicle accident reconstruction requires estimating deformation energy using the Energy Equivalent Speed (EES). This study introduces a novel Bayesian Neural Network model to precisely estimate EES, improving accident analysis.
Area of Science:
- Road Safety
- Computational Mechanics
- Machine Learning
Background:
- Accurate accident reconstruction is crucial for reducing road accidents.
- Estimating vehicle deformation energy during impact is essential for accident modeling.
- Existing tools for deformation energy estimation lack precision and power.
Purpose of the Study:
- To develop a more precise model for estimating the Energy Equivalent Speed (EES).
- To express vehicle deformation energy as a function of EES.
- To address the challenge of estimating EES due to its dependence on vehicle and impact parameters.
Main Methods:
- Development of a novel model combining Bayesian and Neural Network approaches.
- Utilizing the complementary strengths of Bayesian and Neural Network methods.
- Implementing error bars for computed outputs and ensuring optimal model discovery.
Main Results:
- The developed Bayesian Neural Network model accurately estimates the EES of a car.
- The model achieved a mean error of 1.34 m/s in EES estimation.
- A sensitivity analysis was conducted to evaluate the relevance of the model's inputs.
Conclusions:
- The proposed Bayesian Neural Network approach offers a powerful and precise method for EES modeling.
- This model enhances the accuracy of vehicle accident reconstruction and analysis.
- The findings contribute to the development of more sophisticated accident analysis tools.