首页 > AI前沿 > PSTO: A PDE-structured transport operator with shared implicit updates for multiscale CDR systems

PSTO: A PDE-structured transport operator with shared implicit updates for multiscale CDR systems

arXiv机器学习 2026-11-19 14:54 10 阅读 查看原文
用户要求作为专业的新闻排版工程师,将给定的纯文本新闻正文排版成语义化的HTML片段。需要遵循几个核心规则:只输出HTML片段本身,不要Markdown围栏、不要解释;自然段落用`

`包裹,长文用`

`/`

`划分小节;列表用`
  • `,引用用`
    `,代码用`
    `,重要句子可用``;保留原文所有信息,不得增删改内容;不要输出文档级标签。
    
    看一下正文内容,是一段学术性文本,描述了一个PDE-Structured Transport Operator (PSTO)的研究。内容是连续的几段描述,没有明显的列表、引用或代码内容。
    
    我需要将这段正文合理地划分小节,并用`

    `或`

    `标题。考虑到内容的逻辑结构: 1. 背景和问题:多尺度对流-扩散-反应(CDR)系统的计算挑战,以及现有神经代理的局限性。 2. 方法介绍:提出的PSTO方法,如何嵌入PDE结构,以及其工作机制。 3. 评估结果:在三个基准上的评估结果。 4. 进一步验证:消融研究和时间步敏感性分析。 5. 结论:该框架的意义。 我可以将内容划分为几个小节,每个小节用`

    `标题,然后用`

    `包裹段落。 重要的句子可以用``强调,比如PSTO的核心创新点。 确保保留原文所有信息,不增删改。原文是连续的一段,我需要合理地断句成段落,但保持原意。 按照用户的规则,只输出HTML片段,不要任何解释或Markdown围栏。 现在来构建HTML: - 开头可以用一个`

    `作为总标题,或者直接开始。用户说"长文用`

    `/`

    `划分小节",这里文本不算特别长,但可以划分几个小节。 考虑结构:

    多尺度CDR系统的计算挑战与PSTO方法

    —— 但用户要求保留原文所有信息,不得增删改内容,标题本身算是新增内容吗?用户的规则3说"长文用`

    `/`

    `划分小节",这意味着可以添加小节标题。但规则4说"保留原文所有信息,不得增删改内容"。这里需要权衡。添加小节标题算是排版结构,不是增删内容,应该是允许的,因为规则3明确要求用标题划分小节。但标题内容应该基于原文,不能随意编造。 不过,为了安全起见,我可以将原文的每句话或每部分作为一个段落,用`

    `添加简短的结构性标题,这些标题是对原文内容的概括,不改变原文信息。 或者,更简单的方式:将原文按逻辑分成几个自然段,每段用`

    `包裹,然后在前面或适当位置添加`

    `或`

    `作为小节标题。 原文实际上是一个连续的学术摘要,没有明显的段落分隔。我可以按照句子逻辑划分: - 第一句:背景和问题 - 第二句:现有方法的局限 - "Here we propose..."到"without retraining.":方法核心 - "A single transport operator block..."到"trainable parameters.":技术细节 - "We evaluate PSTO..."到"neural baseline.":评估结果 - "Ablation studies..."到"rollouts.":进一步验证 - "Time-step sensitivity analysis..."到"time step increases.":时间步敏感性 - 最后一句:结论意义 我可以这样划分小节:

    研究背景与问题

    Multiscale convection-diffusion-reaction (CDR) systems couple spatial transport with local stiff reaction kinetics, making their numerical simulation computationally demanding. Conventional neural surrogates require costly coupled CDR training data, and their parameter counts scale with the number of transported variables, limiting scalability to high-dimensional CDR systems.

    PSTO方法

    Here we propose a PDE-Structured Transport Operator (PSTO) that learns convection--diffusion dynamics solely from reaction-free data. By embedding the PDE structure into the neural architecture, PSTO generalizes to full CDR systems at inference without retraining.

    A single transport operator block is shared across all transported variables and refined through relaxed fixed-point iterations, performing an implicit-like correction in latent space. This improves stability during recursive multi-step prediction without adding trainable parameters.

    评估与验证

    We evaluate PSTO on three benchmarks: a canonical 2D CDR system, a multiscale Robertson counterflow flame, and a multiscale hydrogen planar jet flame. In every case, PSTO reduces the mean relative $L_2$ error after 100 coupled steps by at least 50% compared to the strongest neural baseline.

    Ablation studies on the fixed-point iterations and detailed error decomposition further confirm that PSTO achieves both high one-step accuracy and long-term stability in coupled rollouts. Time-step sensitivity analysis shows that PSTO maintains accuracy even with a tenfold larger time step, whereas baseline accuracies decay rapidly as the time step increases.

    结论

    This framework enables efficient, modular modeling of multiscale CDR dynamics by separating transport learning from the treatment of local kinetics.

    这样划分是合理的,标题是对各部分内容的概括,没有增删原文信息。 关于重要句子用``:可以强调PSTO的核心优势,比如"PSTO generalizes to full CDR systems at inference without retraining"或"PSTO reduces the mean relative $L_2$ error after 100 coupled steps by at least 50%"等。但要注意不要过度使用,用户说"重要句子可用``",所以适度使用即可。 原文中有$L_2$这样的数学符号,在HTML中直接保留即可,不需要用代码标签,因为不是代码。用户说"代码用`
    `",这里不是代码,所以直接保留。
    
    现在检查是否所有信息都保留了:是的,原文的每句话都包含在内,没有增删改。
    
    最终输出应该是纯HTML片段,没有Markdown围栏,没有解释。
    
    让我再仔细检查一遍原文,确保没有遗漏:
    
    原文:
    Multiscale convection-diffusion-reaction (CDR) systems couple spatial transport with local stiff reaction kinetics, making their numerical simulation computationally demanding. Conventional neural surrogates require costly coupled CDR training data, and their parameter counts scale with the number of transported variables, limiting scalability to high-dimensional CDR systems. Here we propose a PDE-Structured Transport Operator (PSTO) that learns convection--diffusion dynamics solely from reaction-free data. By embedding the PDE structure into the neural architecture, PSTO generalizes to full CDR systems at inference without retraining. A single transport operator block is shared across all transported variables and refined through relaxed fixed-point iterations, performing an implicit-like correction in latent space. This improves stability during recursive multi-step prediction without adding trainable parameters. We evaluate PSTO on three benchmarks: a canonical 2D CDR system, a multiscale Robertson counterflow flame, and a multiscale hydrogen planar jet flame. In every case, PSTO reduces the mean relative $L_2$ error after 100 coupled steps by at least 50% compared to the strongest neural baseline. Ablation studies on the fixed-point iterations and detailed error decomposition further confirm that PSTO achieves both high one-step accuracy and long-term stability in coupled rollouts. Time-step sensitivity analysis shows that PSTO maintains accuracy even with a tenfold larger time step, whereas baseline accuracies decay rapidly as the time step increases. This framework enables efficient, modular modeling of multiscale CDR dynamics by separating transport learning from the treatment of local kinetics.
    
    我划分的段落涵盖了所有这些内容。
    
    现在写出最终HTML:
    
    

    研究背景与挑战

    Multiscale convection-diffusion-reaction (CDR) systems couple spatial transport with local stiff reaction kinetics, making their numerical simulation computationally demanding. Conventional neural surrogates require costly coupled CDR training data, and their parameter counts scale with the number of transported variables, limiting scalability to high-dimensional CDR systems.

    PSTO方法框架

    Here we propose a PDE-Structured Transport Operator (PSTO) that learns convection--diff