首页|Uniaxial stress identification of steel components based on one dimensional-CNN and ultrasonic method
Uniaxial stress identification of steel components based on one dimensional-CNN and ultrasonic method
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Elsevier
The absolute stress of steel components is a key parameter in the construction and service of steel structures. Traditional stress testing methods have drawbacks of high cost and low accuracy. A new method based on deep learning and ultrasonic technique is proposed to obtain the absolute stress of steel components with different thicknesses. Firstly, ultrasonic signals of steel components under different stress levels were collected and used to build datasets. Secondly, the optimal architecture of one-dimensional convolutional neural networks (CNNs) for stress identification of steel components was determined. Finally, parameters of the network with the optimal architecture were optimized and used to identify the absolute stress of the unknown test dataset. The results show that the average stress identification error for the unknown test dataset is 3.83%. The proposed method can overcome the drawbacks of conventional techniques and provide good references for stress identification of steel components in practical engineering.
NSTL
Elsevier
