J. Exp. Theor. Anal., Volume 4, Issue 1 (March 2026) – 6 articles

Multi-laser powder bed fusion is an emerging additive manufacturing technology that enables the production of high-performance components with intricate geometries and large aspect ratios. These tall, slender structures are highly susceptible to steep thermal gradients and residual stress, leading to deformation that compromises dimensional accuracy and structural integrity. This study investigates how geometric compensation, support structure design, and part scaling influence thermal deformation in Inconel 718 components fabricated via multi-laser powder bed fusion. Using pre-compensation, iterative support refinements, and scaled experimental builds, the deformation response across multiple geometries and print strategies is evaluated. Both compensated and original designs are printed on a commercial system equipped with three simultaneously operating lasers. Results show that printing high-angle surfaces without support structures is infeasible, as thermally induced warping and delamination lead to catastrophic failures. Conical support structures spanning critical regions reduce deformation by more than 50% compared to unsupported builds. Reduced-scale parts, however, do not reliably replicate full-scale deformation behavior due to altered boundary conditions and thermal pathways. These findings highlight the need for integrated design-for-AM workflows where compensation, support design, and scale effects are addressed jointly. The study demonstrates that deformation mechanisms do not scale linearly, emphasizing the limitations of small-scale proxies and the necessity of full-scale validation when developing reliable, deformation-aware design strategies for multi-laser powder bed fusion. Full article

This study explores the viability of ultra-high-performance concrete (UHPC) as a structural material for compressed air storage (CAES) systems, combining comprehensive experimental testing and numerical simulations. Scaled (1:20) CAES tanks were designed and tested experimentally under controlled pressure conditions up to 4 MPa (580 psi), employing strain gauges to measure strains in steel cylinders both with and without UHPC confinement. Finite element models (FEMs) developed using ANSYS Workbench 2024 simulated experimental conditions, enabling detailed analysis of strain distribution and structural behavior. Experimental and numerical results agreed closely, with hoop strain relative errors between 0.9% (UHPC-confined) and 1.9% (unconfined), confirming the numerical model’s accuracy. Additionally, the study investigated the role of a rubber interface layer integrated between the steel and UHPC, revealing its effectiveness in mitigating localized stress concentrations and enhancing strain distribution. Failure analyses conducted using the von Mises criterion for steel and the Drucker–Prager criterion for UHPC confirmed adequate safety factors, validating the structural integrity under anticipated operational pressures. Principal stresses from numerical analyses were scaled to real-world operational pressures. These thorough results highlight that incorporating rubber enhances the system’s structural performance. Full article

This paper presents a practical and cost-effective method for in-motion estimation of a vehicle’s CoG position in all three directions by measuring accelerations during two types of maneuvers: braking (longitudinal and vertical CoG estimation) and cornering (lateral and vertical CoG estimation). The proposed method’s main advantage is that it does not require knowledge of vehicle characteristics, such as mass distribution, suspension geometry, or inertia parameters. It relies solely on the known distances between the sensors and their positions relative to a defined reference point on the vehicle. To validate the developed method, experimental tests were conducted on a prototype vehicle, varying the load conditions for the proposed driving scenarios. The CoG position obtained from dynamic maneuvers was compared with reference values derived from static measurements. The results showed that the proposed method could estimate the CoG position with an average error of 3% in the longitudinal direction, a maximum error of 12% in the lateral direction, and a maximum error of 14% in the vertical direction. Full article

Simplified and scalable models of physical systems are extremely valuable in a variety of different engineering fields to test and diagnose particular modes of failure and optimize build conditions. In this work, we develop a practical method to prepare and analyze giant unilamellar vesicles (GUVs) for detailed biophysical interrogations. The method is rapid, scalable, and versatile, where characterization of lipid membrane conformational changes can be performed on multiplexed samples using tissue culture plates and a convenient, high-throughput fluorescence microscopy setup. The simplicity of the setup is enabled by an AI image recognition model that, when trained on the appearance of GUVs in the images, outperforms other image segmentation methods such as the watershed algorithm or the Hough transform. The method allows for the rapid quantification of entire 96-well plates containing in total O (1,000,000) GUVs and provides a potential testbed for the development of drugs. We highlight the power of our system by including large-scale data on the screening of lipophilic analogs of the small molecule antimetabolite carmofur. Full article

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. Full article

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. Full article