Affiliation:
1. Sichuan University of Science & Engineering, School of M
2. Yibin Cowin Automobile Co., Ltd., Automotive Engineering Res
3. Chongqing University, School of Mechanical and Transportatio
Abstract
<div class="section abstract"><div class="htmlview paragraph">The Mobile Progressive Deformable Barrier (MPDB) is a standardized automotive
crash scenario that comprehensively evaluates the safety of battery-electric
vehicles (BEVs) in a crash. In an accident, the deformation pattern of the Front
of Battery Electric Vehicle (FOBEV) structure, the efficiency of energy
absorption, the acceleration pulse, and the degree of intrusion into the
passenger compartment combine to affect the safety of the driver and passengers.
In order to simulate and calculate the damage state of FOBEV in MPDB more
efficiently and to construct a collision damage dataset in the entire velocity
domain, a FOBEV equivalent model is proposed. The acceleration pulses from
numerical simulations and impact tests were compared to verify the model’s
validity. On this basis, the prediction accuracies of the Support Vector Machine
model (SVM), Gaussian Process Regression model (GPR), and BP neural network
model (BP) in FOBEV collision events are compared and analyzed, and BP is taken
as the most suitable model and further improved. Taking a BEV under development
as an example, the application of the accident damage prediction method based on
the FOBEV equivalent model in the optimal design of BEV crashworthiness is
illustrated. The results show that the constructed FOBEV equivalent model
exhibits high consistency in the impact test. The accuracy of the improved
Tent-SSA BP model increased by 34.85%. The neural network prediction technique
with multiple input parameters is used to study the crash damage of FOBEVs over
the entire speed range, revealing the relationship between the parameters of
FOBEVs on the crashworthiness of BEVs in highly nonlinearly varying crashes.</div></div>