Bridge facility health detection method based on DATA random subspace

• Combining singular value jumps with projective reconstruction for bridge detection. • Extending the parameter space of modal-based damage theory. • The method adapts to various structures. • The method is capable of detecting complex connections and localized damage. • Provides a theoretical basis and framework for bridge monitoring. Aiming at the shortcomings of traditional bridge health detection methods in terms of modal recognition accuracy and noise resistance, the study proposes an improved bridge health detection method. The method combines the singular value jump recognition and projection reconstruction techniques to determine the effective modal order of the structural vibration signals, and uses projection reconstruction techniques to improve the accuracy of the modal parameters. Compared with other algorithms, the research algorithm has higher bridge data detection accuracy, with an average data detection accuracy of 95.84 % and an accuracy growth rate of 26.9 %. In the damage mode with weakened mid-span stiffness, the positioning error of the improved method is controlled within 1.42 m, and the damage degree error is 6.48 %. In the case of bridge deck cracks, the positioning error is 1.11 m and the damage degree error is 4.83 %. Compared with the traditional method, the improved method improved 15.24 % and 12.13 % in terms of positioning accuracy and damage assessment accuracy, respectively. The results show that the health detection method proposed in the study can effectively identify the bridge damage location and accurately assess the damage degree. The method provides an efficient and reliable solution for bridge health monitoring and has a wide range of engineering applications.