Predicting interfacial bonding performance of CRTS III slab ballastless track structure via interfacial defects using the PSO-BP algorithm

• Published in: Engineering structures Vol. 341; p. 120807
• Main Authors: Jiang, Wei, Lei, Bin, Wang, En, Ni, Yi-Qing, Chen, Zhengwei, Ding, Siqi, Peng, Yiming, Long, Guangcheng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.10.2025
• Subjects: CRTS III slab ballastless track structure; Interfacial bonding performance; Interfacial defect; PSO-BP algorithm; Interfacial defect; PSO-BP algorithm; Interfacial bonding performance; CRTS III slab ballastless track structure
• ISSN: 0141-0296
• DOI: 10.1016/j.engstruct.2025.120807

The judgment of interfacial bonding performance of the CRTS III slab ballastless track structure in the existing technical standard was mainly an empirical assessment through the area of interfacial defects. The evaluation method lacked the use of appropriate mechanical parameters for assessment. This study introduced a PSO-BP algorithm, which combined the Particle Swarm Optimization (PSO) algorithm and the Back Propagation (BP) algorithm to predict the splitting tensile strength, using the splitting tensile strength as a mechanical parameter to evaluate interfacial bonding performance. During this study, field-fabricated experimental plates were meticulously segmented into samples for testing splitting tensile strength. By analyzing defect area distribution and splitting tensile strength value, 22 critical feature labels were constructed for accurate prediction. Compared to the BP algorithm alone, the PSO-BP algorithm performed even better with a relative error of less than 10 %. In order to further improve the accuracy of the PSO-BP algorithm, various factors affecting its prediction performance were thoroughly investigated. An empirical formula was employed to minimize the mean square error of the training set, leading to the determination of an optimal solution for the number of hidden layer nodes. This study also found that setting the population size to 10 yielded optimal prediction results. Additionally, by fine-tuning the learning factors c1 and c2 to both be 2, the relative error was kept below 10 %. This study underscored the effectiveness of the PSO-BP algorithm in providing a straightforward and efficient solution for intelligent assessment of interfacial defects. •Used PSO-BP algorithm to closely link bubble defects with splitting tensile strength.•Produced experimental plates on-site and conducted field experiment.•Constructed 22 feature label values in the PSO-BP algorithm.•Analyzed various factors affecting the predictive performance of the PSO-BP algorithm.