We develop a data-driven framework for discovering constitutive relations in models of fluid flow and scalar transport. Under the assumption that velocity and/or scalar fields are measured, our approach infers unknown closure terms in the governing equations as neural networks (gray-box discovery). The target to be discovered is the constitutive relations only, while the temporal derivative, convective transport terms, and pressure-gradient term in the governing equations are prescribed. The formulation is rooted in a variational principle from non-equilibrium thermodynamics, where the dynamics is defined by a free-energy functional and a dissipation functional. The unknown constitutive terms arise as functional derivatives of these functionals with respect to the state variables. To enable a flexible and structured model discovery, the free-energy and dissipation functionals are parameterized using neural networks, while their functional derivatives are obtained via automatic differentiation. This construction enforces thermodynamic consistency by design, guaranteeing non-negative entropy production and, in the absence of non-variational (convective) contributions, monotonic decay of the total free energy. The resulting method, termed GIMLET (Generalizable and Interpretable Model Learning through Embedded Thermodynamics), avoids reliance on a predefined library of candidate functions, unlike sparse regression or symbolic identification approaches. The learned models are generalizable in that functionals identified from one dataset can be transferred to distinct datasets governed by the same underlying equations. Moreover, the inferred free-energy and dissipation functions provide direct physical interpretability of the learned dynamics. The framework is demonstrated on several benchmark systems, including the viscous Burgers equation, the Kuramoto–Sivashinsky equation, and the incompressible Navier–Stokes equations for both Newtonian and non-Newtonian fluids.
The safety of long-distance natural gas steel pipelines is increasingly challenged by the coupled effects of progressive corrosion and accidental seismic events. Traditional methods are difficult to accurately reflect the spatiotemporal uncertainty of the structures' mechanical behavior caused by corrosion evolution and seismic loads. The study proposes a probability-based multi-scale stochastic framework to quantify the failure risk of corroded pipelines under spatiotemporal earthquake excitations. The equivalent morphological representation of natural corrosion can be integrated into the stochastic dynamic mechanical analysis model. A bridging-domain formula is introduced to consistently transfer stochastic information across mesoscale-defect and structural scales, enabling uncertainty characterization of nonlinear system response. The extreme value theory-based probability envelope model is developed to evaluate the time-variant failure probability under unexpected earthquake scenarios using the spatiotemporal seismic mechanics behavior. The proposed framework provides a closed-form solution that significantly improves the computational efficiency of seismic risk assessment for large-scale underground structures. The results indicate that the coupling effect between corrosion morphology and seismic excitation randomly amplifies the failure probability of pipelines. The proposed probabilistic modeling can support pipeline integrity management under multiple disasters.
With the continuous growth of data storage demand, precise and stable magnetic-head positioning in hard disk drives (HDDs) remains essential. This paper proposes a disturbance-rejection-oriented control design for triple-stage actuator (TSA) systems. The approach employs Bode-plot analysis to achieve robust performance against plant perturbations while explicitly incorporating the stroke limits of Lead Zirconate Titanate (PZT) micro-actuators into controller synthesis. Stroke constraints are represented as forbidden regions on the Bode diagram, providing an intuitive feasibility criterion for controller design. The physical stroke limitation is formulated as an inequality condition and embedded directly in the loop-shaping procedure. Performance is evaluated through hybrid simulations in which numerical plant models identified from measured actuator frequency responses are driven by disturbance signals generated from measured disturbance characteristics. The results demonstrate that the proposed controllers significantly reduce the risk of micro-actuator saturation while preserving high positioning accuracy.
With the rapid expansion of wind power generation, concerns have grown over bird collisions with turbines, particularly involving rare species such as the White-tailed Eagle (Haliaeetus albicilla). To evaluate potential impacts, we compared bird distribution estimates from roadside surveys with flight activity data from point censuses. Roadside surveys were conducted along a 110.4 km route over a one-year period, while point censuses were carried out at 33 sites—18 wind facilities and 15 control areas—to measure flight frequencies. Species distribution models were developed using MaxEnt, and their outputs were compared with point census results using generalized linear models. White-tailed Eagles showed the highest flight probability near coastlines, influenced by elevation and proximity to secondary forests. Slaty-backed Gulls (Larus schistisagus) were primarily associated with coastal proximity and artificial structures, with frequent flights in coastal and urban areas. Crows (Corvus spp.) exhibited high activity near fishing ports and livestock farms, reflecting their association with human structures. MaxEnt estimates of Slaty-backed Gulls and crows significantly predicted point census flight frequencies in GLM analysis. These results demonstrate that roadside surveys can efficiently and reliably estimate bird distributions in open landscapes with good visibility. Compared with point censuses, they offer a more cost-effective approach, making them a practical tool for environmental impact assessments and bird collision risk evaluations-especially in regions with limited resources.
HiZ-GUNDAM is a candidate for a Japan Aerospace Exploration Agency satellite mission expected to launch in the 2030s, aiming to detect high-redshift gamma-ray bursts (GRBs) and explore the early universe. The wide-field X-ray monitor EAGLE, onboard the HiZ-GUNDAM satellite and responsible for GRB detection and localization, employs lobster-eye-type micro-pore optics (MPOs) and a pnCCD as the focal-plane detector to focus soft X-rays in the 0.4 to 4.0 keV range. We established an alignment strategy for a modular lobster-eye MPO system, and the required alignment accuracy was derived through Monte Carlo simulations that reproduce the point spread function and localization performance. The optical characteristics of individual MPO segments were then evaluated to select and arrange four segments for a breadboard-level alignment demonstration. Subsequently, alignment tuning was performed by adjusting the angular positions of the MPO segments with actuators under X-ray irradiation in a vacuum environment, followed by an evaluation of the resulting direction-determination accuracy. The resulting angular response measurements show that the estimated incident angles derived from the focal positions achieve systematic uncertainties of less than 3 arcmin over more than 95% of the field of view. This level of performance satisfies the localization requirement for the EAGLE instrument when considering systematic errors intrinsic to the MPO-based detection system. These results demonstrate the feasibility of precise MPO alignment in a segmented lobster-eye optics system and provide practical feedback for the design, alignment strategy, and mechanical tolerance definition of future engineering and flight models of the EAGLE X-ray optics.
