Transfer path analysis quantifies how much each load path contributes to a target response in vehicle ride comfort development. Establishing how that contribution shifts with a suspension parameter, however, requires a fresh time-domain solution for every design. This study formulates a seven-degree-of-freedom full-car lumped-mass model in state space and trains a long short-term memory (LSTM) surrogate that predicts the response from the road displacement and the stiffness scale factors. The vertical acceleration at the seat is a constant weighted sum of nine path forces whose coefficients do not depend on stiffness, so the contributions follow from the predicted state by algebra. Four stiffnesses were swept for the training data. The median state prediction error was 4.54 % inside the training range and 8.77 % outside it. Across the swept specifications the projection contribution ratio of the most sensitive path travels 11.8 %p, which the surrogate reproduces with a correlation of at least 0.973 and a mean absolute error of at most 0.36 %p.

Abstract
1. 서 론
2. Full car lumped mass 모델
3. 시간영역 전달경로 기여도
    3.1 경로 하중과 전달계수
    3.2 기여도 지표
4. LSTM 기반 응답 예측
5. 결과 및 고찰
    5.1 기준 모델의 기여도 특성
    5.2 강성 변화 시 응답 예측 및 기여도 계산
6. 결 론
후 기
References

• 원지훈(Hankyong national University, Mechanical Engineering) | Ji Hun won
• 윤병조(Hankyong national University, Mechanical Engineering) | Byeong Jo Yoon
• 정운창(Hankyong national University, Mechanical Engineering) | Un-Chang Jeong Corresponding author