The Lattice Boltzmann method (LBM) is particularly well-suited for implementation on quantum circuits owing to its simple algebraic operations and natural parallelism. However, most quantum LBMs fix τ = 1 to simplify the lattice Boltzmann equation to a linear system with respect to the equilibrium distribution function, which restricts the simulation to a fixed mesh size for a given Reynolds number. To preserve the simplicity associated with τ = 1, we propose a quantum lattice kinetic scheme (LKS). In this approach, a constant parameter A is introduced into the equilibrium distribution function (EDF), enabling independent adjustment of the fluid viscosity. This modification decouples the viscosity from the mesh size, thereby enabling simulations at varying Reynolds numbers on a fixed mesh size. The Chapman-Enskog analysis confirms that the modified EDF still recovers the Navier-Stokes equations without compromising accuracy. We evaluate the method on 2D and 3D Taylor-Green vortex and lid-driven cavity flows, demonstrating that quantum LKS attains the same accuracy and convergence order as classical LKS. The first application of quantum LBM to 3D incompressible flows represents a significant step forward in large-scale fluid dynamics simulation. All the data and code associated with this work have been made publicly available at: https://github.com/XiaoY-1012/QLKS-LBM.
Randomized Kaczmarz methods provide efficient iterative solvers to tackle large linear systems, and their performance depends strongly on the row or column selection strategy. In this paper, we propose an extended semi-randomized Kaczmarz method for large-scale inconsistent linear systems. The proposed method incorporates a local sample-maximum row-selection strategy into a partially randomized extended Kaczmarz framework: the z-iteration is used to handle the inconsistent component, while the x-update uses the latest vector zk+1, with the working row selected from a uniformly sampled subset according to the largest normalized residual. Thus, the method uses current residual information without constructing a global greedy index set or a full probability distribution over all rows. For tall coefficient matrices with full column rank, we prove the convergence in expectation of the proposed method to the least-squares solution, characterize an upper bound for its convergence rate in expectation, and introduce a gain factor to describe the effect of the sample-maximum row-selection rule. Numerical experiments demonstrate that the proposed method is competitive with several existing randomized Kaczmarz-type methods.
Inspired by the fire-making tool (the Iroquois pump drill), a new two-degree-of-freedom (2-DOF) string-driven rotational energy harvester (SREH) is proposed, which can efficiently harvest energy from ambient vibrations at ultra-low frequency (0.2 Hz). The proposed SREH is composed of a 2-DOF string-driven rotor and a magnet-coil transduction unit. The 2-DOF string-driven rotor consists of a rotor, a lid, an elastic string, an inelastic string, and support springs, capable of effectively converting ultra-low frequency vibrations into high-speed rotation (328 rpm). The magnet-coil transduction unit can transfer the kinetic energy to electricity. Numerical and experimental results both show that the proposed SREH can deliver stable electrical outputs under harmonic excitation (1-5 Hz). Under an excitation frequency of 5 Hz and an excitation amplitude of 7.5 mm, the proposed harvester can generate a power of 1.55 mW. Furthermore, with flywheel energy storage mechanism, the proposed harvester can generate a power of 0.64 mW under an impact excitation of 0.2 Hz. In addition, under manual pressing excitation (approximately 6 Hz), the proposed device is capable of generating a power of 1.54 mW and rapidly charging a 1 mF capacitor from 0 V to 1.58 V within approximately 8 s. The results exhibit the great potential of the string-driven harvester for efficiently scavenging ultra-low frequency vibration energy.
Design optimization of offshore wind turbine (OWT) monopiles presents a high-dimensional challenge, as the pile is discretized into multiple segments (typically 2–3 m in length), each with independent diameter and wall thickness. Furthermore, the optimization is governed by various design constraints (ultimate, fatigue, and serviceability limit states), numerous design load cases (DLCs), and complex multi-physics involving aero-hydro-servo-elastic-soil interactions. These factors necessitate computationally expensive time-domain simulations. Conventional optimization methods are susceptible to local optima and struggle with the high-dimensionality, multiple constraints, and numerous load cases inherent to this problem. Moreover, current monopile designs are seldom optimized using integrated multi-physics simulations, which can lead to suboptimal or potentially over-conservative designs. This study proposes an efficient global optimization framework for high-dimensional, multi-constrained, integrated design of monopiles under various load cases. The framework comprises three key components: (i) an integrated aero-hydro-servo-elastic-soil model for providing high-fidelity dynamic responses; (ii) the subset simulation optimization (SSO) algorithm for efficient global exploration of the high-dimensional design space; and (iii) a high-fidelity multivariate metamodel based on Gaussian process regression (GPR) for accelerating the optimization process. The effectiveness of the proposed framework is demonstrated by optimizing a monopile for a 15 MW OWT. Benchmarking against a suite of optimization algorithms (GD, GA, PSO, SA) demonstrated the superiority of SSO, which achieved a 9% reduction in mass and reduced the required computational time by 38%. The results identify fatigue, rather than ultimate strength, as the governing constraint for monopile optimization. The framework equips designers with a computationally efficient, globally optimal route to lighter offshore wind turbine foundations, cutting costs while sustaining structural integrity.
Existing data-driven remaining useful life (RUL) prediction methods often suffer from limited physical interpretability and insufficient uncertainty quantification. To address these limitations, we propose a Bayesian hybrid prognostic framework that integrates local physics priors with global statistical information. First, to capture thermodynamics-consistent local couplings in aero-engines, we develop a physics-constrained Bayesian dynamic graph neural network. The network defines the sensor topology from mechanical transmission pathways and temperature/pressure evolution, and uses Bayesian variational inference to quantify time-varying uncertainty in inter-component couplings. Second, a dynamic Kriging model is used to efficiently retrieve similar historical degradation patterns, providing globally informed references across varying operating conditions. Finally, we design a learnable Kalman filter fusion mechanism, where the predictive covariance serves as an uncertainty indicator to adaptively balance contributions from the local physical state and the global statistical experience. Experiments on the C-MAPSS and N-CMAPSS datasets demonstrate that the proposed method improves prediction accuracy and provides interpretable uncertainty characterization of time-varying sensor couplings.
