Journal of Experimental and Theoretical Analyses

This study investigates the fatigue strength of a motor hanger used in high-speed electric multiple units (EMUs). Finite element analysis and field measurements revealed that reduced weld penetration significantly increases stresses in welded regions. Line tests demonstrated that a 100 Hz torque ripple induces elastic vibration of the hanger, serving as the primary driver of stress propagation, with stress and acceleration levels increasing proportionally with the torque ripple amplitude. This 100 Hz excitation lies close to the hanger’s constrained modal frequency of about 109 Hz, creating a near-resonance condition that amplifies dynamic deformation at the welded joints and accelerates fatigue crack initiation. Hangers with lower in situ modal frequencies exhibited higher equivalent stresses. Joint dynamic simulation further showed that increasing motor mass reduces the longitudinal acceleration of the hanger, while enhancing the radial stiffness of rubber nodes markedly decreases both longitudinal and vertical vibration accelerations as well as stress responses. Based on these insights, a structural improvement scheme was developed. Strength analysis and on-track tests confirmed substantial reductions in overall and weld stresses after modification. Fatigue bench tests indicated that the critical welds of the improved hanger achieved a service life of 15 million km, more than twice that of the original structure (7.08 million km), thereby satisfying operational safety requirements.

Optimising the learning rate is essential for efficient neural network training, but static methods can cause overshooting or undershooting, while adaptive techniques like ADAM often struggle to balance exploration and exploitation. We introduce the Polling Method, an ensemble-based optimisation approach that dynamically selects the most effective learning rate at each step, improving convergence and mitigating issues inherent in traditional optimisation strategies. By evaluating base models with varying learning rates at each epoch, the method adaptively balances exploration and exploitation without being constrained by predefined functions or gradient noise. This study details the theoretical foundation, implementation, and integration of the Polling Method with the ADAM optimiser, demonstrating its effectiveness in Artificial Neural Networks and Bayesian variational inference. The results demonstrate that Polling Method-ADAM reduces absolute error by 50% compared to ADAM alone, while also accelerating convergence. In Bayesian optimisation, it reduces the mean gradient shift from 0.85 to 0.7835 over 500 iterations, indicating improved stability in high-dimensional problems. By introducing adaptive learning rate selection within training, the Polling Method enhances optimisation efficiency while mitigating noise accumulation. This framework provides a computationally efficient, flexible alternative for deep learning applications, offering significant improvements over traditional optimisers and a potential breakthrough in neural network training strategies.

Ionic propulsion, where charged particles, ions, are produced between electrodes and accelerate towards the negative electrode, has practical applications as a propulsion system in the space industry; however, its adoption to in-atmosphere ionic propulsion is relatively new and faces different challenges. A high potential difference is required to achieve a corona discharge between a positive and negative electrode. In this work, we will explore the feasibility of ionic propulsion using CFD modelling to replicate the effect of the ions, with a future aim of improving efficiency. The ionization region is modelled for a 15 kV potential difference, which is replicated with a velocity inlet, based on experimental data. The output velocity from the numerical simulation shows the same trend as theoretical predictions but significantly underestimates the magnitude of the ionic wind when compared with theoretical estimates. Further modelling is highlighted to improve predictions and assess if the theoretical model overestimates the ionic wind.