Convolution neural network model with improved pooling strategy and feature selection for weld defect recognition
- PDF / 837,461 Bytes
- 14 Pages / 595.276 x 790.866 pts Page_size
- 58 Downloads / 241 Views
RESEARCH PAPER
Convolution neural network model with improved pooling strategy and feature selection for weld defect recognition Hongquan Jiang 1,2
&
Qihang Hu 1 & Zelin Zhi 1,3 & Jianmin Gao 1 & Zhiyong Gao 1 & Rongxi Wang 1 & Shuai He 1 & Hua Li 1
Received: 6 July 2019 / Accepted: 6 November 2020 # International Institute of Welding 2020
Abstract Weld defect recognition plays an important role in the manufacturing process of large-scale equipment. Traditional methods generally include several serial steps, such as image preprocessing, region segmentation, feature extraction, and type recognition. The results of each step have significant impact on the accuracy of the final defect identification. The convolutional neural network (CNN) has strong pattern recognition ability, which can overcome the above problem. However, there are two problems: one is that the pooling strategy has poor dynamic adaptability and the other is the insufficient feature selection ability. To overcome these problems, we propose a CNN-based weld defect recognition method, which includes an improved pooling strategy and an enhanced feature selection method. According to the characteristics of the weld defect image, an improved pooling strategy that considers the distribution of the pooling region and feature map is introduced. Additionally, in order to enhance the feature selection ability of the CNN, an enhanced feature selection method integrating the ReliefF algorithm with the CNN is proposed. A case study is presented for demonstrating the proposed techniques. The results show that the proposed method has higher accuracy than the traditional CNN method, and establish that the proposed CNN-based method is successfully applied for weld defect recognition. Keywords Weld defect recognition . Convolution neural network . Improved pooling strategy . Feature selection enhancement
Abbreviations CNN ReliefF SVM CGP
HD-CNN Convolutional neural networks Feature selection algorithm Support vector machine Cartesian genetic programming
P-CNN RNN DCNN
Hierarchical Deep Convolutional Neural Network Pose-based Convolutional Neural Network Recurrent Neural Network Deep Convolutional Neural Network
Recommended for publication by Commission XVIII - Quality Management in Welding and Allied Processes * Hongquan Jiang [email protected]
Shuai He [email protected]
Qihang Hu [email protected] Zelin Zhi [email protected] Jianmin Gao [email protected]
Hua Li [email protected] 1
State Key Laboratory for Manufacturing System Engineering, Xi’an Jiaotong University, Xi’an, China
2
Laboratory for Manufacturing and Productivity, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA
3
Shaanxi Special Equipment Inspection and Testing Institute, Xi’an, China
Zhiyong Gao [email protected] Rongxi Wang [email protected]
Weld World
ResNet RCNN ROI Fij n S σP σFM tmin tmax tave μ R W(A) m Mj(C) p(C) Class(Ri) diff(A, Ri, Rj) Cn Cm PO SL LF LP CR
Residual network Region Convolutional Neural Network Region o
Data Loading...
