Akademik Veri Yönetim Sistemi
Mechanics Based Design of Structures and Machines, cilt.54, sa.1, 2026 (SCI-Expanded, Scopus)
Train operation safety has gained significant attention in recent years. Due to the large mass and high operation velocity, train collision accidents often result in catastrophic damage. Current train crashworthiness design focuses on fixed impact velocity by unchangeable energy absorption structures. However, the real train impact velocity is unpredictable and can vary significantly, thus the current train crashworthiness design strategy cannot achieve the optimal crashworthiness under various impact velocities. To address this limitation, this work proposes a real-time adaptive optimization method of energy dissipation in a train collision. First, a one-dimensional train collision model is established based on the multibody dynamic theory. The dynamic response and energy distribution during train collision are analyzed. Then, by adjusting the crushing force of each vehicle, the optimal crushing forces of the energy-absorbing device for the given impact velocity are obtained by the genetic algorithm, which regards the maximum average moving acceleration as the objective response. Results show that the safe collision velocity of the train is increased from 17 m/s to 27 m/s without changing the stroke of the energy-absorbing device. A back-propagation neural network is introduced and trained, which regards the velocities of two trains as input indicators and the optimal crashworthiness indicators of the energy-absorbing device as output indicators. Finally, a real-time prediction framework is established, which can predict the optimal crashworthiness indicators during train collisions under different impact velocities in real time. The results show that the maximum error of the maximum moving average acceleration between the optimization and prediction results is below 2%, which proves the accuracy of the prediction framework.