Purpose To address the problems of extensive inspection blind zones, low inspection efficiency and high safety risks in traditional manual inspection of industrial steam pipelines, this study aims to develop an integrated automatic inspection system based on infrared thermography to improve inspection safety, efficiency and defect identification accuracy. Design/methodology/approach Taking steam pipelines in an underground utility tunnel of a power workshop as the application scenario, an automatic inspection system with a three-layer technical architecture of “mobile inspection–intelligent analysis–remote early warning” is designed and implemented. A tracked inspection vehicle equipped with a four-degree-of-freedom robotic arm is used to achieve multiangle coverage of pipeline surface regions. An integrated infrared thermal imaging sensor is used for temperature acquisition, while defect recognition and classification are realized through image preprocessing and machine learning-based analysis. In addition, a low-power wide-area communication link based on the Long Range Radio protocol is established to support real-time data transmission and remote monitoring. Findings Experimental results demonstrate that the proposed system can achieve stable temperature acquisition and reliable detection of surface thermal anomalies in steam pipelines. Under training with a simulation-based data set, the defect recognition model achieves an overall classification accuracy of 98.148%, and the defect type identification accuracy reaches 87.963%. Field validation further confirms the effectiveness and robustness of the system in complex underground environments. Originality/value The performance of the defect recognition model is influenced by the diversity and representativeness of the training data set, and further improvements can be achieved by incorporating more real-world operating conditions and defect samples. Nevertheless, the proposed system provides a practical technical reference for automated inspection and intelligent monitoring of industrial pipeline infrastructure. This study proposes an integrated automatic inspection framework that combines mobile robotic platforms, infrared thermography, intelligent defect recognition and low-power wireless communication for steam-pipeline inspection. The system effectively reduces inspection blind zones and safety risks while improving inspection efficiency and accuracy, offering a practical and replicable solution for intelligent inspection of industrial steam-pipeline networks.
Gao et al. (2026) studied this question.