Data-driven network analysis of construction hazard characteristics and warning strategy
LIU Mei1, XU Linyu2, LIAO Pin-Chao2, HUANG Yuecheng2, SUN Chengshuang1
1. School of Urban Economics and Management, Beijing University of Civil Engineering and Architecture, Beijing 100044, China; 2. Department of Construction Management, Tsinghua University, Beijing 100084, China
Abstract:[Objective] The investigation and management of construction hazards have transformed with the digitization and intellectualization of the construction industry. However, the efficient analysis and application of large-scale and high-dimensional hazard data remain farfetched. In this study, a hazard characteristic network is constructed by extracting and reducing the dimensions of the hazard record data on the construction site and data-driven hazard warning strategies are preseneted. [Methods] First, the hazard records are standardized using the text mining method and 111 safety hazard characteristics are extracted. Second, the correlations of these characteristics are determined using the correlation test, and the characteristic network is established based on the correlation strength calculation. Third, critical hazard characteristics are identified based on the analysis of the structural and individual indices in the network. In addition, 11 groups are obtained by the hierarchical clustering of all hazard characteristics. Finally, based on the above methods and analyses, data-driven warning strategies are preseneted in view of hazard characteristics. [Results] Through charateristic extraction and hierarchical clustering of potential safety hazards, 111 potential safety hazard characteristics and 11 potential safety hazard characteristic groups are determined to reduce the dimensions of potential safety hazard data. Based on the correlation test and correlation strength calculation, a safety hazard characteristic network is constructed, and the network and individual indices of safety hazard characteristics (out-degree, in-degree, intermediate, and eigenvector centralities) are analyzed to identify the key hazard characteristics (“facilities/equipment/apparatus/devices”, “settings”, “scaffolding”, and “railings”) and their associated paths with different roles. The early-warning strategies based on the data-driven network analysis are presented to improve timeliness from two aspects. In the early-warning process, relevant potential hazard information and importance ranking are obtained through the correlation analysis of the safety hazards via onsite inspection. Under limited manual labor, the safety management personnel are equipped with troubleshooting ideas and clues to mitigate the limitations of the original unplanned hazard inspection method. Meanwhile, a programmed information system can be further developed to provide early-warning tools. Safety managers can input the hazard information, and the system will conduct rapid standardization in the background to deal with the information and provide the associated hazards and sequencing to ensure timely feedback of hazard inspection clues. [Conclusions] This study establishes the standardization and characteristic extraction method of hazard record data which reduces data dimensionality from 3 [KG-*7]267 nonstandard hazard records to 111 hazard characteristics and 11 hazard characteristic groups to clarify the key information regarding inspection and governance, such as hazard types and scenarios. Based on the analysis of individual and structural indices of the hazard characteristic network, the deduction and importance ranking of potentially associated hazard characteristics are realized, thus providing early-warning clues for effective safety inspection and governance. The early-warning strategies based on data-driven hazard characteristics can not only address the inefficient original unscheduled search method in the early-warning process but also improve safety management efficiency through timely feedback of inspection clues. This study introduces a foundational method for mining regular strategy information with early-warning clues of engineering construction safety hazard data, as well as for effective inspection and governance of construction safety hazards.
[1] 方东平, 黄新宇, HINZE J. 工程建设安全管理[M]. 2版. 北京: 中国水利水电出版社, 2005. FANG D P, HUANG X Y, HINZE J. Safety management of engineering construction[M]. 2nd ed. Beijing: China Water & Power Press, 2005. (in Chinese) [2] MIHIĆ M. Classification of construction hazards for a universal hazard identification methodology[J]. Journal of Civil Engineering and Management, 2020, 26(2): 147-159. [3] CHEN F Y, WANG H W, XU G Y, et al. Data-driven safety enhancing strategies for risk networks in construction engineering[J]. Reliability Engineering and System Safety, 2020, 197: 106806. [4] GANBAT T, CHONG H Y, LIAO P C, et al. Identification of critical risks in international engineering procurement construction projects of Chinese contractors from the network perspective[J]. Canadian Journal of Civil Engineering, 2020, 47(12): 1359-1371. [5] LIU J Y, ZHAO X B, YAN P. Risk paths in international construction projects: Case study from Chinese contractors[J]. Journal of Construction Engineering and Management, 2016, 142(6): 05016002. [6] LIAO P C, GUO Z H, TSAI C H, et al. Spatial-temporal interrelationships of safety risks with dynamic partition analysis: A mechanical installation case[J]. KSCE Journal of Civil Engineering, 2018, 22(5): 1572-1583. [7] WAMBEKE B W, LIU M, HSIANG S M. Task variation and the social network of construction trades[J]. Journal of Management in Engineering, 2014, 30(4): 05014008. [8] OKUDAN O, BUDAYAN C, DIKMEN I. A knowledge-based risk management tool for construction projects using case-based reasoning[J]. Expert Systems with Applications, 2021, 173: 114776. [9] ZHANG M Y, ZHU M, ZHAO X F. Recognition of high-risk scenarios in building construction based on image semantics[J]. Journal of Computing in Civil Engineering, 2020, 34(4): 04020019. [10] XU N, MA L, LIU Q, et al. An improved text mining approach to extract safety risk factors from construction accident reports[J]. Safety Science, 2021, 138: 105216. [11] THOMPSON P, YATES T, INAN E, et al. Semantic annotation for improved safety in construction work[C]// Proceedings of the 12th Language Resources and Evaluation Conference. Marseille, France: European Language Resources Association, 2020: 1990-1999. [12] SHRESTHA S, MORSHED S A, PRADHANANGA N, et al. Leveraging accident investigation reports as leading indicators of construction safety using text classification[C]// Construction Research Congress 2020: Safety, Workforce, and Education. Tempe, USA, 2020: 490-498. [13] ZHONG B T, PAN X, LOVE P E D, et al. Hazard analysis: A deep learning and text mining framework for accident prevention[J]. Advanced Engineering Informatics, 2020, 46: 101152. [14] 黑永健. 基于文本挖掘的地铁施工隐患分析及可视化研究[D]. 武汉: 华中科技大学, 2019. HEI Y J. Research on analysis and visualization of subway construction safety hazards based on text mining[D]. Wuhan: Huazhong University of Science & Technology, 2019. (in Chinese) [15] NGUYEN L D, TRAN D Q, CHANDRAWINATA M P. Predicting safety risk of working at heights using Bayesian networks[J]. Journal of Construction Engineering and Management, 2016, 142(9): 04016041. [16] HALLOWELL M R, GAMBATESE J A. Construction safety risk mitigation[J]. Journal of Construction Engineering and Management, 2009, 135(12): 1316-1323. [17] ZHAO T S, LIU W, ZHANG L M, et al. Retracted: Cluster analysis of risk factors from near-miss and accident reports in tunneling excavation[J]. Journal of Construction Engineering and Management, 2018, 144(6): 04018040. [18] PETERSEN D. Human error reduction and safety management[M]. New York, USA: Van Nostrand Reinhold, 1996. [19] REASON J. Human error: Models and management[J]. BMJ, 2000, 320(7237): 768-770. [20] LIU M, XU L Y, LIAO P C. Character-based hazard warning mechanics: A network of networks approach[J]. Advanced Engineering Informatics, 2021, 47: 101240. [21] 崔阳, 陈勇强, 徐冰冰. 工程项目风险管理研究现状与前景展望[J]. 工程管理学报, 2015(2): 76-80. CUI Y, CHEN Y Q, XU B B. Research of risk management in construction project: Current situation and future directions[J]. Journal of Engineering Management, 2015(2): 76-80. (in Chinese) [22] KIM B C. Integrating risk assessment and actual performance for probabilistic project cost forecasting: A second moment bayesian model[J]. IEEE Transactions on Engineering Management, 2015, 62(2): 158-170. [23] 秦旋, 李怀全, 莫懿懿. 基于SNA视角的绿色建筑项目风险网络构建与评价研究[J]. 土木工程学报, 2017, 50(2): 119-131. QIN X, LI H Q, MO Y Y. Study on establishment and evaluation of risk network in green building projects based on SNA[J]. China Civil Engineering Journal, 2017, 50(2): 119-131. (in Chinese) [24] LEE C Y, CHONG H Y, LIAO P C, et al. Critical review of social network analysis applications in complex project management[J]. Journal of Management in Engineering, 2018, 34(2): 04017061. [25] YANG R J, ZOU P X W. Stakeholder-associated risks and their interactions in complex green building projects: A social network model[J]. Building and Environment, 2014, 73: 208-222. [26] LI C Z, HONG J K, XUE F, et al. Schedule risks in prefabrication housing production in Hong Kong: A social network analysis[J]. Journal of Cleaner Production, 2016, 134: 482-494. [27] CHU C Y, PARK K, KREMER G E. A global supply chain risk management framework: An application of text-mining to identify region-specific supply chain risks[J]. Advanced Engineering Informatics, 2020, 45: 101053. [28] ZAMORA J, MENDOZA M, ALLENDE H. Hashing-based clustering in high dimensional data[J]. Expert Systems with Applications, 2016, 62: 202-211. [29] SONG B, YAN W, ZHANG T J. Cross-border e-commerce commodity risk assessment using text mining and fuzzy rule-based reasoning[J]. Advanced Engineering Informatics, 2019, 40: 69-80. [30] JAZIZADEH F, BECERIK-GERBER B, BERGES M, et al. An unsupervised hierarchical clustering based heuristic algorithm for facilitated training of electricity consumption disaggregation systems[J]. Advanced Engineering Informatics, 2014, 28(4): 311-326. [31] DE OLIVEIRA D P, GARRETT J H JR, SOIBELMAN L. A density-based spatial clustering approach for defining local indicators of drinking water distribution pipe breakage[J]. Advanced Engineering Informatics, 2011, 25(2): 380-389. [32] 许林宇. 基于多层网络的建设工程安全隐患特征预警研究[D]. 北京: 清华大学, 2021. XU L Y. A character-based hazard warning research in construction projects: Multi-layer network approach[D]. Beijing: Tsinghua University, 2021. (in Chinese) [33] WARD J H J. Hierarchical grouping to optimize an objective function[J]. Journal of the American Statistical Association, 1963, 58(301): 236-244. [34] LOUREIRO A, TORGO L, SOARES C. Outlier detection using clustering methods: A data cleaning application[C]// Proceedings of KDNet Symposium on Knowledge-Based Systems for the Public Sector. Bonn, Germany, 2004. [35] YATES F. Contingency tables involving small numbers and the χ2 test[J]. Supplement to the Journal of the Royal Statistical Society, 1934, 1(2): 217-235. [36] 丁嘉威. 网络视角下的安全风险关联机理: 以电梯安装工程为例[D]. 北京: 清华大学, 2016. DING J W. Research on the mechamism of risk interdependence from the network perspective: Taking the elevator installation project as an example[D]. Beijing: Tsinghua University, 2016. (in Chinese)