基于神经网络技术的管道机电阻抗健康状况定量评估研究

Study of Electromechanical Impedance Quantitative Evaluation for Pipeline Structure Based on Neural Network Technique

将机电阻抗法用于管道法兰和主干结构健康监测,并利用BP神经网络对结构损伤进行定量评估。首先实验研究了管道法兰与主干结构健康状况对阻抗谱的影响,不同的结构损伤可通过阻抗分析仪测量的阻抗实部谱变化反映出来;然后利用BP神经网络技术对管道不同工况下得到的阻抗实部谱进行定量分析。采用阻抗值实部作为输入样本对神经网络进行训练,并使用受训的神经网络实现了对管道中不同结构损伤状况的定量评估。研究结果表明,将机电阻抗法与神经网络数据处理技术结合起来用于复杂管道的结构健康监测,不仅可实现对不同类型损伤的定量评估,同时还具有较高的稳定性。

Electromechanical impedance method was adopted to monitor the health condition of flange and main body of pipeline structure, furthermore, BP neural network was used to evaluate structure damage quantitatively. In experiment, the influence of health condition of flange and pipeline structure main body on the impedance spectra was studied firstly, then, the variation of real parts of impedance spectra obtained from impedance analyzer can characterize the different kinds of structure damage. BP neural network technique was used to quantitatively analyze the real parts of impedance spectrum under pipeline structure different working conditions. The real part of impedance spectra were selected as input samples to train developed neural network, finally, the trained neural network can achieve the quantitative evaluation of different kinds of damage in pipeline structure. Results show that the combination of electromechanical impedance method and neural network data processing technique for structural health monitoring of complex pipeline structure not only can effectively achieve quantitative evaluation of different kinds of defect but also have high stability.