Quantifying the reliability of defects located by bridge inspectors through human observation behavioral analysis

被引：5

作者：

Liu, Pengkun ^{[1
]}

Shi, Ying ^{[1
]}

Xiong, Ruoxin ^{[1
]}

Tang, Pinbo ^{[1
]}

机构：

[1] Carnegie Mellon Univ, Dept Civil & Environm Engn, 5000 Forbes Ave, Pittsburgh, PA 15213 USA

来源：

DEVELOPMENTS IN THE BUILT ENVIRONMENT | 2023年 / 14卷

基金：

美国国家科学基金会;

关键词：

Bridge inspection; Visual inspection; Defects localization; Process mining; Crowdsourcing; Ensemble learning; PROCESS MODELS; AGREEMENT; COLLAPSE; QUALITY;

D O I：

10.1016/j.dibe.2023.100167

中图分类号：

TU [建筑科学];

学科分类号：

0813 ;

摘要：

Distinct site conditions and individual expertise contribute to the subjective nature of bridge inspection processes, which involve uncertain human factors. Assessing inspection reliability can be achieved by examining inspectors' behaviors that lead to inaccurately identified or overlooked defects. However, the scarcity of comprehensive behavioral data regarding defect observation poses a challenge in evaluating inspection consistency. This paper investigates observation behaviors in bridge inspection to quantify the reliability of structural defect localizations. We employ defect inspection strategies correlated with more dependable defect localization records to construct a behavioral process graph that quantifies inspectors' performance and predicts their "inspection reliability index." The generated reliability index for inspectors serves as a weighting factor to emphasize the opinions of more reliable inspectors when consolidating inspection records. The findings reveal that aggregating the inspection records of 96 human subjects based on their reliability indices effectively filters out false alarms while retaining reliable defect records.

引用

页数：18