To enhance automobile fuel efficiency, it is crucial to develop lightweight and compact components with improved strength. This research specifically addresses the increasing demand for higher fatigue strength in components that undergo surface hardening processes. The study focuses on the impact of refining the prior austenite (prior γ) grain size on the fatigue strength of low carbon steel sheets that have been processed through vacuum carburizing followed by quenching. To investigate this relationship, specimens were prepared under various quenching temperatures, and their plane bending fatigue strength was evaluated. The results indicated a clear trend: finer prior γ grain sizes generally correlate with improved fatigue strength. However, it was observed that when the maximum grain size fell below 47 μm, the enhancement in fatigue strength became minimal. Further analysis of the fracture surfaces revealed a significant shift in the initiation point of fatigue failure, transitioning from intergranular to transgranular fracture. This shift is crucial as it contributes to the overall fatigue behavior of the materials. The findings of this study underscore the importance of achieving a sufficiently fine prior γ grain size in vacuum carburized and quenched materials. Such refinement is essential for attaining stable and high fatigue strength, as it facilitates transgranular fracture initiation, ultimately leading to enhanced durability and reliability of automotive components.

Generally, the aerodynamic force coefficients of railway vehicles are determined through wind tunnel tests that simulate the interaction between the vehicle and track structures under turbulent boundary layer flow conditions. In most conventional crosswind wind tunnel tests, aerodynamic forces are measured on a stationary vehicle model placed on representative track structures. In actual operation, however, the relative wind direction acting on a train may vary owing to changes in natural wind direction, train acceleration or deceleration, or passage through curved track sections. To evaluate aerodynamic forces under such varying wind-direction conditions, this study measured the surface pressure distribution on a vehicle model mounted on a rotating turntable. Wind tunnel tests were carried out in the large low-noise wind tunnel of the Railway Technical Research Institute using a 1/40-scale model of a typical commuter train placed on a low-embankment structure. A turbulent boundary layer representing natural wind was generated upstream of the turntable. Surface pressures measured on the vehicle body were used to calculate the side force and lift force coefficients. The results obtained under rotational conditions, in which the relative wind direction changed continuously from 0-90°, were compared with those under stationary conditions corresponding to conventional tests. The results showed that no clear difference was observed in the side-force coefficient between stationary and rotational conditions. In contrast, the lift-force coefficient under rotational conditions differed clearly from that under stationary conditions, and its tendency also depended on the direction of rotation. Although the maximum difference in lift-force coefficient was notable, its influence on the critical wind speed for overturning was considered limited because lift force contributes less to overturning safety than side force. These findings suggest that wind-direction changes can affect the aerodynamic characteristics of railway vehicles.

In power generation using vibration energy to rotate an electro-magnetic coil, we encounter the problem that the force due to the coil current suppresses the coil rotation, vibration and power generation. This paper studies a method to avoid this problem by noting that the vibration of a rotor is promoted by internal damping. This method uses an electro-magnetic coil system such that it produces internal damping for the rotor vibration. To suppress the unstable vibration caused by the internal damping, another electro-magnetic coil system is applied to supply external damping. By increasing the external damping according to the increase in the internal damping, we can augment the powers produced by the two kinds of damping without raising the vibration amplitude (structural load).

The use of GPUs in large-scale technical and scientific computing is spreading, and several libraries for distributed parallel sparse linear equation solvers are GPU-compatible. While these libraries process a system consisting of a single coefficient matrix and right-hand-side (RHS) vector, in fields such as structural analysis based on the finite element method (FEM), the analysis mesh itself is divided into multiple subdomains and matrices are constructed independently in each subdomain. This represents a structural mismatch with the input format assumed by general-purpose direct solver libraries. In this study, a direct linear equation solver using NVIDIA CUDA libraries was newly implemented as an extension to FrontISTR, a parallel finite element structural analysis program. The objective is to enable solving linear equations based on domain decomposition by the direct method, which is a robust solution method applicable to a wide range of analyses. For sequential execution targeting models that fit within a single GPU’s memory, the implementation produced solutions with very small relative residuals on NVIDIA A100 GPUs, achieving speedups of approximately 7 to 25 times compared with CPU execution of the same routine. For larger models requiring distributed parallel execution across multiple GPUs, a parallel algorithm based on the multiplicative Schwarz procedure combined with CUDA-aware MPI was designed and implemented, and numerical experiments confirmed the algorithmic correctness of the iteration, although its wall-clock performance does not yet surpass that of FrontISTR’s CPU-based parallel direct solver. The presented framework is therefore positioned as an alternative GPU-compatible solution pathway that operates directly on the domain-decomposed layout, using a single-GPU sparse-direct routine as a per-subdomain local solver, for models that exceed the memory of a single GPU.

In automotive torsional damper testing systems, nonlinear characteristics such as dead zones and multi-stage stiffness can cause torque overshoot, which degrades evaluation accuracy and may cause mechanical damage. Conventional feedforward one-pulse command-shaping methods suppress overshoot by applying a predesigned reference-shaping signal; however, parameter retuning is required when torsional characteristics vary. This paper proposes a relative energy-based feedback command-shaping method that generates a reference-shaping signal only when the relative energy exceeds a prescribed threshold and modifies the torque reference accordingly. The proposed method evaluates the relative energy online from the measured torque and relative angular velocity, and generates a corrective reference-shaping signal to dissipate excess rotational energy while maintaining a fast response. Simulation results show that the proposed method effectively suppresses torque overshoot and maintains robust performance under variations in torsional stiffness without parameter re-optimization.