基于光纤传感器的桥梁预应力施加质量评估

Quality Assessment of Prestress Application in Bridges Based on Fiber Optic Sensors

摘要:

以海南省椰林桥20 m预应力混凝土空心板为研究对象，提出了一种基于光纤传感器与BP神经网络的桥梁预应力施加质量评估方法。首先，采用光纤布拉格光栅（FBG）传感器实时采集空心板关键截面应变数据，并与有限元模型理论计算值进行对比分析，量化预应力损失率和极差；然后，构建基于TensorFlow框架的BP神经网络模型，并将张拉力和伸长量作为输入特征；最后，输出预应力施加质量的质量等级分类，实现预应力质量的动态评估。结果表明，光纤传感器实测数据与有限元理论计算值之间的最大偏差为5.9%，证明数据采集可靠，神经网络模型经优化后的验证集准确率提升至83.3%。该评估方法通过数据驱动的多参数映射，实现了预应力施加质量的自动化评估，为桥梁工程的质量控制提供了科学依据。

Abstract:

Taking the 20-m prestressed concrete hollow slab of the Yelin Bridge in Hainan Province as the research object, a bridge prestress application quality assessment method based on fiber optic sensors and BP neural networks is proposed. Fiber Bragg Grating (FBG) sensors are used to collect the real-time strain data at key sections of the hollow slab, and the comparative analysis is conducted with finite element model theoretical values to quantify the prestress loss rates and range differences. A BP neural network model based on the TensorFlow framework is built, and the tension force and elongation are used as input features. Finally, the quality grade classification of the prestressed application quality is outputted to achieve the dynamic assessment of the prestressed quality. The results show that the maximum deviation between the measured data of the fiber optic sensor and the theoretical calculation value of the finite element is 5.9%, which confirms the data reliability. After optimization, the validation accuracy of the neural network model reaches 83.3%. This assessment method realizes the automated evaluation of prestress application quality through data-driven multi-parameter mapping, which provides a scientific basis for quality control of bridge engineering.