The formulation of an ordinary differential equation (ODE) to describe a dynamic system is generally the first and most crucial step before numerical investigations may be performed. However, this process step will fail if the underlying physics is not fully understood or cannot be described. In such cases, running real-world experiments may be the only possibility to get some insight into the system behavior. Based on the collected data, a surrogate for the ODE may be adjusted until it behaves like the original system. Such a substitution is also recommended if time-intensive simulations need to be combined e.g. with global optimization algorithms. Here, an approach is presented where a neural network surrogate for the state function of the system is trained while being embedded into a numerical integration scheme with step-size control. During forward propagation, an initial value problem is solved while the embedded neural network is treated as state function. To update the adjustable weights of the neural network, the output system states are compared to target system states and the corresponding loss is propagated backward using error backpropagation methods with the goal to minimize the discrepancy against measured data. The surrogate state function optimized in this way may be finally used as a substitute for the unknown ODE under investigation.
Long-span precast prestressed concrete girders are susceptible to lateral instability during temporary phases such as lifting, transport and placement on bearings. As spans have increased, failures have occurred, driven by initial imperfections from fabrication tolerances, creep, shrinkage and thermal effects. This paper presents a robust optimisation framework that combines a model of hanging beam stability with a most likely heteroscedastic Gaussian process surrogate to identify minimum-weight cross-sections satisfying a probabilistic stress constraint under geometric and material uncertainty. The surrogate learns both the mean stress response and its input-dependent variance, and the resulting optima are validated against Monte Carlo solutions. Decomposing uncertainty into epistemic and aleatoric components distinguishes regions of the design space where the surrogate lacks training data from those where the structural response is inherently sensitive to parameter variability, indicating where additional data or design margin is most required. Results show that the area premium associated with a high target reliability is modest.
Electric vehicle (EV) battery packs strongly influence vehicle performance, safety, cost, and energy efficiency. This paper presents a structured review of optimization approaches for EV battery-pack design, focusing on four interacting domains: structural design, thermal management, spatial packaging, and system-level multidisciplinary design optimization (MDO). The reviewed literature is examined to identify common design objectives, simulation tools, optimization algorithms, and remaining research gaps. The review shows that structural optimization, particularly crashworthiness and lightweight enclosure design using finite element analysis, surrogate modelling, and multi-objective algorithms, is comparatively well developed. Thermal management is also one of the most actively studied areas, with extensive use of computational fluid dynamics, surrogate modelling, and multi-objective optimization. In contrast, spatial packaging optimization, defined here as the three-dimensional placement and orientation of battery modules or pack structures within a vehicle-defined envelope, remains less systematically addressed. Few studies integrate structural, thermal, packaging, electrical, and vehicle-level requirements within a single solver-based MDO framework. The review therefore identifies vehicle-level spatial battery-pack optimization as a key research gap and a promising direction for future integrated battery-pack design.
Highway autonomous driving demands coordinated longitudinal and lateral decision-making to ensure safety, comfort, and traffic efficiency. This paper proposes an explainable fuzzy navigation approach addressing two independently developed functions: adaptive cruise control (ACC) and lane-change decision-making. For ACC, a Sugeno-type fuzzy inference system governs longitudinal control using inter-vehicle gap error, relative velocity, and speed-limit deviation, with membership function parameters tuned via a simulation-based Bayesian ε -constraint multi-objective optimization framework that explicitly balances safety and comfort. For lane-change decisions, a Nash-product-based multiobjective fuzzy decision framework is adopted. A Safety and a Progress utility channels each evaluate lane-change opportunities through dedicated Sugeno fuzzy inference systems, generating utility scores over target-lane gap geometry, relative speed dynamics, and speed deficit. A lane change is executed if and only if both utility channels strictly exceed their disagreement-point scores - the utility of remaining in the current lane - enforcing a mutual surplus condition that prevents geometrically safe but strategically unnecessary maneuvers. The lane-change module is evaluated as a standalone decision layer across static traffic configurations. The Nash-FIS configuration outperforms both the hand tuned baseline and a MOBIL-style reference on all borderline scenarios while preserving the intended safety barrier across all hard-veto cases. Integration of both modules into a unified dynamic simulation framework, encompassing car-following, lane-change, and overtaking maneuvers, constitutes the next stage of this work.
The design of a ship energy system can be complex due to multiple design options (e.g choice for component types or occurence) and constraints (e.g environmental, economical, military). The optimization problem that can help to find an optimal architecture (components and associated connections) is typically referred as a synthesis, design and operation optimization (SDO) problem. Evolutionary algorithms are usually employed to tackle SDO problems and they provide good results. However, due to the very important number of objective function evaluations involved, the physical modeling must remain very simple (causal behavior without differential equations or matrix inversions). A Bayesian optimization (BO) approach is proposed in this article to tackle this issue. The expected advantage of BO is its ability to represent a global approximation of the objective function with few data points while its weakness lies in the number of optimization variables it can handle. This high-dimensional challenge is tackled here by a decomposition of the optimization problem into the different modes that represent the ship mission profile. A numerical example is presented for a ship energy system than can achieve propulsion and electricity production by multiple means. The effect of the formulation of constraints in the objective function will also be discussed as it can have a strong impact on the optimization results.
