神经网络在桥梁结构等级状态评估中的应用

探讨了模糊数学中的隶属函数在桥梁技术等级状态评估中的应用。在研究现有桥梁状态评估方法的基础上,把人工神经网络和模糊数学理论结合起来应用于大跨度预应力斜拉桥的结构等级状态评估,建立了基于3层神经元的模糊神经网络模型,并建立结构损伤度函数及结构等级隶属度模型,通过样本学习训练,获取评估专家的知识及直觉思维,最终确定桥梁所对应的技术状态等级。以检测的150组线形检测数据作为学习样本,另外8组作为验证样本进行了线形状态评估预测。计算结果表明,网络预测值与期望值吻合良好。

Application of Neural Network on Grade Condition Assessment of Bridge Structure

The applications of fuzzy mathematics in grade condition assessment of bridge structure were discussed. Based on the research of present methods, artificial neural network and fuzzy mathematics were integrated and applied in grade condition assessment of long span PC(prestressed concrete) cable bridge structure. 3-layer fuzzy BPNN (back-propagation neural network) was proposed, DV(damage value) function and degree of grade membership model were applied then. Expert knowledge was obtained, due to sample learning. And technical state grade of bridge was eventually acquired. In the end, 150 sets of inspected data of cable tensions were used for training samples, while the other 8 for testing to predict the condition of line shape. Results show that it agrees quite well with the expectation.