工程研究——跨学科视野中的工程2025，Vol.17Issue(3)：364-377,14.DOI:10.3724/j.issn.1674-4969.20240016
基于改进OATA的物体打击事故致因-责任人关联分析
"Cause-Responsible Person"Correlation Analysis of Struck-by Accidents Based on Improved Occupational Accidents Tree Analysis
摘要
Abstract
The construction industry serves as a vital force driving China's economic development.However,due to the characteristics of construction projects such as changeable working environments,high personnel mobility,and frequent cross-operations,the accident rate remains generally high,posing a serious threat to the lives and property of personnel at construction sites and surrounding areas.Among them,Object strike injury is one of the common"five major injuries"in construction.The risks caused by struck-by accident are sudden and unpredictable.It will cause casualties once such accident occurs.Therefore,in order to improve the severe safety situation,this thesis analyzes the relationship between accident cause and responsible person,which is conducive to strengthening the understanding of the accident mechanism of all participants in the project.Consequently,it will enable them to adopt more targeted accident prevention measures. This study addresses a critical gap in construction safety management by developing a hybrid analytical framework to dissect the intricate relationships between struck-by accident causations and individual responsibilities.Regarding the research on safety accidents,Initially,the research merely focused on extracting accidents causes.Subsequently,it evolved to analyzing the interrelationships among these accident causes.Currently,the research has further advanced to exploring the connections between accident causes and various other factors.By assigning the implementation of these measures to specific individuals,the execution efficiency of such safety strategies can be significantly enhanced.However,existing studies have failed to clearly assign accidents causes to each specific responsible party.Moreover,most of the selected responsible parties are responsible units,and there is a relative lack of research on specific responsible individuals,such as project managers.This work pioneers a granular approach to link specific causes(e.g.,procedural violations,equipment malfunctions)to discrete roles(e.g.,project managers,safety officers)within the accident chain.This not only enables various responsible parties to clearly understand their duties but also helps decision-makers formulate more targeted preventive measures.Drawing on a dataset of 125 detailed accident reports from diverse construction projects across China(2018-2023)—spanning small-scale residential developments to large-scale infrastructure initiatives-the study integrates advanced computational and systems-analysis techniques to enhance precision in root-cause identification and responsibility attribution. The methodology combines text mining and social network analysis with an improved occupational accident tree analysis(OATA)to achieve three breakthroughs.First,extract the"Accident Causes"and"Liability Determination"sections from accident reports to construct a text corpus,using a hybrid algorithm fuses TF-IDF(quantifying term frequency-statistical significance)with TextRank(capturing semantic-contextual relevance)to extract 28 actionable causes from qualitative accident narratives.For example,"inadequate supervision"emerged as the highest-weighted cause(TF-IDF score:0.0748;TextRank score:0.0737),validated against industry standards.These causes were then linked to 15 predefined responsibility categories(e.g.,project managers,safety officers)through an adapted OATA model,revealing asymmetric causal-responsibility networks.For instance,project managers exhibited the highest aggregated liability(20.71％)due to systemic failures in oversight and corrective actions. Key findings highlighted management deficiencies as the dominant factor(62％of total liability),with"inadequate supervision"and"delayed hazard rectification"emerging as top contributors.In terms of environmental factors,there is only one category,namely adverse weather,with the impact of wind being the main influence,accounting for 2％of the total liability.The main responsible persons recorded in construction struck-by accident investigation report are:project managers,safety officers,mechanic and other 15 categories.Among them,the chief supervision engineer,safety officer,project leader,general technical workers,special operators and enterprise leaders are in the core position of responsible person network.By analyzing the correlations between accident causes and responsible parties,the specific causes borne by each type of responsible person and their corresponding responsibility rates were obtained.It was found that project managers,general technicians,and safety officers were identified as high-risk roles due to their central roles in decision-makingand compliance enforcement.Smallerprojects showed weaker safety protocols compared to larger ones,with informal subcontracting practices exacerbating risks. Practically,the study introduces a scalable framework to translate qualitative accident reports into actionable insights.Organizations can prioritize targeted interventions(e.g.,mandatory supervisor training),deploy IoT sensors for real-time hazard detection,and align regulatory frameworks with quantitative risk profiles.Methodologically,the research advances safety governance by validating OATA's applicability to construction contexts and demonstrating the synergistic value of blending computational linguistics with systems-based analysis.This work operationalizes the"Safety 3.0"paradigm,shifting from reactive incident response to proactive,responsibility-driven risk mitigation.关键词
物体打击事故/社会网络分析/职业事故树分析(OATA)/词频-逆文档频率/TextRank算法Key words
struck-by accident/social network analysis/occupational accidents tree analysis(OATA)/term frequency-inverse document frequency/TextRank algorithm分类
安全科学引用本文复制引用
李珏,武诗怡..基于改进OATA的物体打击事故致因-责任人关联分析[J].工程研究——跨学科视野中的工程,2025,17(3):364-377,14.基金项目
湖南省自然科学基金资助项目(2021JJ30744) （2021JJ30744）
湖南省教育厅科研项目(20K011) （20K011）
