ABSTRACT High‐fidelity computational fluid dynamics (CFD) simulation usually carries a heavy burden, especially for parametric CFD simulations requiring multiple calculations. To address this challenge, researchers have developed reduced‐order modeling (ROM) to significantly decrease the burden by building simplified model. This article proposes hybrid method of weighted proper orthogonal decomposition and Kriging, novel method. improves accuracy model assigning appropriate weights samples...

Monitoring the state of battery, including charge (SOC) and health (SOH), is crucial for ensuring safety reliability electrical equipment. The paper presents a novel hybrid network that combines nonlinear autoregressive model with exogenous inputs (NARX) DS-attention. proposed DS-attention method establishes robust mapping relationship between outputs, it specialized recurrent neural enhances estimation performance by incorporating division function self-adaptive into attention mechanism....

Abstract When solving the black-box dynamic optimization problem (BDOP) in sophisticated system, finite difference technique requires significant computational efforts on numerous expensive system simulations to provide approximate numerical Jacobian information for gradient-based optimizer. To save budget, this work introduces a BDOP framework based right-hand side (RHS) function surrogate model (RHSFSM), which RHS derivative functions of state equation are approximated by models, and is...

<title>Abstract</title> This paper introduces a novel Feature-Extended Neural Ordinary Differential Equation (Fe-NODE) method for modeling black-box control systems by incorporating parameters into the ordinary differential equation. Since input to model should consider both system state and parameters, this is challenge process. The primary contribution of proposed employing feature expansion add information enhance perception ability while utilizing intermediate supervision teacher forcing...

In this paper, a response band-based method for time-dependent reliability-based robust design optimization is proposed. The proposed provides novel alternative framework, consist of two-step transformation stage and solving stage, to solve the problem. original problem transformed into an equivalent deterministic in settled stage. dynamic modal decomposition technique kriging are combined overcome that there no standard both time division observation sampling commonly used methods. approach...

Abstract In the cooling fan optimization, there are many local minima near optima, which improves accuracy requirement of Kriging model. Due to unexpected prediction errors caused by some deceptive samples, model exploration capability traditional method is not enough. To overcome this problem, an adaptive based on trust index proposed in paper. By considering sample distribution and region nonlinearity, used evaluate reliability can enhance sampling strategy for new candidates. Several...

Due to the presence of capacity regeneration and random disturbances during degradation process lithium-ion batteries, decay curve batteries exhibits a non-linear unstable trend. Therefore, accurately estimating SOH is critical ensuring safety reliability electrical device. This paper presents hybrid method nonlinear auto-regressive model with exogenous inputs neural network (NARX) DS-Attention. The NARX improves estimation performance by incorporating as auxiliary factors, while...