Segmentation method of U-net sheet metal engineering drawing based on CBAM attention mechanism

Oct 24, 2026

Heavy industrial equipment manufacturing often requires manual redrawing of specific units in welding engineering drawings before cutting sheet metal parts, which reduces efficiency. This paper presents a U-net based method for segmenting and extracting specific units in welding drawings, enabling automatic segmentation from visual information and shape-matched cutting. Performance is improved by replacing weak U-net skip connections with a CBAM attention U-net jump structure and dual-pooling convolutional fusion to reduce feature-dimension discrepancies. Experiments on a welding drawing dataset validate higher IoU, mAP, and accuracy versus the traditional U-net baseline.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Related Work

Segmentation method of U-net sheet metal engineering drawing based on CBAM attention mechanism Songzhiwei 1，Yao Hui2 [1](1 songzhiwei@st.xatu.edu.cn)[ songzhiwei@st.xatu.edu.cn](1 songzhiwei@st.xatu.edu.cn)；[2](2 yaohui@xatu.edu.cn)[ yaohui@xatu.edu.cn](2 yaohui@xatu.edu.cn) Abstract In the manufacturing process of heavy industrial equipment, the specific unit in the welding diagram is first manually redrawn and then the corresponding sheet metal parts are cut, which is inefficient. To this end, this paper proposes a U-net-based method for the segmentation and extraction of specific units in welding engineering drawings. This method enables the cutting device to automatically segment specific graphic units according to visual information and automatically cut out sheet metal parts of corresponding shapes according to the segmentation results. This process is more efficient than traditional human-assisted cutting. Two weaknesses in the U-net network will lead to a decrease in segmentation performance: first, the focus on global semantic feature information is weak, and second, there is a large dimensional difference between shallow encoder features and deep decoder features. Based on the CBAM (Convolutional Block Attention Module) attention mechanism, this paper proposes a U-net jump structure model with an attention mechanism to improve the network's global semantic feature extraction ability. In addition, a U-net attention mechanism model with dual pooling convolution fusion is designed, the deep encoder's maximum pooling + convolution features and the shallow encoder's average pooling + convolution features arefused vertically to reduce the dimension difference between the shallow encoder and deep decoder. The dual-pool convolutional attention jump structure replaces the traditional U-net jump structure, which can effectively improve the specific unit segmentation performance of the welding engineering drawing. Using vgg16 as the backbone network, experiments have verified that the IoU, mAP, and Accu of our model in the welding engineering drawing dataset segmentation task are 84.72%, 86.84%, and 99.42%, respectively. The method in this paper is 22.10, 19.09, and 0.05 percentage points higher than the traditional U-net, and has better segmentation performance for welding engineering drawings. Keywords: Intelligent Manufacturing; Welding Engineering; CBAM (Convolution Block Attention Module); U-net; Double-pool Convolution 1. Introduction Heavy industry equipment generally adopts customized manufacturing. In such projects, Fig 1. Comparison of our method (d) with skip structure schemes of other models. Dashed lines denote skip connections. large sheet metal parts need to be cut out according to the content of customized engineering drawings and finally manufactured by welding and stamping operations. Efficiency and precision are critical to the manufacture of customized heavy industrial equipment. The specific process of the traditional way to obtain sheet metal parts in the manufacturing process of heavy industry equipment is manual recognition of engineering drawings[1,2]-manual redrawing of specific graphics[3,4]-sheet metal cutting and stamping based on CAD/CAM integrated system[5,6,7] . The process of obtaining specific sheet metal parts in traditional ways is cumbersome and inefficient. With the concept of intelligent manufacturing being proposed, the efficient automatic cutting of customized sheet metal parts is an urgent problem to be solved in the current heavy industry equipment manufacturing process. In recent years, with the development of artificial intelligence, the industrial field is also more inclined to use deep learning methods to solve engineering problems. Some studies have shown that this method has an excellent performance in solving engineering problems. For example, Lau et al[8] . used deep learning image segmentation technology to realize the automatic detection of road cracks, an

FAQs

What problem does the proposed U-net method address in welding engineering drawings?

It addresses the low efficiency of manually recognizing and redrawing specific units in welding diagrams before cutting sheet metal parts.

How does CBAM improve the U-net segmentation performance?

CBAM enhances the network’s global semantic feature extraction ability and strengthens attention through an improved jump structure.

What metrics were used to evaluate the method, and how did it perform?

The study uses IoU, mAP, and Accu; the model achieves 84.72% IoU, 86.84% mAP, and 99.42% Accu, outperforming the traditional U-net.

Oct 24, 2026

Upload by Maya Linwood