| 제목 |
다층 신경망을 이용한 보행자 머리 충격 량 예측에 관한 연구 |
| 분야 |
차체 및 구조 안전 |
| 언어 |
Korean |
| 저자 |
김현석(순천향대학교), 박민규(순천향대학교), 박성근(순천향대학교), 이신선(현대자동차), 이태희(현대자동차) |
| Key Words |
Head injury criterion(머리 상해도), Pedestrian safety(보행자 안전), Passive safety(수동 안전), NCAP(신차 안전도 평가프로그램), Impact prediction(충격 량 예측), Multi-layer perceptron(다층 퍼셉트론) |
| 초록 |
According to statistics on fatalities in traffic accidents, about 38 % of all deaths were pedestrian related. For this reason, car manufacturers are conducting various research and commercialization from the perspective of both active safe and passive safe technology in order to prevent traffic accidents and minimize damage in the event of an accident. During the stage of vehicle development, vehicles are being designed in order to ensure the safety of pedestrians. However, in the current development phase, methods are in place to predict the number of pedestrian injuries after the prototype has been completed or to predict the number of injuries by means of a design-based mechanical analysis. This study proposes a method for predicting the acceleration of the standard value of pedestrian head injury using vehicle design data, impact volume test values, and multi-layer neural network. Vehicle design files are encoded in image form, and HIC raw values are predicted by using encoded data and multi-layer perceptron. The raw data is in the form of time series data. The models were trained and the results were predicted by using a parallel multi-layer perceptron model in order to predict the impact amount for each time. The results showed an error rate of 21.6 based on the RMSE. This would allow us to identify the safety of the vehicle from the design stage in advance, thereby enhancing efficiency and productivity during the development process. |
| 원문(PDF) |
다운로드
Journal Site
|
