Research progress on the application and development of the material point method in dynamic simulation of geological hazards
-
摘要:
在解决崩塌、滑坡、泥石流等大变形地质灾害问题时,常采用数值模拟的方法。如何准确高效地模拟这类问题一直以来都是个难题。物质点法(material point method,MPM)作为一种新兴的数值方法,克服了传统有限元和有限差分等数值方法在模拟大变形时产生的网格畸变问题,并已成功应用于地质灾害中的大变形分析。为了解物质点法在地质灾害大变形模拟中的研究进展,文章在现有研究的基础上简要介绍了物质点法的基本原理,主要总结了物质点法在模拟滑坡、泥石流、地裂缝等地质灾害大变形问题中的应用以及最新的研究进展。同时,指出了物质点法在现有研究中存在的精度、计算效率、多物理场耦合等问题,并展望了物质点法在工程地质中进一步发展的趋势。
Abstract:Numerical simulation is commonly used to address large deformation geological disasters such as collapses, landslides, and debris flows. Accurately and efficiently simulating these issues has always been a challenge. The material point method (MPM), as emerging numerical method, overcomes the grid distortion problems of traditional numerical methods such as the finite element method (FEM) and finite difference method (FDM) when simulating large deformations, and has been successfully applied in the large deformation analysis of geological disasters. In order to understand the research progress of MPM in the large deformation simulation of geological disasters, this paper briefly introduces the basic principles of MPM based on current research. It also summarizes the application of MPM in simulating large deformations of geological disasters such as landslide, debris flow, and ground fracture, highlighting the latest research progress. Furthermore, it identifies issues in existing MPM research, such as accuracy, computational efficiency, and coupling of multi-physics fields, and discusses future trends in MPM development withinengineering geology.
-
0. 引言
近年来,人工智能(artificial intelligence, AI)技术发展迅猛。特别是大语言模型(large language models, LLMs),在多模态数据分析、复杂场景建模、领域知识融合等关键领域实现了突破。这些突破促使AI从通用领域向各类垂直领域渗透[1]。如今,AI在专业领域应用的门槛越来越低,应用场景不断扩展,成果更加丰富,服务链条持续延伸。在各个研究领域,AI催生出一系列新理念、新思路和新范式,尤其是LLMs等AI技术,正在加速改变甚至颠覆传统研究模式[2]。
当前LLMs通过自监督学习,从海量文本中总结语言的统计规律、语义关联和句法特征,从而实现了文本生成与理解的智能化飞跃,在很多场景中展现出类似人类的任务处理能力。随着拥有千亿级参数的模型不断涌现和训练数据的呈指数级增长,语言模型已逐步突破传统应用限制,在自然科学领域展现出重组知识和进行创新性推理的可能[3]。比如,基于深度学习研发的AlphaFold模型用于蛋白质结构预测,改进了传统生物学研究方法,提高了研究效率[4]。这充分体现了AI技术在打破学科数据界限、推动跨学科研究方面的变革性作用,也意味着继数据归纳、理论推演和仿真模拟之后,“AI-for-science”这一研究范式正在兴起,其意义深远且具有开创性[4 − 5]。
在地球科学领域,LLMs等AI技术主要从数据分析、复杂建模和知识发现3方面[5 − 6],创新地球科学的研究范式。具体包括扩大对大规模多模态数据采集和处理的能力范围,在高维度、多尺度的复杂场景下进行更精准的建模预测,以及深入揭示和加强理解地球物理机制及其演化过程。目前,研究前沿和难点集中在多模态数据的对齐与融合处理、小样本学习、将地学领域知识融入模型以提高模型的透明度和可解释性、解决复杂非线性问题进而突破计算精度与效率等方面[6 − 7]。
在地质灾害智慧防治研究领域,AI技术在识别[8 − 9]、评估[10 − 11]、监测预警[12 − 13]以及智慧服务[14]等场景中的应用越来越广泛和深入,在提高隐患智能识别精度和效率[15 − 17]、监测预警可靠性、评估准确性[18]以及信息服务精准性等关键问题上取得了重要进展。同时,AI技术也为解决地质灾害智慧防治研究中的瓶颈问题提供了技术可能,比如:多模态数据的对齐融合、噪声清洗和语义提取、基于知识推理揭示的易灾特征、规模化制备高质量样本、复杂多维多尺度场景下的精细建模等 [19 − 20]。
随着以LLMs为代表的AI技术快速发展,垂直领域的应用门槛会进一步降低,新的应用场景将不断涌现[21 − 22]。这将促进AI关键技术更深度地融入各个领域,解决关键问题,为不同领域科学研究提供通用AI技术支持[23]。因此,如何缩小地质灾害防治与LLMs等AI技术之间的距离,更加高效和针对性运用AI技术解决地质灾害智慧防治研究关键问题,让AI在更多场景中助力地质灾害研究,从数据、模型和知识层面挖掘交叉研究的创新潜力,催生出“AI+地质灾害”研究新思路和新范式,成为一个值得深入思考的重要课题。
本文以DeepSeek为例,梳理归纳了LLMs技术的发展特点、研发策略和关键技术,探讨了LLMs等AI技术与地质灾害隐患智慧识别的结合思路。在此基础上,基于“知识-数据-模型”互相促进协同发展的理念,提出了包括“应用场景、问题机制、数据模态、样本特征、模型研发、人机协同”等核心要素的“AI+地质灾害”研究框架、技术思路,并结合典型场景进行了简要分析[24]。希望通过本文,在AI技术深刻影响和助力防灾减灾领域的大背景下,在未来“数智智能、知识平权、技术平权”时代,更好推动LLMs等AI技术在地质灾害防治场景中的应用,利用AI技术破解地质灾害防治中的多维度、多尺度复杂非线性问题,实现AI技术从数据、模型和知识的更深层次和更多场景融入地质灾害防治工作,为我国地质灾害智慧防治关键技术研究和体系建设赋能。
1. 大语言模型(LLMs)技术发展和领域应用
当前在全球范围内,LLMs领域研究发展迅猛[24],竞争激烈,呈现“多模竞技”的爆发态势。国外包括GPT系列(OpenAI)、PaLM系列和DeepMind系列(Google)、LLaMA系列(Facebook)等;国内有文心一言(百度)、通义千问(阿里巴巴)、盘古(华为)、豆包(字节跳动)、Kimi(月之暗面)等。LLMs等AI技术在各领域的使用门槛不断降低,对各领域研究技术支撑的能力逐渐增强。
其中,中国深度求索公司(DeepSeek Inc.)发布的DeepSeek模型,以更低数据成本和更小算力规模投入等优势,进一步提升我国在LLMs领域的水平[25]。因此,有必要通过梳理DeepSeek等大语言模型的发展特点,更好了解LLMs技术的优势,弄清相关技术在地质灾害防治中的应用场景以及如何应用、有多大作用,进而给地质灾害智慧防治工作提供参考和启示。
1.1 技术发展特征
作为当前LLMs的典型代表,DeepSeek模型研发思路也体现了LLMs的发展特征(图1)。模型能力提升层面,DeepSeek基于改进型Transformer架构,融合了混合专家模型(mixture of experts,MoE)与稀疏化训练,通过动态路由网络优化计算资源分配,提升了模型训练与推理效率;DeepSeek引入多头潜在注意力(multi-head latent attention,MLA)机制,增强了其对不同语义特征的捕捉能力,并降低显存占用;DeepSeek还采用了八位浮点数精度(floating point 8-bit,FP8)混合精度训练、多令牌预测(multi-token prediction,MTP)等一系列前沿技术,进一步提升模型训练效率和推理速度;基于MoE、参数高效微调(parameter-efficient fine-tuning,PEFT)等轻量化策略,实现在有限样本下的高精度建模。数据处理能力提升方面,通过自监督学习与跨模态对齐技术(contrastive language-image pretraining,CLIP),提升对多源异构数据的融合处理效率。知识嵌入与协同方面,借助人类反馈强化学习(reinforcement learning fromhuman feedback,RLHF)构建“数据感知-知识推理-决策输出”闭环,增强模型可解释性[25 − 26]。
上述一系列研发策略和关键技术(表1)推动DeepSeek模型在多模态数据融合处理、数据噪声清洗、小样本学习、模型轻量化与可迁移性、人机协同与专家知识融合等方面取得重要进展,并在文本生成、语言理解、数学推理、代码生成以及多个垂直领域的逻辑推理、数据融合任务中表现显著,为通用人工智能(artificial general intelligence,AGI)的发展提供新技术路径[26],为驱动不同领域研究提供了通用AI技术新选择。
表 1 LLMs研发策略与关键技术描述Table 1. LLMs development strategies and core technical components思路 策略 主要技术 数据
高效
利用强化数据制备与
噪声清洗利用Attention-Driven Temporal Filtering、Self-Supervised Denosing、Multimodal Joint Denoising、生成对抗网络(generative adversarial network,GAN)及其变种(如CycleGAN)等技术,对视频帧、音频信号、时序数据等多模态数据进行噪声清洗 强化对多模态数据的融合对齐与集成训练 采用数据特征级对齐、语义级对齐、模态间映射和联合学习的策略,利用CLIP-Style等技术,将不同模态的数据映射到统一多维空间内;通过对比学习、掩码预测等多模态自监督学习方法提升融合效果 强化样本生成与
特征增强利用物理驱动的合成数据生成方法,减少对真实数据的依赖;应用Conditional GAN生成特定条件下的数据,扩展训练样本的覆盖度 模型
减重
优化减少模型参数 利用PEFT、Adapters和Prefix Tuning等方法,在保持模型性能的同时减少需要训练的参数 优化任务分配 基于MoE、Sparse Sparse MoE等模型架构,通过动态路由机制将任务分配给特定专家模块 加速模型推理 利用Flash Attention优化技术或线性注意力机制(如Performer、Linformer等)技术,提升注意力机制效率 缩小模型体积 利用INT8量化技术、模型剪枝和知识蒸馏,在精度损失可接受范围内压缩模型参数长度 强化模型迁移 利用低秩自适应(low-rank adaptation,LoRA)、元学习等技术,利用跨领域小样本特征对模型参数进行微调,适应新任务 知识蒸馏技术 利用机器学习技术,将体积庞大、结构复杂、具有庞大参数的模型知识迁移到轻量化模型中,在保持较高性能的同时,显著降低模型的计算复杂度和存储需求 人机
协同
融合强化专家知识嵌入 利用RLHF或交互式机器学习(interactive machine learning,IML)方法,整合专家经验,嵌入模型训练计算,优化模型决策逻辑 构建人机协同
决策体系利用众包理念吸纳专家参与制备高质量样本数据,实现专家介入修正模型偏差;
通过协同过滤和集体智慧等理念提升决策准确性针对地学领域,DeepSeek模型早期版本(2021—2022年)聚焦中文地质文本的理解与生成,利用动态词表扩展技术解决了专业术语的语义嵌入难题,为后续多模态数据融合奠定基础;2023年的架构升级引入视觉编码器与时空数据处理模块,能够实现遥感影像、传感器监测数据与文本描述数据等多类型、跨模态数据的对齐分析,这一阶段的技术突破推动DeepSeek模型在专业化文字理解和生成、地质灾害隐患识别等不同任务中的工程实用化水平进一步提升,为专业技术人员更加便捷和针对性地利用LLMs工具奠定了基础;当前版本(2024年)进一步强化了领域专业化能力,通过强化学习与人类反馈的协同优化,通过与领域知识逐步融合,建立了“数据感知-知识推理-决策输出”的完整技术闭环,为进一步构建地学垂直领域模型奠定基础[26]。
可见,DeepSeek的发展历程反映了LLMs从“规模驱动”向“价值驱动”的发展创新范式的转变[25]。这种转变不仅体现了模型自身技术成熟度的快速提升,也反映了AI技术从辅助工具向决策主体、从单一服务向综合服务的角色转变,还体现了LLMs从通用领域向垂直领域的延伸拓展、持续融合专家知识服务领域需求、破解场景难题的能力提升,这一转变思路也为地质灾害智慧防治研究提供了启示和借鉴。
1.2 领域应用
LLMs技术通用性和解决复杂非线性问题的能力以及低应用门槛使其在不同领域中发挥愈加重要的研究驱动和决策辅助作用,形成差异化的应用生态。在金融领域,通过融合企业财报、供应链数据以及遥感影像、交通物流等多模态数据,构建起动态信用评估体系和模型[3];在医疗领域,基于跨模态数据分析、特征判识及推理能力,助力复杂临床诊断准确率的大幅提升[27];在气象预报[28]、机械结构、热传导、流体动力学、电磁场、材料设计、施工建造等多物理场、多尺度耦合作用的复杂非线性问题领域[29 − 30],构建基于作用机制的物理模型较为复杂,主要基于AI通用模型和领域知识嵌入,聚焦关键问题构建领域和场景模型,通过对海量历史、实测及其模拟数据进行融合分析,开展关联推理,揭示关键特征,发现并学习多场多尺度数据要素间的复杂映射关系,从而实现对机制过程的增强理解和模拟预测[31]。
在地质灾害防治领域,智能化研究主要面临多模态数据对齐融合、噪声清洗和语义提取、基于知识推理的易灾特征揭示、高质量样本规模化制备、复杂多维跨尺度场景精细建模以及数据到应用服务“端到端”难以贯通等数据、模型、服务方面的技术瓶颈。LLMs一系列关键技术突破为上述问题解决提供了新思路:在数据层面,通过整合遥感影像、地质文本、传感器时序数据等多模态数据,实现语义级对齐,破解传统防治体系中的“信息孤岛”困境,解决多模态数据融合处理难题,从不同数据模态强化问题机理和样本特征表达;在模型层面,基于物理约束的神经网络架构,将物理原理机制规则和成熟的专家经验知识嵌入模型推理过程,实现数据驱动和机理驱动的平衡,提升模型科学性和轻量化水平,支撑模型边缘化部署;在服务层面,基于逻辑链技术,实现多模态和多来源数据的定制化、精准化生成和投送,实现从数据到服务的“端到端”贯通,提升日常管理、抢险救援等不同场景下面向管理、技术和公众的快速精准信息服务能力。可以预见,随着LLMs在数据处理、因果推理、模型轻量化、领域知识图谱等方面持续完善,LLMs在赋能地质灾害隐患识别、风险评估、预警预报等方面将发挥更加显著的赋能驱动作用,推动防灾减灾从“经验依赖型”向“智能驱动型”的范式革新。
总结来看,LLMs遵循“通用能力专业化,专业能力工程化”的发展路径,为不同领域研究提供了通用AI技术基础。在通用AI基础上发展垂直领域AI,核心在于构建“AI+领域”深度融合的研究范式。这一范式涉及两个核心问题:“融合什么”、“如何融合”,进一步从“知识-数据-模型”层面展开:基于LLMs的语义理解、知识整合与因果推理能力,从理解领域复杂场景问题出发,结合专家知识揭示灾害演化机理,判识易灾主控因子;以机理认识为指导,重点采集主控因子的多模态表征数据,通过知识图谱、特征量化等技术进一步提炼易灾特征并开展影响量化排序,可进一步通过多种技术手段扩充样本规模,强化对机理特征的表达;将易灾特征、理论知识与专家经验嵌入模型架构与算法优化过程,提升模型推理效率、可解释性与跨场景迁移能力;将众包理念和人机协同思路贯穿“数据标注-特征提取-模型训练-决策校验”全流程,赋能机理分析、因子识别、数据表征、模型优化等技术环节。这一思路的实践意义在于:通过实现“知识-数据-模型”深度融合,推动突破传统“黑箱”局限,更好利用AI技术破解地质灾害防治中的多维多尺度复杂非线性问题,构建具有普适意义的地质灾害防治智慧研究范式。
2. 大语言模型(LLMs)技术赋能地质灾害智慧识别
基于遥感、地质、地理、地质灾害等海量多模态数据和机器学习、深度学习等智能模型的地质灾害智慧识别是当前领域研究的热点[10],基本思路是:基于实地调查和文献分析,提炼区域地质灾害地形地貌、地层岩性、地表覆盖等主控易灾特征,利用综合遥感、无人机航测、地表探测等“空-天-地”技术从不同尺度上获取易灾特征基础数据,制备识别样本,开展架构优化和调参等模型研发工作,利用典型案例开展模型校验与反馈优化,提升识别精度和计算效率,为野外调查、应急救援等防灾减灾工作提供高效支撑,形成“知识驱动-数据融合-模型优化-服务贯通”的技术思路。结合LLMs的一系列关键技术的发展与突破,从知识、数据、模型层面进一步细化和完善技术思路,划分为4个研究环节(图2),分别梳理不同环节上的主要思路、关键技术和研发方向。
(1)知识蒸馏与动态图谱构建阶段(阶段I):基于文献数据资料和专家知识,构建起具有时空属性的知识图谱,突破传统知识的静态局限。进一步利用语义理解技术挖掘孕灾环境、诱发机制、演化路径的潜在关联规则,形成涵盖多模态、多维度的地质灾害隐患智慧识别多维知识图谱。为样本制备、模型研发提供典型特征和理论约束,利用智能算法实现知识的自主进化与持续完善。利用知识蒸馏技术,从已有隐患识别模型中提炼针对不同孕灾环境和灾害类型识别模型关键参数和算法架构特征,降低模型训练数据成本,提升模型推理和训练效率。利用自监督学习技术,持续提升模型对隐患典型特征的识别和提取能力。
(2)多模态数据对齐与质量优化阶段(阶段II):依托“空-天-地”立体观测技术,采集遥感、无人机、地面监测以及岩土体物理参数等多元异构数据,构建统一的时空数据基准框架和跨模态特征映射体系,利用跨模态数据对齐技术,实现影像-文本-时序数据的语义级映射,支撑多模态数据的深度融,提升对隐患识别主控特征的完整表达。建立数据噪声判识指标,利用自监督去噪与注意力驱动滤波技术开展数据清洗、降噪等工作,提升数据信噪比,强化样本质量。建立样本质量评估指标,实现对样本数据质量的动态评估,确保特征完整性与可靠性。
(3)机理-数据双驱动模型构建优化阶段(阶段III):模型基础架构层面,基于MoE和稀疏训练架构,构建任务自适应动态路由机制,将图像处理、文本处理、时序传感器监测数据等不同类型数据由不同模型处理,例如遥感影像处理由Vision Transformer模块解析,地质文本由BERT-GAT模块建模,传感器时序数据由LSTM-Transformer模块分析;机理约束和知识嵌入层面,通过物理损失函数和规则逻辑树等方式将易灾特征、机理模型等领域知识嵌入智能模型,克服单纯靠数据进行驱动分析的过拟合风险,利用知识图谱构建的隐患主控因子判识指标和权重规则为模型提供先验条件,克服数据驱动模型的盲目性;特征挖掘层面,采用多尺度卷积如嵌套式U型网络架构(nested u-net architecture,U-Net++)与图注意力网络(graph attention network,GAT)挖掘隐患隐蔽特征(如坡体裂缝变化、地表温度异常等),利用深度学习技术通过多层次非线性运算提取隐患点的隐蔽特征,通过强化学习反馈优化模型,提升识别精度。AI模型在此过程中形成自主学习能力,采用人机协同的反馈校验模式,逐步提升对特殊、疑难地质环境的泛化识别能力。
(4)服务链条贯通与决策支撑阶段(阶段IV):利用自然语言理解和语义推理技术,从海量的灾情报告、隐患调查、公众报灾等数据资料中,吸收最新科研成果、专家经验以及舆情信息,围绕应用场景特征形成服务产品,不仅输出对风险隐患的识别结果,并利用CoT(chain of thought,CoT)技术拆分和分析理解复杂需求,通过整合风险评估场景模型计算,输出包含风险等级评估、演化趋势预测、防治对策建议等内容的立体化解决方案,为地质灾害防治提供完整的技术支撑。
在上述4个环节中,将众包思路和人机协同的交互模式贯穿始终,基于LLMs的自然语言交互技术,将专家知识高效转化为识别模型的决策逻辑与可量化、可嵌入的约束条件,增强模型的透明性,使专家以对话方式验证模型推断、修正分析路径,提升人机协同支撑下的地质灾害隐患智能识别效率。
3. “AI+地质灾害”研究范式
通过梳理、归纳、总结“AI+地球科学”研究的演化趋势及发展特征,借鉴和利用LLMs等AI技术的研究策略和关键技术,融合考虑场景、问题、知识、数据、模型、人机协同等核心因素,展望提出“AI+地质灾害”的研究范式,突出AI技术在地质灾害多维多尺度非线性复杂关系建模中的重要价值,实现AI从更深层次和更多场景赋能地质灾害智慧防治,是一个值得研究的重要问题。
3.1 研究范式演化
AI与地球科学的协同创新正推动研究范式从“经验驱动”向“数据-机理双驱”转型,两个领域的互促-协同-创新研究已成为热门前沿问题之一[29 − 32]。这一转型主要体现在AI对地球科学研究能力的拓展和地球科学对AI技术研发和应用场景拓展的反哺。AI技术拓展了对地球科学数据采集-处理-分析的数据能力边界和对高纬度复杂场景下精细反演-模拟-预测的模型能力边界,持续推动传统地球科学研究范式发展创新(图3)。同时,地球科学领域研究也在为AI技术发展提供更多应用场景,通过建立基于地球科学知识理论、物理机制及典型案例的专家知识库,不仅提升了智能模型针对高维、多模、海量、复杂数据的理解处理分析能力,也逐步推动模型由“黑箱”到“白箱”、面向复杂模拟过程场景中模型可解释性、可迁移性能力快速发展[28]。
地质灾害研究经历了从经验科学、理论科学、计算科学3个范式和发展阶段,并正在向第4范式数据科学演进(图4)。在数据采集分析能力和AI技术快速发展的大背景下,通过大数据挖掘分析实现科研范式的创新,拓宽研究创新路径,提升知识发现效率,拓展数据处理分析和模型预测模拟的能力边界。
AI技术发展面临的数据质量依赖性强、多模态数据融合分析能力弱、模型体积庞大、模型可解释性低、模型理解和推理能力弱、算力资源消耗大等主要问题,而地质灾害孕灾特征和成灾机理复杂,实现基于智能化技术的地质灾害隐患精准识别、精准化监测预警、精细化风险评估以及定制信息泛在精准服务等方面同样面临特征样本小、数据融合难、计算模型弱、知识嵌入难、信息高效精准投送难等挑战难题(图5)。
因此,结合LLMs“小样本数据-轻量化模型-人机协同”研发策略和关键技术,展望构建基于“知识-数据-模型”互促协同的地质灾害智慧研究范式(图6)。
知识层面:分析提炼隐患识别、风险评估、监测预警等不同类别场景中的关键具体问题,梳理问题的主控因素以及孕灾-致灾的因果(X-Y)关系,为确定基础数据的采集类型和精度提供理论依据。同时,将领域专业知识嵌入样本数据特征和模型研发等关键环节,提升样本特征性、模型研发的效率和科学性。
数据层面:通过对多模态数据的采集和处理,强化对研究对象和关键问题主控特征的精细刻画与定量表征;建立面向场景问题的数据质量评估方法和噪声判识指标体系;构建指标完整、内涵清晰、噪声可控、动态更新的高质量样本数据,从孤立模态向时空对齐的多源融合升级,为智能模型研发应用夯实样本数据基础。
模型层面:基于机理-数据双驱架构和多场景(识别、评估、监测等)、多尺度、复杂非线性关系等问题特征,通过模型架构优化、规则约束等方式,建立领域知识嵌入模型的技术路径,提升智能化模型基于领域知识的自学习、可解释、可迁移和自约束能力,推动AI模型从“黑箱预测”向“白箱推理解释”升级。
该“知识-数据-模型”互促协同的地质灾害智慧研究范式,可进一步概括为“领域场景化、场景问题化、问题因果化、数据多模化、样本精准化、特征定量化,模型可解释化、人机协同化”,即:梳理提炼AI技术应用的具体防治场景和需要解决的具体问题,基于领域知识梳理、归纳导致该问题的主控因素及造成后果,围绕主控因素开展数据采集与样本制备,提升数据和样本对主控特征的精细化、定量化描述,为研发可解释性强的AI模型夯实样本基础,将人机协同模式贯穿始终,提升地质灾害防治智慧研究的效率和成效。
3.2 技术思路
在“知识-数据-模型”互促协同的地质灾害研究范式基础上,利用数据清洗、数据融合、知识对齐、特征提取与增强以及样本补充等数据处理和样本制备技术和知识嵌入、模型参数微调、建立模型自学习机制等一系列关键技术,探索建立“AI+地质灾害”融合研究的技术框架思路(图7a)。
在地质灾害智慧防治研究中,除了隐患智能识别问题,针对地质灾害隐患体潜在致灾范围的高效精准计算是提升风险防控精准度的另一项关键问题,特别是在当前规模化、高精度、多模态、高时效性的“空天对地”探测、观测数据支撑下,利用智能模型,为在区域尺度上实现对群发性中小型灾害潜在致灾范围的精准计算提供技术可能。当前围绕该问题虽已取得一定进展[32 − 33],但在区域尺度上对失稳碎屑体运移轨迹的计算方法仍难以有效融合滑源区、流动区、堆积区等局部地形特征。因此,利用智能模型破解区域尺度上滑坡致灾范围精细量化这一典型的复杂非线性关系问题,也是AI赋能地质灾害智慧防治研究和实践的一个典型场景(图7b)。以地质灾害多发频发、风险精准防控难度大的乌蒙山典型地区中小型高位崩滑碎屑灾害的致灾范围计算为例进行说明:
知识层面:区内中小型规模崩滑碎屑灾害群发多发频发[32],高效精准计算崩滑碎屑体在沟谷等复杂地形控制下的致灾范围是进行开展风险精准防控的基础。基于提出的“知识-数据-模型”互促协同范式,从地形控制(上陡下缓的“靴状”地形)、孕灾地层(飞仙关组地层和煤系地层的易滑地层组合)、构造节理特征(破碎混杂岩体内节理随机分布)等机理方面[34],提炼总结乌蒙山区中小型高位崩滑灾害的孕灾-致灾主控,为多模态的基础数据采集提供理论依据。
数据层面:针对性采集区内地形地貌、地表覆盖、地层岩性、构造节理和灾害编目等基础数据,开展灾害隐患和灾情特征(地形、地表、地质等)精细化提取,重点围绕滑源区、流动区、堆覆区开展高精度遥感解译,提取涵盖地形、地表、地质以及运移轨迹线几何属性的致灾范围样本特征;同时补充开展区内典型致灾案例数值模拟,强化对不同地形控制条件下的碎屑体运移轨迹特征提取和样本制备,进而建立一套涵盖中小型崩滑碎屑灾害发生不同地形、地表、地质等特征的致灾轨迹线样本集。
模型层面:通过对不同类型机器学习(随机森林等)或深度学习(神经网络等)模型进行训练和筛选,选定精度最好的模型并结合典型案例进行算法优化,重点提升模型算法针对典型孕灾地形(上陡下缓的“靴状”地形)、地质(飞仙关组+煤系地层;破碎混杂岩内随机分布节理)等致灾范围主控因素的表达,强化runout能量角等物理规律和易灾机理知识的嵌入,进而完成区域尺度上对灾害致灾范围的精细计算,并在数据样本和模型算法中融入局部地形地貌等特征影响。通过实地校验,开展样本特征、模型参数等优化,提升计算精度。
4. 结论
(1)大语言模型(LLMs)技术拓展了地质灾害研究的能力边界。地质灾害研究经历了经验科学、理论科学、计算科学3个范式和发展阶段,正在向第4范式数据科学演进。在数据采集能力和AI技术快速发展大背景下,借鉴数据高效利用、小样本学习、模型减重优化、人机协同融合的研究策略,将信息归集、特征挖掘、数据处理、知识嵌入、模型迁移等AI技术与地质灾害防治场景的具体问题相结合,拓展对规模化地球数据采集和处理能力的边界,提升对多维度、多尺度、复杂场景下的精细建模与模拟能力,对助力我国防灾减灾精准化、智能化发展具有重要意义。
(2)在大语言模型(LLMs)等AI技术赋能下,围绕地质灾害智慧识别研究,展望提出涵盖了“知识蒸馏与动态图谱构建阶段、多模态数据对齐与质量优化阶段、机理-数据双驱动模型构建优化阶段、服务链条贯通与决策支撑阶段”4个阶段的研究技术路径,进一步拓展地质灾害隐患智慧识别的研究思路。
(3)展望提出了“领域场景化、场景问题化、问题特征化、数据多模化、样本精准化、特征定量化,模型可解释化、人机协同化”的“AI+地质灾害”研究范式,并以区域尺度崩滑碎屑灾害致灾范围的精细计算场景为例进行分析,论述AI技术从数据、模型和知识嵌入的更深层次融入地质灾害典型场景研究的技术思路,以期更好利用AI技术发现、学习和解决地质灾害多维度、多尺度、复杂场景下的非线性问题。
(4)需要强调的是,“赋能”的本质是增强而非替代,地质灾害研究的核心和基础仍依赖于对孕灾-成灾机理的实地调查、深入认知与理论创新,需要通过扎实的基础研究构建可持续积累、不断迭代更新的地质灾害防治专业知识体系,作为“AI+地质灾害防治”的基础,通过人机协同的研究范式,将行业知识大数据不断投喂给AI系统,进而不断提升其知识厚度、思考深度、能力广度。
-
-
[1] 陈建平,辛亚波,王泽鹏,等. 样本选取对地质灾害易发性评价的影响——以山西柳林县为例[J]. 中国地质灾害与防治学报,2024,35(3):152 − 162. [CHEN Jianping,XIN Yabo,WANG Zepeng,et al. Effect of sample selection on the susceptibility assessment of geological hazards:A case study in Liulin County,Shanxi Province[J]. The Chinese Journal of Geological Hazard and Control,2024,35(3):152 − 162. (in Chinese with English abstract)] CHEN Jianping, XIN Yabo, WANG Zepeng, et al. Effect of sample selection on the susceptibility assessment of geological hazards: A case study in Liulin County, Shanxi Province[J]. The Chinese Journal of Geological Hazard and Control, 2024, 35(3): 152 − 162. (in Chinese with English abstract)
[2] 孙玉进. 岩土大变形问题的物质点法研究[D]. 北京:清华大学,2017. [SUN Yujin. Study on large deformation of rock and soil by material point method[D]. Beijing:Tsinghua University,2017. (in Chinese with English abstract)] SUN Yujin. Study on large deformation of rock and soil by material point method[D]. Beijing: Tsinghua University, 2017. (in Chinese with English abstract)
[3] 张雄,廉艳平,刘岩,等. 物质点法[M]. 北京:清华大学出版社,2013. [ZHANG Xiong,LIAN Yanping,LIU Yan,et al. Material point method[M]. Beijing:Tsinghua University Press,2013. (in Chinese)] ZHANG Xiong, LIAN Yanping, LIU Yan, et al. Material point method[M]. Beijing: Tsinghua University Press, 2013. (in Chinese)
[4] 廉艳平,张帆,刘岩,等. 物质点法的理论和应用[J]. 力学进展,2013,43(2):237 − 264. [LIAN Yanping,ZHANG Fan,LIU Yan,et al. Material point method and its applications[J]. Advances in Mechanics,2013,43(2):237 − 264. (in Chinese with English abstract)] LIAN Yanping, ZHANG Fan, LIU Yan, et al. Material point method and its applications[J]. Advances in Mechanics, 2013, 43(2): 237 − 264. (in Chinese with English abstract)
[5] YERRO C A,ALONSO P A E,PINYOL P N M. The material point method:A promising computational tool in geotechnics[C]//Challenges and Innovations in Geotechnics. 2013:853 − 856.
[6] SULSKY D,ZHOU S J,SCHREYER H L. Application of a particle-in-cell method to solid mechanics[J]. Computer Physics Communications,1995,87(1/2):236 − 252.
[7] ZIENKIEWICZ O C,SHIOMI T. Dynamic behaviour of saturated porous media; The generalized Biot formulation and its numerical solution[J]. International Journal for Numerical and Analytical Methods in Geomechanics,1984,8(1):71 − 96.
[8] VERRUIJT A. An introduction to soil dynamics[M]. DordrechtSpringer Netherlands,2010
[9] BEUTH L,BENZ T,VERMEER P A,et al. Large deformation analysis using a quasi-static material point method[J]. Journal of Theoretical and Applied Mechanics,2008,38(1 − 2):45 − 60
[10] BEUTH L,WIĘCKOWSKI Z,VERMEER P A. Solution of quasi-static large-strain problems by the material point method[J]. International Journal for Numerical and Analytical Methods in Geomechanics,2011,35(13):1451 − 1465.
[11] 黄鹏,张雄. 岩土边坡失效分析的物质点法研究:中国计算力学大会2010(CCCM2010) 暨第八届南方计算力学学术会议(SCCM8),中国四川绵阳,2010[C][143]. [HUANG Peng,ZHANG Xiong. Material point method for failure analysis of rock and soil slopes:Chinese congress of computational mechanics 2010 (CCCM2010) and the 8th Southern Conference of Computational Mechanics (SCCM8),Mianyang,Sichuan,China,2010:143. (in Chinese with English abstract)] HUANG Peng, ZHANG Xiong. Material point method for failure analysis of rock and soil slopes: Chinese congress of computational mechanics 2010 (CCCM2010) and the 8th Southern Conference of Computational Mechanics (SCCM8), Mianyang, Sichuan, China, 2010: 143. (in Chinese with English abstract)
[12] HUANG P,LI S L,GUO H,et al. Large deformation failure analysis of the soil slope based on the material point method[J]. Computational Geosciences,2015,19(4):951 − 963.
[13] LLANO-SERNA M A,FARIAS M M,PEDROSO D M. An assessment of the material point method for modelling large scale Run-out processes in landslides[J]. Landslides,2016,13(5):1057 − 1066.
[14] LI X P,WU Y,HE S M,et al. Application of the material point method to simulate the post-failure runout processes of the Wangjiayan landslide[J]. Engineering Geology,2016,212:1 − 9.
[15] 宰德志,庞锐. 基于物质点法的土体强度对边坡失稳滑动距离影响研究[J]. 水利与建筑工程学报,2021,19(5):46 − 51. [ZAI Dezhi,PANG Rui. Influence of soil strength on sliding distance of slope instability based on material point method [J]. Journal of Water Resources and Architectural Engineering,21,19(5):46 − 51. (in Chinese with English abstract)] ZAI Dezhi, PANG Rui. Influence of soil strength on sliding distance of slope instability based on material point method [J]. Journal of Water Resources and Architectural Engineering, 21, 19(5): 46 − 51. (in Chinese with English abstract)
[16] 姚云. 基于物质点法考虑剪胀特性的土质边坡稳定性分析[J]. 湖南交通科技,2017,43(2):82 − 85. [YAO Yun. Stability analysis of soil slope considering dilatancy characteristics based on material point method[J]. Hunan Communication Science and Technology,2017,43(2):82 − 85. (in Chinese with English abstract)] YAO Yun. Stability analysis of soil slope considering dilatancy characteristics based on material point method[J]. Hunan Communication Science and Technology, 2017, 43(2): 82 − 85. (in Chinese with English abstract)
[17] 杨婷婷,杨永森,邱流潮. 基于物质点法的土体流动大变形过程数值模拟[J]. 工程地质学报,2018,26(6):1463 − 1472. [YANG Tingting,YANG Yongsen,QIU Liuchao. Mpm based numerical simulation of large deformation process of soil flow[J]. Journal of Engineering Geology,2018,26(6):1463 − 1472. (in Chinese with English abstract)] YANG Tingting, YANG Yongsen, QIU Liuchao. Mpm based numerical simulation of large deformation process of soil flow[J]. Journal of Engineering Geology, 2018, 26(6): 1463 − 1472. (in Chinese with English abstract)
[18] QU C X,WANG G,FENG K W,et al. Large deformation analysis of slope failure using material point method with cross-correlated random fields[J]. Journal of Zhejiang University-Science A (Applied Physics & Engineering),2021,22(11):856 − 870.
[19] YERRO A,ALONSO E E,PINYOL N M. Run-out of landslides in brittle soils[J]. Computers and Geotechnics,2016,80:427 − 439.
[20] LIU X,WANG Y,LI D Q. Investigation of slope failure mode evolution during large deformation in spatially variable soils by random limit equilibrium and material point methods[J]. Computers and Geotechnics,2019,111:301 − 312.
[21] 刘磊磊,梁昌奇,徐蒙,等. 考虑参数旋转各向异性空间变异性的边坡大变形概率分析[J]. 地球科学,2023,48(5):1836 − 1852. [LIU Leilei,LIANG Changqi,XU Meng,et al. Probabilistic analysis of large slope deformation considering soil spatial variability with rotated anisotropy[J]. Earth Science,2023,48(5):1836 − 1852. (in Chinese with English abstract)] LIU Leilei, LIANG Changqi, XU Meng, et al. Probabilistic analysis of large slope deformation considering soil spatial variability with rotated anisotropy[J]. Earth Science, 2023, 48(5): 1836 − 1852. (in Chinese with English abstract)
[22] LIU L L,LIANG C Q,HUANG L,et al. Parametric analysis for the large deformation characteristics of unstable slopes with linearly increasing soil strength by the random material point method[J]. Computers and Geotechnics,2023,162:105661.
[23] LI X,TANG S,ZHENG Y,et al. Influence of the matrix of the soil-rock mixture on deformation and failure behaviors of the slope based on material point method[J]. Frontiers in Earth Science,2022,10:997928.
[24] 蒋涛,崔圣华,冉耀. 开挖和降雨耦合诱发滑坡机理分析——以四川万源前进广场滑坡为例[J]. 中国地质灾害与防治学报,2023,34(3):20 − 30. [JIANG Tao,CUI Shenghua,RAN Yao. Analysis of landslide mechanism induced by excavation and rainfall:A case study of the Qianjin Square landslide in Wanyuan City,Sichuan Province[J]. The Chinese Journal of Geological Hazard and Control,2023,34(3):20 − 30. (in Chinese with English abstract)] JIANG Tao, CUI Shenghua, RAN Yao. Analysis of landslide mechanism induced by excavation and rainfall: A case study of the Qianjin Square landslide in Wanyuan City, Sichuan Province[J]. The Chinese Journal of Geological Hazard and Control, 2023, 34(3): 20 − 30. (in Chinese with English abstract)
[25] 张彦博, 孙俊, 陈涛, 等. 贵州“7•23” 水城滑坡触发机制及二次滑坡动力致灾分析[J]. 中国地质灾害与防治学报,2025,36(3):18 − 26. [ZHANG Yanbo, SUN Jun, CHEN Tao, et al. Triggering mechanism and secondary landslide analyses of the “7•23” Shuicheng landslide in Guizhou[J]. The Chinese Journal of Geological Hazard and Control,2025,36(3):18 − 26. (in Chinese with English abstract)] ZHANG Yanbo, SUN Jun, CHEN Tao, et al. Triggering mechanism and secondary landslide analyses of the “7•23” Shuicheng landslide in Guizhou[J]. The Chinese Journal of Geological Hazard and Control, 2025, 36(3): 18 − 26. (in Chinese with English abstract)
[26] 顾福计,钱龙,王梦洁,等. 太行山河北段 “23•7” 强降雨引发的地质灾害规律研究[J]. 中国地质灾害与防治学报,2024,35(2):55 − 65. [GU Fuji,QIAN Long,WANG Mengjie,et al. Analysis of geological hazards caused by the “23•7” heavy rainfall in the northern section of Taihang Mountain in Hebei Province[J]. The Chinese Journal of Geological Hazard and Control,2024,35(2):55 − 65. (in Chinese with English abstract)] GU Fuji, QIAN Long, WANG Mengjie, et al. Analysis of geological hazards caused by the “23•7” heavy rainfall in the northern section of Taihang Mountain in Hebei Province[J]. The Chinese Journal of Geological Hazard and Control, 2024, 35(2): 55 − 65. (in Chinese with English abstract)
[27] QIN J Y,MEI G,XU N X. Meshfree methods in geohazards prevention:A survey[J]. Archives of Computational Methods in Engineering,2022,29(5):3151 − 3182.
[28] WANG B,VARDON P J,HICKS M A. Rainfall-induced slope collapse with coupled material point method[J]. Engineering Geology,2018,239:1 − 12.
[29] JASSIM I,STOLLE D,VERMEER P. Two-phase dynamic analysis by material point method[J]. International Journal for Numerical and Analytical Methods in Geomechanics,2013,37(15):2502 − 2522.
[30] GIRARDI V,YERRO A,CECCATO F,et al. Modelling deformations in water retention structures with unsaturated material point method[J]. Proceedings of the Institution of Civil Engineers - Geotechnical Engineering,2021,174(5):577 − 592.
[31] HIGO Y,OKA F,KIMOTO S,et al. A coupled MPM-FDM analysis method for multi-phase elasto-plastic soils[J]. Soils and Foundations,2010,50(4):515 − 532.
[32] ZHENG X C,PISANÒ F,VARDON P J,et al. Fully implicit,stabilised MPM simulation of large-deformation problems in two-phase elastoplastic geomaterials[J]. Computers and Geotechnics,2022,147:104771.
[33] ABE K,SOGA K,BANDARA S. Material point method for coupled hydromechanical problems[J]. Journal of Geotechnical and Geoenvironmental Engineering,2014,140(3):04013033.
[34] SOŁOWSKI W T,SEYEDAN S. Granular material point method:unsaturated soil modelling[J]. Geomechanics for Energy and the Environment,2023,34:100471.
[35] ZHAN Z Q,ZHOU C,LIU C Q,et al. Modelling hydro-mechanical coupled behaviour of unsaturated soil with two-phase two-point material point method[J]. Computers and Geotechnics,2023,155:105224.
[36] FENG K W,HUANG D R,WANG G. Two-layer material point method for modeling soil–water interaction in unsaturated soils and rainfall-induced slope failure[J]. Acta Geotechnica,2021,16(8):2529 − 2551.
[37] CECCATO F,YERRO A,GIRARDI V,et al. Two-phase dynamic MPM formulation for unsaturated soil[J]. Computers and Geotechnics,2021,129:103876.
[38] TRAN Q A,SOŁOWSKI W. Generalized interpolation material point method modelling of large deformation problems including strain-rate effects - Application to penetration and progressive failure problems[J]. Computers and Geotechnics,2019,106:249 − 265.
[39] 王升,曾鹏,李天斌,等. 土质滑坡失稳、运动及冲击压力物质点法模拟研究[J]. 工程地质学报,2022,30(4):1362 − 1370. [WANG Sheng,ZENG Peng,LI Tianbin,et al. Initiation,movement and impact simulation of soil landslide with material point method[J]. Journal of Engineering Geology,2022,30(4):1362 − 1370. (in Chinese with English abstract)] WANG Sheng, ZENG Peng, LI Tianbin, et al. Initiation, movement and impact simulation of soil landslide with material point method[J]. Journal of Engineering Geology, 2022, 30(4): 1362 − 1370. (in Chinese with English abstract)
[40] WANG D,WANG B,YUAN W H,et al. Investigation of rainfall intensity on the slope failure process using GPU-accelerated coupled MPM[J]. Computers and Geotechnics,2023,163:105718.
[41] HE K,XI C J,LIU B,et al. MPM-based mechanism and runout analysis of a compound reactivated landslide[J]. Computers and Geotechnics,2023,159:105455.
[42] LEE W L,MARTINELLI M,SHIEH C L. An investigation of rainfall-induced landslides from the pre-failure stage to the post-failure stage using the material point method[J]. Frontiers in Earth Science,2021,9:764393.
[43] MOORMANN C,HAMAD F. MPM dynamic simulation of a seismically induced sliding mass[J]. IOP Conference Series:Earth and Environmental Science,2015,26:012024.
[44] ALSARDI A,YERRO A. Runout modeling of earthquake-triggered landslides with the material point method[C]//IFCEE 2021. Dallas,Texas. Reston,VA:American Society of Civil Engineers,2021:21–31.
[45] ALSARDI A,COPANA J,YERRO A. Modelling earthquake-triggered landslide runout with the material point method[J]. Proceedings of the Institution of Civil Engineers - Geotechnical Engineering,2021,174(5):563 − 576.
[46] HE M C,RIBEIRO E SOUSA L,MÜLLER A,et al. Numerical and safety considerations about the Daguangbao landslide induced by the 2008 Wenchuan earthquake[J]. Journal of Rock Mechanics and Geotechnical Engineering,2019,11(5):1019 − 1035.
[47] XU X R,JIN F,SUN Q C,et al. Three-dimensional material point method modeling of runout behavior of the Hongshiyan landslide[J]. Canadian Geotechnical Journal,2019,56(9):1318 − 1337.
[48] FENG K W,HUANG D R,WANG G,et al. Physics-based large-deformation analysis of coseismic landslides:A multiscale 3D SEM-MPM framework with application to the hongshiyan landslide[J]. Engineering Geology,2022,297:106487.
[49] FENG K W,WANG G,HUANG D R,et al. Material point method for large-deformation modeling of coseismic landslide and liquefaction-induced dam failure[J]. Soil Dynamics and Earthquake Engineering,2021,150:106907.
[50] BIELAK J. Domain reduction method for three-dimensional earthquake modeling in localized regions,part I:theory[J]. Bulletin of the Seismological Society of America,2003,93(2):817 − 824.
[51] YOSHIMURA C. Domain reduction method for three-dimensional earthquake modeling in localized regions,part II:Verification and applications[J]. Bulletin of the Seismological Society of America,2003,93(2):825 − 841.
[52] KOHLER M,STOECKLIN A,PUZRIN A M. A MPM framework for large-deformation seismic response analysis[J]. Canadian Geotechnical Journal,2022,59(6):1046 − 1060.
[53] KOHLER M,HODEL D,KELLER L,et al. Case study of an active landslide at the flank of a water reservoir and its response during earthquakes[J]. Engineering Geology,2023,323:107243.
[54] KOHLER M,PUZRIN A M. Mechanism of co-seismic deformation of the slow-moving La sorbella landslide in Italy revealed by MPM analysis[J]. Journal of Geophysical Research (Earth Surface),2022,127(7):e2022JF006618.
[55] TRAN Q A,GRIMSTAD G,ALI GHOREISHIAN AMIRI S. MPMICE:A hybrid MPM-CFD model for simulating coupled problems in porous media. Application to earthquake-induced submarine landslides[J]. International Journal for Numerical Methods in Engineering,2024,125(5):e7383.
[56] FERNÁNDEZ F,VARGAS E,MULLER A L,et al. Material point method modeling in 3D of the failure and run-out processes of the Daguangbao landslide[J]. Acta Geotechnica,2024,19(7):4277 − 4296.
[57] 吴方东,张巍,史卜涛,等. 堆载诱发型土质滑坡运动特征物质点法模拟[J]. 水文地质工程地质,2017,44(6):126 − 134. [WU Fangdong,ZHANG Wei,SHI Butao,et al. Run-out characteristic simulation of a surcharge-induced soil landslide using the material point method[J]. Hydrogeology & Engineering Geology,2017,44(6):126 − 134. (in Chinese with English abstract)] WU Fangdong, ZHANG Wei, SHI Butao, et al. Run-out characteristic simulation of a surcharge-induced soil landslide using the material point method[J]. Hydrogeology & Engineering Geology, 2017, 44(6): 126 − 134. (in Chinese with English abstract)
[58] XIE T C,ZHU H H,ZHANG C X,et al. Modeling strip footings on slopes using the material point method[J]. Bulletin of Engineering Geology and the Environment,2023,82(4):99.
[59] ZHU H H,XIE T C,ZHANG W,et al. Numerical simulations of a strip footing on the soil slope with a buried pipe using the material point method[J]. International Journal of Geomechanics,2023,23(11):04023190.
[60] TRONCONE A,PUGLIESE L,CONTE E. Analysis of an excavation-induced landslide in stiff clay using the material point method[J]. Engineering Geology,2022,296:106479.
[61] GONZÁLEZ ACOSTA J L,VARDON P J,HICKS M A. Study of landslides and soil-structure interaction problems using the implicit material point method[J]. Engineering Geology,2021,285:106043.
[62] NG C W W,WANG C,CHOI C E,et al. Effects of barrier deformability on load reduction and energy dissipation of granular flow impact[J]. Computers and Geotechnics,2020,121:103445.
[63] CECCATO F,REDAELLI I,DI PRISCO C,et al. Impact forces of granular flows on rigid structures:comparison between discontinuous (DEM) and continuous (MPM) numerical approaches[J]. Computers and Geotechnics,2018,103:201 − 217.
[64] WYSER E,ALKHIMENKOV Y,JABOYEDOFF M,et al. Analytical and numerical solutions for three-dimensional granular collapses[J]. Geosciences,2023,13(4):119.
[65] LEI Z D,WU B S,WU S S,et al. A material point-finite element (MPM-FEM) model for simulating three-dimensional soil-structure interactions with the hybrid contact method[J]. Computers and Geotechnics,2022,152:105009.
[66] LEI X Q,CHEN X Q,YANG Z J,et al. A simple and robust MPM framework for modelling granular flows over complex terrains[J]. Computers and Geotechnics,2022,149:104867.
[67] CUOMO S,DI PERNA A,MARTINELLI M. Material point method (MPM) hydro-mechanical modelling of flows impacting rigid walls[J]. Canadian Geotechnical Journal,2021,58(11):1730 − 1743.
[68] LI X P,YAN Q W,ZHAO S X,et al. Investigation of influence of baffles on landslide debris mobility by 3D material point method[J]. Landslides,2020,17(5):1129 − 1143.
[69] ABE K,KONAGAI K. Numerical simulation for runout process of debris flow using depth-averaged material point method[J]. Soils and Foundations,2016,56(5):869 − 888.
[70] ZHAO Y D,CHOO J,JIANG Y P,et al. Coupled material point and level set methods for simulating soils interacting with rigid objects with complex geometry[J]. Computers and Geotechnics,2023,163:105708.
[71] LI X P,YAO J,SUN Y L,et al. Material point method analysis of fluid–structure interaction in geohazards[J]. Natural Hazards,2022,114(3):3425 − 3443.
[72] DI PERNA A,CUOMO S,MARTINELLI M. Empirical formulation for debris flow impact and energy release[J]. Geoenvironmental Disasters,2022,9(1):8.
[73] CUOMO S,DI PERNA A,MARTINELLI M. Analytical and numerical models of debris flow impact[J]. Engineering Geology,2022,308:106818.
[74] KAZMI Z A,KONAGAI K,IKEDA T. Field measurements and numerical simulation of debris flows from dolomite slopes destabilized during the 2005 Kashmir earthquake,Pakistan[J]. Journal of Earthquake Engineering,2014,18(3):364-388.
[75] WONG T,WEI Y J,JIE Y X. Global sensitivity analysis on debris flow energy dissipation of the artificial step-pool system[J]. Computers and Geotechnics,2022,147:104758.
[76] VICARI H,TRAN Q A,NORDAL S,et al. MPM modelling of debris flow entrainment and interaction with an upstream flexible barrier[J]. Landslides,2022,19(9):2101 − 2115.
[77] XIE X C,CECCATO F,ZHOU M L,et al. Hydro-mechanical behaviour of soils during water-soil gushing in shield tunnels using MPM[J]. Computers and Geotechnics,2022,145:104688.
[78] 谢小创,张冬梅. 基于物质点法的盾构隧道涌水涌砂模拟与应用[J]. 隧道建设(中英文),2023,43(6):1045 − 1056. [XIE Xiaochuang,ZHANG Dongmei. Simulation of water-soil gushing in shield tunnel based on material point method and its application[J]. Tunnel Construction,2023,43(6):1045 − 1056. (in Chinese with English abstract)] XIE Xiaochuang, ZHANG Dongmei. Simulation of water-soil gushing in shield tunnel based on material point method and its application[J]. Tunnel Construction, 2023, 43(6): 1045 − 1056. (in Chinese with English abstract)
[79] 张春新,朱鸿鹄,李豪杰,等. 支护压力控制下隧道周围砂土变形破坏物质点法模拟[J]. 浙江大学学报(工学版),2021,55(7):1317 − 1326. [ZHANG Chunxin,ZHU Honghu,LI Haojie,et al. Material point method simulations of sand deformation and failure around tunnel controlled by support pressure[J]. Journal of Zhejiang University (Engineering Science),2021,55(7):1317 − 1326. (in Chinese with English abstract)] ZHANG Chunxin, ZHU Honghu, LI Haojie, et al. Material point method simulations of sand deformation and failure around tunnel controlled by support pressure[J]. Journal of Zhejiang University (Engineering Science), 2021, 55(7): 1317 − 1326. (in Chinese with English abstract)
[80] 王曼灵,李树忱,周慧颖,等. 基于改进对流粒子域插值物质点法的隧道大变形分析[J]. 岩土工程学报,2024,46(8):1632 − 1643. [WANG Manling,LI Shuchen,ZHOU Huiying,et al. Improved convective particle domain interpolation material point method for large deformation analysis of tunnels[J]. Chinese Journal of Geotechnical Engineering,2024,46(8):1632 − 1643. (in Chinese with English abstract)] WANG Manling, LI Shuchen, ZHOU Huiying, et al. Improved convective particle domain interpolation material point method for large deformation analysis of tunnels[J]. Chinese Journal of Geotechnical Engineering, 2024, 46(8): 1632 − 1643. (in Chinese with English abstract)
[81] 王桂林,余浩,翟俊,等. 带裂损隧道在爆炸作用下的二次损伤响应规律[J]. 高压物理学报,2023,37(5):197 − 208. [WANG Guilin,YU Hao,ZHAI Jun,et al. Secondary damage response of cracked tunnels under explosion[J]. Chinese Journal of High Pressure Physics,2023,37(5):197 − 208. (in Chinese with English abstract)] WANG Guilin, YU Hao, ZHAI Jun, et al. Secondary damage response of cracked tunnels under explosion[J]. Chinese Journal of High Pressure Physics, 2023, 37(5): 197 − 208. (in Chinese with English abstract)
[82] TU S Q,LI W,ZHANG C P,et al. Face stability analysis of tunnels in saturated soil considering soil-fluid coupling effect via material point method[J]. Computers and Geotechnics,2023,161:105592.
[83] CHENG X S,ZHENG G,SOGA K,et al. Post-failure behavior of tunnel heading collapse by MPM simulation[J]. Science China Technological Sciences,2015,58(12):2139 − 2152.
[84] LI S,ZHANG Y,WU J C,et al. Numerical modeling of pumping-induced earth fissures using coupled quasi-static material point method[J]. Journal of Geotechnical and Geoenvironmental Engineering,2023,149(9):04023065.
[85] 李克智,边树涛,郭晓辉. 应用物质点法实现非常规油气藏水平井压前模拟——以鄂尔多斯盆地杭锦旗区块为例[J]. 石油地质与工程,2023,37(6):109 − 113. [LI Kezhi,BIAN Shutao,GUO Xiaohui. Pre-pressure simulation of horizontal wells in conventional oil and gas reservoirs with material point method:By taking Hangjinqi Block in Ordos Basin as an example[J]. Petroleum Geology and Engineering,2023,37(6):109 − 113. (in Chinese with English abstract)] LI Kezhi, BIAN Shutao, GUO Xiaohui. Pre-pressure simulation of horizontal wells in conventional oil and gas reservoirs with material point method: By taking Hangjinqi Block in Ordos Basin as an example[J]. Petroleum Geology and Engineering, 2023, 37(6): 109 − 113. (in Chinese with English abstract)
[86] LIANG D F,ZHAO X Y,SOGA K. Simulation of overtopping and seepage induced dike failure using two-point MPM[J]. Soils and Foundations,2020,60(4):978 − 988.
[87] TALBOT L E D,GIVEN J,TJUNG E Y S,et al. Modeling large-deformation features of the Lower San Fernando Dam failure with the material point method[J]. Computers and Geotechnics,2024,165:105881.
[88] 徐云卿,周晓敏,赵世一,等. 基于B样条物质点法的溃坝流模拟研究[J]. 应用数学和力学,2023,44(8):921 − 930. [XU Yunqing,ZHOU Xiaomin,ZHAO Shiyi,et al. Simulation study on dam break flow based on the B-spline material point method[J]. Applied Mathematics and Mechanics,2023,44(8):921 − 930. (in Chinese with English abstract)] XU Yunqing, ZHOU Xiaomin, ZHAO Shiyi, et al. Simulation study on dam break flow based on the B-spline material point method[J]. Applied Mathematics and Mechanics, 2023, 44(8): 921 − 930. (in Chinese with English abstract)
[89] REMMERSWAAL G,VARDON P J,HICKS M A. Evaluating residual dyke resistance using the random material point method[J]. Computers and Geotechnics,2021,133:104034.
[90] ZHAO X Y,LIANG D F,MARTINELLI M. Numerical simulations of dam-break floods with MPM[J]. Procedia Engineering,2017,175:133 − 140.
[91] 冯晓青,叶斌,苗沪生,等. 基于固液两相耦合物质点法的饱和地基液化数值计算方法[J]. 哈尔滨工业大学学报,2024,56(11):15 − 25. [FENG Xiaoqing,YE Bin,MIAO Husheng,et al. Numerical calculation method of saturated ground liquefaction based on solid-liquid two-phase coupling material point method[J]. Journal of Harbin Institute of Technology,2024,56(11):15 − 25. (in Chinese with English abstract)] FENG Xiaoqing, YE Bin, MIAO Husheng, et al. Numerical calculation method of saturated ground liquefaction based on solid-liquid two-phase coupling material point method[J]. Journal of Harbin Institute of Technology, 2024, 56(11): 15 − 25. (in Chinese with English abstract)
[92] LIANG W J,ZHAO J D,WU H R,et al. Multiscale modeling of anchor pullout in sand[J]. Journal of Geotechnical and Geoenvironmental Engineering,2021,147(9):04021091.
[93] 高宇新,朱鸿鹄,张春新,等. 砂土中锚板上拔三维物质点法模拟研究[J]. 岩土工程学报,2022,44(2):295 − 304. [GAO Yuxin,ZHU Honghu,ZHANG Chunxin,et al. Three-dimensional uplift simulation of anchor plates in sand using material point method[J]. Chinese Journal of Geotechnical Engineering,2022,44(2):295 − 304. (in Chinese with English abstract)] GAO Yuxin, ZHU Honghu, ZHANG Chunxin, et al. Three-dimensional uplift simulation of anchor plates in sand using material point method[J]. Chinese Journal of Geotechnical Engineering, 2022, 44(2): 295 − 304. (in Chinese with English abstract)