基于物理的维度约简用于无模拟运算符学习的刚性微分代数系统
Physics-Guided Dimension Reduction for Simulation-Free Operator Learning of Stiff Differential-Algebraic Systems
AI总结:
本文提出基于物理的深度ONet扩展层,通过精确约束和快速动态约简,提升刚性微分代数系统运算符学习的精度与效率。
摘要:
神经代理用于刚性微分代数方程(DAEs)面临两个障碍:软约束方法留下代数残差,刚性放大导致误差;硬约束方法需要来自刚性积分器的轨迹数据。我们引入扩展的牛顿隐层,通过单次可微求解精确强制代数约束并减少快速动态到准稳态值。嵌入物理感知深度ONet中,该层从慢态预测中精确恢复所有快速和代数状态,去除每窗口刚性放大路径,并产生无惩罚方法的刚性缩放隐函数定理梯度。级联隐层扩展到多组分系统,具有可证明的收敛性。在形成逆变器(刚性比约为4712)上,扩展牛顿达到1.42%误差(对比惩罚方法39.3%和标准牛顿57.0%);增强拉格朗日和反馈线性化发散。两个独立训练模型无需重新训练(0.72%到1.16%误差,精确约束满足)。跨领域验证在罗伯逊刚性DAE(刚性比高达$10^5$)上确认泛化性。符合预测提供90%覆盖,自动检测分布外情况。
英文摘要:
Neural surrogates for stiff differential-algebraic equations (DAEs) face two barriers: soft-constraint methods leave algebraic residuals that stiffness amplifies into errors, and hard-constraint methods require trajectory data from stiff integrators. We introduce an extended Newton implicit layer that enforces algebraic constraints exactly and reduces fast dynamics to their quasi-steady-state values in a single differentiable solve. Embedded in a physics-informed DeepONet, the layer recovers all fast and algebraic states exactly from slow-state predictions, removes the per-window stiffness-amplification pathway, and yields a stiffness-scaled Implicit Function Theorem gradient absent from penalty methods. Cascaded implicit layers extend this to multi-component systems with provable convergence. On a grid-forming inverter (stiffness ratio of about 4712), extended Newton attains 1.42% error versus 39.3% (penalty) and 57.0% (standard Newton); augmented Lagrangian and feedback linearization diverged. Two independently trained models compose without retraining (0.72% to 1.16% error, exact constraint satisfaction). Cross-domain validation on the Robertson stiff DAE (stiffness ratio up to $10^5$) confirms generalization. Conformal prediction provides 90% coverage with automatic out-of-distribution detection.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。