Scholarlink

<p>The article presents research results on applying Bayesian Network Structure Learning (BNSL) methodology to assess the risk of additional construction works occurring in railway infrastructure projects. The justification for conducting research using Bayesian networks is the fact that its effectiveness has been confirmed in many issues in civil engineering and related disciplines. On this background, implementations of this methodology in the issue of structural reliability are particularly distinguished. Bayesian network inference models can also serve as effective decision support systems in the context of complex design processes and construction project implementation. The research was based on a database of information about the determinants of the risk process associated with the necessity of additional construction work. Using this collected information, the Bayesian network methodology was implemented, particularly the GOBNILP network structure extraction algorithm. Due to the properties of Bayesian networks, the created model can be applied both in forecasting the risk of additional works becoming necessary, as well as in diagnosing the causes of such necessity. An additional advantage of this model is its interpretability, resulting directly from the connections (edges) between individual nodes forming the DAG (Directed Acyclic Graph) structure. The validated Bayesian network, understood as a generative AI tool, can also be used for data augmentation. This, in turn, can lead to a more detailed description of the process and make it possible to analyse the sensitivity of the model with respect to its variables encoded in the form of nodes of the DAG structure. It should be emphasized that in previous attempts to use Bayesian networks in engineering applications, their structure was typically specified by the user (expert) and had an arbitrary character. Therefore, the research presented in this work, concerning the autonomous extraction of Bayesian network structure from data, represents a significant achievement in the field of civil engineering.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GOBNILP algorithm in identifying optimal risk model structure for additional work in railway project realizations

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年5月20日

查看全文
  • Janusz Rusek,
  • Agnieszka Leśniak,
  • Filip Janowiec

摘要

The article presents research results on applying Bayesian Network Structure Learning (BNSL) methodology to assess the risk of additional construction works occurring in railway infrastructure projects. The justification for conducting research using Bayesian networks is the fact that its effectiveness has been confirmed in many issues in civil engineering and related disciplines. On this background, implementations of this methodology in the issue of structural reliability are particularly distinguished. Bayesian network inference models can also serve as effective decision support systems in the context of complex design processes and construction project implementation. The research was based on a database of information about the determinants of the risk process associated with the necessity of additional construction work. Using this collected information, the Bayesian network methodology was implemented, particularly the GOBNILP network structure extraction algorithm. Due to the properties of Bayesian networks, the created model can be applied both in forecasting the risk of additional works becoming necessary, as well as in diagnosing the causes of such necessity. An additional advantage of this model is its interpretability, resulting directly from the connections (edges) between individual nodes forming the DAG (Directed Acyclic Graph) structure. The validated Bayesian network, understood as a generative AI tool, can also be used for data augmentation. This, in turn, can lead to a more detailed description of the process and make it possible to analyse the sensitivity of the model with respect to its variables encoded in the form of nodes of the DAG structure. It should be emphasized that in previous attempts to use Bayesian networks in engineering applications, their structure was typically specified by the user (expert) and had an arbitrary character. Therefore, the research presented in this work, concerning the autonomous extraction of Bayesian network structure from data, represents a significant achievement in the field of civil engineering.