An energy absorbing tube combining multi-corner and multi-cell configurations was designed in this study. Machine learning was adopted to predict and optimize the crashworthiness of the proposed tube because it can handle both numerical and categorical responses. The results showed the increases in the considered geometric parameters caused the increases in the specific energy absorption and peak crushing force, while also made the unstable deformation mode prone to appear. Besides, with the help of machine learning, the accurate optimization results were obtained, in which the unstable deformation was removed. This work highlights the prospect of machine learning in structural optimizations.

Hybrid structures have the advantage of combining different types of materials at the same time. The trend of lightweight design in the transportation industry has promoted the development and application of composite materials with good crashworthiness performance. Low-density crushable foam-filled metal-composite hybrid structures have potential advantages as energy-absorbing components. This study investigated the mechanical characteristics of four different polyurethane foam-filled hybrid structures and their individual components under quasi-static axial compression. The experimental results showed foam-filled hybrid structures could change the deformation mode and improve stability during the compression process. Meanwhile, these hybrid structures could also improve energy absorption compared with their individual components. Among the different configurations, specimen C-PU-C (i.e., polyurethane foam filler between an outer CFRP tube and an inner CFRP tube) had the highest energy absorption capacity, at 5.4 kJ, and specific energy absorption, at 37.3 kJ/kg. Finally, a finite element (FE) model was established to analyze the mechanical characteristics of the hybrid structures by validating the simulation results against the experimental results.

When a train crashes with another train at a high speed, it will lead to significant financial losses and societal costs. Carrying out a train-to-train crash test is of great significance to reproducing the collision response and assessing the safety performance of trains. To ensure the testability and safety of the train collision test, it is necessary to analyze and predict the dynamic behavior of the train in the whole test process before the test. This paper presents a study of the dynamic response of the train in each test stage during the train-to-train crash test under different conditions. In this study, a 1D/3D co-simulation dynamics model of the train under various load conditions of driving, collision and braking has been established based on the MotionView dynamic simulation software. The accuracy of the numerical model is verified by comparing with a five-vehicle formations train-to-train crash test data. Sensitivities of several key influencing parameters, such as the train formation, impact velocity and the vehicle mass, are reported in detail as well. The results show that the increase in the impact velocity has an increasing effect on the movement displacement of the vehicle in each process. However, increasing the vehicle mass and train formation has almost no effect on the running displacement of the braking process of the traction train. By sorting the variables in descending order of sensitivity, it can be obtained that impact speed > train formation > vehicle mass. The polynomial response surface method (PRSM) is used to construct the fitting relationship between the parameters and the responses.

In order to improve crashworthiness, the optimum energy-absorption design of locomotives was performed in MADYMO multibody dynamics software, and a three-dimensional computational model for locomotive vehicles was built. The results of calculations using the model were then fitted by multiple linear-regression functions, and a scaled equivalent model test was carried out to validate the fitting formula results. The results showed how to influence the energy absorption of a locomotive by studying the mass and velocity of the locomotive, as well as the plastic deformation platform force of the middle car's crushing tube. The fitting formula related to the plastic deformation platform force of the crushing of the middle car's tube and the kinetic energy of the locomotive was also obtained. The fitting accuracy of the model was 99.98%, and the error between the calculation results from the fitting formula and the scaled equivalent model test results was less than 5%. These results reveal the mode of energy dissipation, and they can be applied to the preliminary conceptual design of train. The dynamic tests of prototype are essential before serial production.