Pedestrian Safety Performance Prediction Using Machine-Learning Techniques Tata Motors Limited

Format:

Book

Conference/Event

Author/Creator:

Kaushik, Bharat, author.

Contributor:

Bera, Satadru

Daphal, Pratap

Khare, Pratyush

Koralla, Sivaprasad

Conference Name:

Symposium on International Automotive Technology (2021-09-29 : Pune, India)

Language:

English

Physical Description:

1 online resource cm

Place of Publication:

Warrendale, PA SAE International 2021

Summary:

As per WHO 2018 report, pedestrian fatalities accounts for 23% of world road accident fatalities. Every day 850 pedestrians lose their lives in the world. As per MoRTH 2018 report, 16% of road accident fatalities are of pedestrians in India. Everyday 64 pedestrians lose their lives in India.Based on accident data, one of the most common reason for the pedestrian fatality is head injury due to primary contact from vehicle front-end structure. Pedestrian head injury performance is highly depended on front-end styling, bonnet stiffness, clearance with aggregates underneath the bonnet and hard contact points.During concept stage of vehicle development, safety recommendation on front-end design is provided based on geometric assessment of the class A surface, as the maturity of the geometry is insufficient to carryout CAE simulations. This paper presents the novel approach of using machine-learning algorithms to predict the head injury performance at the early stage of the vehicle design using the knowledge of existing vehicle simulation data and new vehicle design features. Machine learning based mathematical model developed considering critical design parameters such as clearance with aggregates, Impact point location with respect to hard points, stiffness of bonnet as an input variables and head injury criteria (HIC) as an output variable from existing vehicles. Different supervised machine learning algorithms such as random forest, neural networks, logistic regression and supporting vector regression has trained and tested using available data. Selected the suitable mathematical model based on the model score. Identified model was able to predict the pedestrian head injury criteria (HIC) within 15% of margin of error. This approach has significant potential and provides opportunities for giving directional feedback at early stage of the vehicle programs

Notes:

Vendor supplied data

Publisher Number:

2021-26-0026

Access Restriction:

Restricted for use by site license