Multi-hazard susceptibility mapping based on Convolutional Neural Networks

Multi-hazard susceptibility prediction is an important component of disasters risk management plan.An effective multi-hazard risk mitigation strategy includes assessing individual hazards as well as their interactions.However,with the rapid development of artificial intelligence technology,multi-hazard susceptibility prediction techniques based on machine learning has encountered a huge bottleneck.In order to effectively solve this problem,this study proposes a multi-hazard susceptibility mapping framework using the classical deep learning algorithm of Convolutional Neural Networks(CNN).First,we use historical flash flood,debris flow and landslide locations based on Google Earth images,extensive field surveys,topography,hydrology,and environmental data sets to train and validate the proposed CNN method.Next,the proposed CNN method is assessed in comparison to conventional logistic regression and k-nearest neighbor methods using several objective criteria,i.e.,coefficient of determination,overall accuracy,mean absolute error and the root mean square error.Experimental results show that the CNN method outperforms the conventional machine learning algorithms in predicting probability of flash floods,debris flows and landslides.Finally,the susceptibility maps of the three hazards based on CNN are combined to create a multi-hazard susceptibility map.It can be observed from the map that 62.43%of the study area are prone to hazards,while 37.57%of the study area are harmless.In hazard-prone areas,16.14%,4.94%and 30.66%of the study area are susceptible to flash floods,debris flows and landslides,respectively.In terms of concurrent hazards,0.28%,7.11%and 3.13%of the study area are susceptible to the joint occurrence of flash floods and debris flow,debris flow and landslides,and flash floods and landslides,respectively,whereas,0.18%of the study area is subject to all the three hazards.The results of this study can benefit engineers,disaster managers and local government officials involved in sustainable land management and disaster risk mitigation.

Is multi-hazard mapping effective in assessing natural hazards and integrated watershed management?

Natural hazards are often studied in isolation.However,there is a great need to examine hazards holistically to better manage the complex of threats found in any region.Many regions of the world have complex hazard landscapes wherein risk from individual and/or multiple extreme events is omnipresent.Extensive parts of Iran experience a complex array of natural hazards-floods,earthquakes,landslides,forest fires,subsidence,and drought.The effectiveness of risk mitigation is in part a function of whether the complex of hazards can be collectively considered,visualized,and evaluated.This study develops and tests individual and collective multihazard risk maps for floods,landslides,and forest fires to visualize the spatial distribution of risk in Fars Province,southern Iran.To do this,two well-known machine-learning algorithms-SVM and MARS-are used to predict the distribution of these events.Past floods,landslides,and forest fires were surveyed and mapped.The locations of occurrence of these events(individually and collectively) were randomly separated into training(70%) and testing(30%) data sets.The conditioning factors(for floods,landslides,and forest fires) employed to model the risk distributions are aspect,elevation,drainage density,distance from faults,geology,LULC,profile curvature,annual mean rainfall,plan curvature,distance from man-made residential structures,distance from nearest river,distance from nearest road,slope gradient,soil types,mean annual temperature,and TWI.The outputs of the two models were assessed using receiver-operating-characteristic(ROC) curves,true-skill statistics(TSS),and the correlation and deviance values from each models for each hazard.The areas-under-the-curves(AUC) for the MARS model prediction were 76.0%,91.2%,and 90.1% for floods,landslides,and forest fires,respectively.Similarly,the AUCs for the SVM model were 75.5%,89.0%,and 91.5%.The TSS reveals that the MARS model was better able to predict landslide risk,but was less able to predict flood-risk patterns and forest-fire risk.Finally,the combination of flood,forest fire,and landslide risk maps yielded a multi-hazard susceptibility map for the province.The better predictive model indicated that 52.3% of the province was at-risk for at least one of these hazards.This multi-hazard map may yield valuable insight for land-use planning,sustainable development of infrastructure,and also integrated watershed management in Fars Province.

Multi-hazard performance assessment of a transfer-plate high-rise building

Many urban areas are located in regions of moderate seismicity and are subjected to strong wind. Buildings in these regions are often designed without seismic provisions. As a result, in the event of an earthquake, the potential for damage and loss of lives may not be known. In this paper, the performance of a typical high-rise building with a thick transfer plate （TP）, which is one type of building structure commonly found in Hong Kong, is assessed against both earthquake and wind hazards. Seismic- and wind-resistant performance objectives are first reviewed based on relevant codes and design guidelines for high-rise buildings. After a brief introduction of wind-resistant design of the building, various methodologies, including equivalent static load analysis （ESLA）, response spectrum analysis （RSA）, pushover analysis （POA）, linear and nonlinear time-history analysis （LTHA and NTHA）, are employed to assess the seismic performance of the building when subjected to frequent earthquakes, design based earthquakes and maximum credible earthquakes. The effects of design wind and seismic action with a common 50-year return period are also compared. The results indicate that most performance objectives can be satisfied by the building, but there are some objectives, such as inter-story drift ratio, that cannot be achieved when subjected to the frequent earthquakes. It is concluded that in addition to wind, seismic action may need to be explicitly considered in the design of buildings in regions of moderate seismicity.

Towards establishing practical multi-hazard bridge design limit states

In the U.S., the current Load and Resistance Factor Design （LRFD） Specifications for highway bridges is a reliability-based formulation that considers failure probabilities of bridge components due to the actions of typical dead load and frequent vehicular loads. Various extreme load effects, such as earthquake and vessel collision, are on the same reliability-based platform. Since these extreme loads are time variables, combining them with not considered frequent. non- extreme loads is a significant challenge. The number of design limit state equations based on these failure probabilities can be unrealistically large and unnecessary from the view point of practical applications. Based on the opinion of AASHTO State Bridge Engineers, many load combinations are insignificant in their states. This paper describes the formulation of a criterion to include only the necessary load combinations to establish the design limit states. This criterion is established by examining the total failure probabilities for all possible time-invariant and time varying load combinations and breaking them down into partial terms. Then, important load combinations can be readily determined quantitatively,

Multi-Hazard Evaluation Using Cluster Analysis—For Designated Evacuation Centers of Yokohama

Hazard maps are usually prepared for each disaster, including seismic hazard maps, flood hazard maps, and landslide hazard maps. However, when the general public attempts to check their own disaster risk, most are likely not aware of the specific types of disaster. So, first of all, we need to know what kinds of hazards are important. However, the information that integrates multiple hazards is not well maintained, and there are few such studies. On the other hand, in Japan, a lot of hazard information is being released on the Internet. So, we summarized and assessed hazard data that can be accessed online regarding shelters (where evacuees live during disasters) and their catchments (areas assigned to each shelter) in Yokohama City, Kanagawa Prefecture. Based on the results, we investigated whether a grouping by cluster analysis would allow for multi-hazard assessment. We used four natural disasters (seismic, flood, tsunami, sediment disaster) and six parameters of other population and senior population. However, since the characteristics of the population and the senior population were almost the same, only population data was used in the final examination. From the cluster analysis, it was found that it is appropriate to group the designated evacuation centers in Yokohama City into six groups. In addition, each of the six groups was found to have explainable characteristics, confirming the effectiveness of multi-hazard creation using cluster analysis. For example, we divided, all hazards are low, both flood and Seismic hazards are high, sediment hazards are high, etc. In many Japanese cities, disaster prevention measures have been constructed in consideration of ground hazards, mainly for earthquake disasters. In this paper, we confirmed the consistency between the evaluation results of the multi-hazard evaluated here and the existing ground hazard map and examined the usefulness of the designated evacuation center. Finally, the validity was confirmed by comparing this result with the ground hazard based on the actual measurement by the past research. In places where the seismic hazard is large, the two are consistent with the fact that the easiness of shaking by actual measurement is also large.

Resilience of city underground infrastructure under multi-hazards impact:From structural level to network level

Underground infrastructure(UI)plays a great important role in the urbanization and modernization of megacities in the world.However,the intensive development of the UI during the past decades has posed great risks to the safety of city infrastructures under the impact of multi-hazards,especially with the condition of global climate change.In this paper,a general conceptualized framework to assess the resilience of UI in cities under multihazards impact is proposed.The urban tunnel system,e.g.,metro tunnel,road tunnel etc.,is selected as the typical underground infrastructure discussed with the emphasis both on the structural level in terms of mechanical behaviors and system level in terms of network efficiency.The hazards discussed in this paper include the natural hazards and human-related ones,with emphasis on earthquake,flood,and aggressive disturbances caused by human activities.After the general framework proposed for resilience of the structural and network behavior of the UI,two application examples are illustrated.The structural resilience of the shield tunnel under earthquake impact is analyzed by using the proposed resilience model,and the network resilience of the road tunnel system under the flood impact due to climate change is analyzed,respectively.The resilience enhancement by using the adaptive design strategy of real-time observational method is mathematically presented in this case.Some other practical engineering recovery measures are briefly discussed at the end of this application example.The findings in the application examples should be helpful to enhance the resilience-based design of the structural and network of tunnels from the component to the system level.

Integration of Probabilistic and Multi-Hazard Risk Assessment Within Urban Development Planning and Emergency Preparedness and Response: Application to Manizales,Colombia

The details of a multi-hazard and probabilistic risk assessment, developed for urban planning and emergency response activities in Manizales, Colombia, are presented in this article. This risk assessment effort was developed under the framework of an integral disaster risk management project whose goal was to connect risk reduction activities by using open access and state-of-theart risk models. A probabilistic approach was used for the analysis of seismic, landslide, and volcanic hazards to obtain stochastic event sets suitable for probabilistic loss estimation and to generate risk results in different metrics after aggregating in a rigorous way the losses associated to the different hazards. Detailed and high resolution exposure databases were used for the building stock and infrastructure of the city together with a set of vulnerability functions for each of the perils considered. The urban and territorial ordering plan of the city was updated for socioeconomic development and land use using the hazard and risk inputs and determinants, which cover not only the current urban area but also those adjacent areas where the expansion of Manizales is expected to occur. The emergency response capabilities of the city were improved by taking into account risk scenarios and after updating anautomatic and real-time post-earthquake damage assessment.

Multi-hazard disaster scenario method and emergency management for urban resilience by integrating experiment-simulation-field data

Due to the frequent occurrence of multi-hazard disasters worldwide in recent years,effective multi-hazard sce-nario analysis is imperative for disaster rescue and emergency management.The response procedure for different single hazards were investigated and formulated before.However,the investigations of disaster scenario rarely systematically address the entire development and response process of multi-hazards,including the coupling mechanisms,evolution dynamics,scenario assessment and emergency response.To this end,this paper presents our methodology of multi-hazard disaster scenario that integrates experiment-simulation-field data,focusing on three dimensions consisting of multi-hazard coupling,structures and systems,and emergency management.The newly proposed scenario method mainly comprises three aspects:experiments and simulations,multi-hazard field investigation,scenario analysis and response.Specifically,in order to study the large-scale,high-intensity and multi-hazard coupling effects,we carried out reduced-scale experiments and field measurement experiments to develop experimental similarity theory and prototype simulations of multi-hazards.In addition,a variety of field rescue and survey equipment,such as robots,Unmanned Aerial Vehicle(UAV),and Virtual Reality/Augmented Reality(VR/AR)technologies were utilized to acquire real-time data of multi-hazard field.Furthermore,we also examine the mechanism and framework of multi-hazard scenarios to formulate the detailed procedures of man-agement and response.They are incorporated with the experiments,simulations,field data and models to con-struct a new scenario platform.The proposed scenario method was applied in a case study of the coupled wind and snow multi-hazard to verify its effectiveness.The new method contributes to the disaster relief,decision-making and emergency management for multi-hazard disaster to improve the urban resilience.

A Hybrid Multi-Hazard Susceptibility Assessment Model for a Basin in Elazig Province,Türkiye

Preparation of accurate and up-to-date susceptibility maps at the regional scale is mandatory for disaster mitigation,site selection,and planning in areas prone to multiple natural hazards.In this study,we proposed a novel multi-hazard susceptibility assessment approach that combines expert-based and supervised machine learning methods for landslide,flood,and earthquake hazard assessments for a basin in Elazig Province,Türkiye.To produce the landslide susceptibility map,an ensemble machine learning algorithm,random forest,was chosen because of its known performance in similar studies.The modified analytical hierarchical process method was used to produce the flood susceptibility map by using factor scores that were defined specifically for the area in the study.The seismic hazard was assessed using ground motion parameters based on Arias intensity values.The univariate maps were synthesized with a Mamdani fuzzy inference system using membership functions designated by expert.The results show that the random forest provided an overall accuracy of 92.3%for landslide susceptibility mapping.Of the study area,41.24%were found prone to multi-hazards(probability value>50%),but the southern parts of the study area are more susceptible.The proposed model is suitable for multi-hazard susceptibility assessment at a regional scale although expert intervention may be required for optimizing the algorithms.

WARNINGS ON TROPICAL CYCLONE FOR WMO GLOBAL MULTI-HAZARD ALERT SYSTEM

The World Meteorological Organization(WMO) is planning to implement a Global Multi-hazard Alert System(GMAS) to aggregate official warning^1 information issued by authorities around the world and to serve as a one-stop shop to support the humanitarian organizations of the United Nations(UN), National Meteorological and Hydrological Services(NMHSs) and other global users including the media. It aims to enhance the authority and visibility of NMHSs and other alerting authorities. To aid effective dissemination of warnings to GMAS, the Common Alerting Protocol(CAP) was considered as a standard and robust format to use. In respect of GMAS infrastructure, the World Weather Information(WWIS) and the Severe Weather Information Centre(SWIC) of WMO as well as the WMO Alert Hub now being implemented are identified as core components, among others. The SWIC is being upgraded with GIS capability for displaying authoritative warnings and tropical cyclone(TC) information, and for use as a display platform of GMAS. Apart from warnings from NMHSs, authoritative TC warnings and advisories issued by Regional Specialized Meteorological Centres(RSMCs) and Tropical Cyclone Warning Centres(TCWCs) are also indispensable information for GMAS. As the existing TC warnings and advisories, now more or less in free text format, are mainly targeted for human users and are not intended for automatic parsing by computer software, it is proposed to make available the TC advisories in a machine-readable format so that TC information can be effectively ingested into GMAS and made available to the UN humanitarian organizations, NMHSs and other global users. In this respect, some enhancement measures to TC advisories are proposed. This calls for active collaboration of Members of the Typhoon Committee in the GMAS project.

Preliminary identification of earthquake triggered multi-hazard and risk in Pleret Sub-District(Yogyakarta,Indonesia)

Yogyakarta is one of the large cities in Central Java,located on Java Island,Indonesia.The city,and the Pleret sub-district,where the study has taken place,is prone to earthquake hazards,because it is close to several seismically active zones,such as the Sunda Megathrust and the active fault known as the Opak Fault.Since a devastating earthquake of 2006,the population of the Pleret sub-district has increased significantly.Thus,the housing demand has increased,and so is the pace of low-cost housing that does not meet earthquake-safety requirements,and furthermore are often located on unstable slopes.The local alluvial material covering a jigsaw of unstable blocks and complex slope is conditions that can amplify the negative impacts of earthquakes.Within this context,this study is aiming to assess the multi-hazards and risks of earthquakes and related secondary hazards such as ground liquefaction,and coseismic landslides.To achieve this,we used geographic information systems and remote sensing methods supplemented with outcrop study and existing seismic data to derive shear-strain parameters.The results have revealed the presence of numerous uncharted active faults with movements visible from imagery and outcrops.show that the middle part of the study area has a complex geological structure,indicated by many unchartered faults in the outcrops.Using this newly mapped blocks combined with shear strain data,we reassessed the collapse probability of buildings that reach level>0.75 near the Opak River,in central Pleret sub-district.Classifying the buildings and from population distribution,we could determine that the highest risk was during nighttime as the buildings susceptible to fall are predominantly housing buildings.The secondary hazards follow a slightly different distribution with a concentration of risks in the West.

Mapping the multi-hazards risk index for coastal block of Sundarban,India using AHP and machine learning algorithms

Global climate change,climate extremes,and overuse of natural resources are all major contributors to the risk brought on by cyclones.In I West Bengal state of India,the Pathar Pratima Block frequently experiences a variety of risks that result in significant loss of life and livelihood.In order to govern coastal society,it is crucial to measure and map the multi-hazards risk status.To depict the multi-hazards vulnerability and risk status,no cutting-edge models are currently being applied.Predicting distinct physical vulnerabilities is possible using a variety of cutting-edge machine learning techniques.This study set out to precisely describe multi-hazard risk using powerful machine learning methods.This study involved the use of Analytic Hierarchical Analysis and two cutting-edge machine-learning algorithms-Random Forest and Artificial Neural Network,which are yet underutilized in this area.The multi-hazards risk was determined by taking into account six criteria.The southern and eastern regions of the research area are clearly identified by the multi-hazards risk maps as having high to extremely high hazards risk levels.Cyclonic hazards and embankment breaching are the main dominant factors among the multi-hazards.The machine learning approach is the most accurate model for mapping the multi-hazards risk where the ROC result of Random forest and artificial neural network is more than the conventional method AHP.Here RF is the most validated model than the other two.The effectiveness,root mean square error,true skill statistics,Friedman and Wilcoxon rank test,and area under the curve of receiver operating characteristic tests were used to evaluate the prediction capacity of newly constructed models.The RMSE values of 0.24 and 0.26,TSS values of 0.82 and 0.73,and AUC values of 88.20%and 89.10%as produced by RF and ANN models,respectively,were all excellent.

地震和风耦合作用下超高层连体结构易损性分析

为探究超高层连体结构在多灾害耦合作用下的设计与安全性能,以苏州超塔为例,开展地震和风耦合作用下的多灾害易损性分析.根据设计资料建立结构有限元模型,分别开展结构在地震作用下以及地震和风耦合作用下的易损性分析,探究结构在多灾害耦合作用下的响应规律.结果表明,地震和风耦合作用下结构联合易损性随着地震动和风荷载强度的增大而增加,且地震动对结构的破坏占主导作用.地震与风耦合作用时结构达到中等破坏、严重破坏和倒塌破坏的概率均比仅地震作用时有所增加.当风速达到40 m/s时,结构在罕遇地震下达到3种破坏状态的概率分别为99.77%、91.56%和46.54%,相比于仅罕遇地震作用下分别提高了1.11%、10.73%和14.65%,相比于一般高层连体结构在同样的灾害耦合作用下分别提高了0.27%、6.26%和14.34%.

西部矿区采动损害及减损开采的地质保障技术框架体系

【目的】煤炭在我国的主体能源地位短期内难以发生改变,是我国能源发展的兜底保障。煤炭资源大规模、高强度的开采损害煤矿区地质环境,同时导致突水、冲击地压等灾害。【方法】围绕煤矿安全高效开采面临的地质环境损害预测预警难和减损开采保障压力大等难题,基于煤炭采动损害预测与减损地质保障多学科交叉特点,突出地质条件的控灾机理和采动致灾模式,强调煤炭开采模式与地质结构演化全时空多场响应,分析采动效应下地质条件物质场、能量场、信息场的耦合机制,厘清损害模式和预警信息关键参量的映射关系,形成安全高效开采的地质保障策略和减损控灾工程技术体系。【结果和结论】研究思路为“孕灾环境→损害机制→过程响应→损害预测→防灾减损”,技术路线是“煤矿开采地质环境条件→地质结构演化规律及损害模式→多物理场演化全时空信息响应→损害监测及预测预警→减损保障工程技术”。核心内容包括:(1)剖析地质条件与致灾地质体的空间关系和成因联系,构建高精度的三维地质力学模型,阐明地质结构、开采条件和损害模式的映射关系,建立主控要素的特征参量数据库;(2)构建采动效应下的工程地质力学模型,研究开采方式、空间布局、采动速率等影响下地质条件结构的时空演化特征和损害机制,提出考虑地质体关键结构破坏演化规律的损害模式判识方法;(3)获取采动过程地质结构演化背景下的全时空多源信息响应,提出主控参量作用下的损害模式识别标准,进而厘清裂隙场、应力场、渗流场与地球物理场信息参量的镜像关系,建立基于地质条件物质场、能量场和信息场耦合响应的全时空信息映射模型;(4)构建地面-钻孔-井下的全空间、多方位、主被动的一体化多源监测体系,提出煤矿采动损害预测模型和预报方法;(5)构建基于损害源、损害模式、损害动力、损害通道剖析以及减损技术与效果评价于一体的减损控灾体系,追求煤炭安全开采与地质环境保护协调发展,破解资源开发与地质环境制约之间的矛盾,为煤矿安全高效开采和防灾减损提供地质、力学、物理基础的科学依据。

滑坡-泥石流灾害链特征分析与风险评价——以川西石棉县马颈子沟为例

我国川西地区地形地貌和地质构造复杂,滑坡-泥石流灾害链易发育,灾害链中主要灾害向次要灾害逐级传递,促使灾害的风险性进一步增强。为研究滑坡-泥石流灾害链的特征与风险性,以四川马颈子沟滑坡-泥石流灾害链为典型研究对象,对该区域滑坡和泥石流分别进行风险度分析,并对区域滑坡-泥石流灾害链进行特征分析与风险评价。首先构建AHP模型对滑坡危险度进行分析,并应用易损性等级计算模型计算滑坡易损性;然后应用泥石流危险区指标来估算泥石流的危险度,并根据区域统计数据得出区域泥石流灾害损失脆弱性;最后根据多灾害风险理论公式,计算区域滑坡-泥石流灾害链风险度,并对灾害链进行风险评价。结果表明:马颈子沟滑坡-泥石流灾害链为中风险度地区,该研究结果与现场勘察结果一致,验证了基于多灾害风险理论的灾害链分析评价方法体系结合多种模型进行计算的可行性。通过分析马颈子沟灾害链风险度,为类似研究提供理论依据。

基于多源数据与机器学习的降雨型滑坡灾害危险性动态预警

降雨型滑坡灾害是三峡库区发育的主要地质灾害之一,区域滑坡灾害危险性预警对滑坡风险减缓有重要意义。针对区域滑坡灾害危险性预警的时效性等问题,本文以重庆市云阳县为研究区,提出了基于多源数据与机器学习的滑坡灾害危险性动态预警方法(图1)。1滑坡易发性空间预测滑坡易发性预测可准确地预测出潜在滑坡空间分布规律(黄发明等,2020;林高聪等,2023)。

自然灾害风险监测框架及产品体系设计

自然灾害风险监测是防灾减灾救灾领域的基础性和前沿性工作,为风险评估、灾害预警和综合减灾提供基础数据和研判信息。自然灾害风险监测当前处于理论研究和业务探索阶段,存在业务服务不精准、产品体系不健全、实践难题未凝练、关键技术未突破和监测站网不完备等难题,还不能满足灾害风险早发现、早识别和早防范的现实需求。结合多灾种灾害链基础理论和业务实践,阐明“为何监测、何时监测、怎么监测、监测哪里、监测什么”等业务系统化、体系化发展思路。为全面提升自然灾害风险监测产品决策支撑能力,设计了包括业务体系构建、关键技术突破、标准规范形成、数据库建设、应用平台开发和典型应用推广等在内的总体框架,构建应用场景、产品类型和产品要素一体化的科学精细的业务产品体系。

热带气旋多致灾因子综合风险评估研究进展

【目的】热带气旋伴随的大风、暴雨及引发的风暴潮、洪涝等致灾因子,为热带气旋风险评估带来了极大的挑战。针对当前热带气旋多致灾因子综合风险评估方法不一的问题,【方法】通过文献分析方法,阐述了热带气旋多致灾因子间的相互关系,并对热带气旋多致灾因子的危险性、承灾体脆弱性和风险评估研究现状进行了梳理。【结果】结果表明:综合考虑热带气旋多因子复合链生关系的定量化风险评估方法极为缺乏,如何降低气候变化背景下未来热带气旋风险评估结果的不确定性是当前的一个难点。【结论】未来可以在厘清热带气旋多致灾因子复杂关系的基础上,基于大数据和机器学习等计算机手段,从机理角度研发热带气旋多致灾因子复合链生关系的承灾体损失和风险量化模型,从而实现热带气旋多致灾因子损失和风险的精准和量化评估,为应急管理和风险防范提供重要支撑。

暴雨山洪与山坪塘漫坝多灾害耦合风险分析

为研究强降雨下山洪与漫坝耦合灾害对下游居民的影响程度和范围,应用HEC-RAS软件构建暴雨山洪与山坪塘漫坝多灾害耦合模型。首先,以重庆市巴南区某山坪塘为研究对象,根据巴南区暴雨强度公式设计不同重现期的降雨雨型;其次,借助数字高程数据对山坪塘进行原始构建,采用参数化模型对溃坝楔口参数进行预测,结合物理溃坝机制模拟漫坝过程;最后,选取洪水深度、洪水流速和洪水滞留时间等3个评价指标进行洪水风险分析。研究结果表明,突发强降雨下的山洪形成迅速且山坪塘发生漫坝的机率极大;山坪塘漫坝会加重附近区域的洪水风险等级。

水电工程施工安全隐患多标签文本智能分类方法

水利水电工程施工安全隐患体量大、形式多元、类型多样,同一隐患可能涉及多个类型,且隐患类型的界定存在模糊不清的现象。隐患分类多以人工经验为主导,极易导致隐患管理混淆,增加了隐患管理的难度。针对上述问题,本文提出了一种水电工程施工安全隐患文本多标签智能分类方法。首先,利用ALBERT模型对文本信息进行编码,实现非结构化安全隐患文本的高精度量化;然后,以文本量化结果为基础,考虑安全隐患中文本内容权重,利用Attention机制改进的双向门控循环单元(Bi-GRU),构建安全隐患文本多标签智能分类模型,提升施工安全隐患识别效率;最后,利用水电工程施工安全隐患文本,测试方法性能,得到方法的F1值达到了92.11%,证明了该方法的适用性,有望为水电工程施工的安全管理、事故隐患排查和分析提供信息支撑。