|
Citation
|
Kawabuchi T, Dokko Y, Mikami H, Katsushima K, Nagai Y. 27th International Technical Conference on the Enhanced Safety of Vehicles (ESV); April 3-6, 2023; Abstract #: 23-0219, pp. 15p. Washington, DC USA: US National Highway Traffic Safety Administration, 2023 open access. |
|
Abstract
|
27th International Technical Conference on the Enhanced Safety of Vehicles (ESV): Enhanced and Equitable Vehicle Safety for All: Toward the Next 50 Years
https://www-esv.nhtsa.dot.gov/Proceedings/27/27ESV-000219.pdf
The fatality rate of thoracic injury for elderly occupants in vehicle accidents is significantly high. Its major cause is the rise of internal organ injury rates due to an increase in the number of fractured ribs (NFR). Therefore, NFR reduction is crucial to enhance elderly occupant protection and is one of the key issues for achieving zero fatalities. In order to improve NFR prediction accuracy, the previous study proposed the criterion using the weighted averaged displacement of all ribs (WADAR), which indicated a higher correlation coefficient with NFR than that of the criterion, Rmax, using four Infra-Red Telescoping Rod for the Assessment of Chest Compression (IR-TRACC) installed on the thorax of the Test device for Human Occupant Restraint Anthropometric Test Dummy (THORATD). While WADAR requires all rib deflections, it is difficult to install IR-TRACCs on all ribs inside the limited space in the thorax of THOR-ATD. The objective of this research is to predict the deflections of all ribs by means of a neural network model using time-histories of rib deflections from four IR-TRACCs and the crash velocity without any installation of additional measurement devices. The architecture of the neural network model is based on U-Net, which is one of the convolutional neural network models. The model was trained by time-historical X, Y and Z displacements of 14 ribs and the crash velocity derived from the 56 FEM simulation data, which represented frontal and oblique sled experiments with THORATD. The model learned the physical relationships among the ribs with and without IR-TRACCs. The predicted rib deflections were validated by the THOR-ATD experiment, where the displacements of the 2nd to 6th ribs on the left side were measured three-dimensionally by the set of two cameras installed on the upper and lower thoracic spines. The predicted deflections during 0 to 150 ms were processed into a resultant deflection and compared to the actual deflection through the 2nd to 6th ribs on the left side. The maximum differences in the peak deflection were 2.3 mm, respectively. Furthermore, the root mean square error (RMSE) was calculated at each rib for prediction accuracy evaluation, which resulted in minimum and maximum RMSE of 0.6 mm and 2.7 mm, respectively. Although the number of training datasets was small, the neural network model trained by FEM simulation data could predict all the rib deflections with small error without physical measurement devices.
Language: en |