Weld Defect Inspection of Battery Pack Based on Deep Learning of Linear Array Image

Tong Wu; Jincheng Yang; Ruiying Liao; Linghui Yang

doi:10.3788/LOP57.221502

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 22 >Page 221502 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 22, 221502 (2020)

Weld Defect Inspection of Battery Pack Based on Deep Learning of Linear Array Image

Tong Wu, Jincheng Yang, Ruiying Liao, and Linghui Yang^*

Author Affiliations

State Key Laboratory of Precision Measuring Technology and Instruments, Tianjin University, Tianjin 300072, China

show less

DOI: 10.3788/LOP57.221502 Cite this Article Set citation alerts

Tong Wu, Jincheng Yang, Ruiying Liao, Linghui Yang. Weld Defect Inspection of Battery Pack Based on Deep Learning of Linear Array Image[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221502 Copy Citation Text

show less

Fig. 1. Detection method of weld defect. (a) Measuring system of line scan camera; (b) flow chart of image acquisition; (c) flow chart of defect detection

Download full size

Fig. 2. Schematic diagram of multi-scale fusion

Download full size

Fig. 3. Data set.

Download full size

Fig. 4. Measurement system of line scan camera

Download full size

Fig. 5. Test results of our network. (a) Workpiece image; (b) weld inspection result

Download full size

Fig. 6. Weld defect and weld end inspection results. (a) Weld1; (b) weld2; (c) weld3

Download full size

$(w, x, y, z)$	P	R
(1,0,0,0)	90.7	88.1
(0,1,0,0)	91.3	93.5
(0,0,1,0)	92.7	95.2
(0,0,0,1)	90.6	88.6
(0.25,0.25,0.25,0.25)	95.2	95.1
(0.4,0.2,0.2,0.2)	95.4	94.9
(0.2,0.4,0.2,0.2)	98.9	91.7
(0.2,0.2,0.4,0.2)	97.6	94.9
(0.2,0.2,0.2,0.4)	98.1	92.3

Table 1. Weld defect identification results between different weighting combination unit: %

Inspection method		Weld seam	Pinhole	Wormhole	Dent
Number ofinspections	Test sample	20	65	12	31
	Improved YOLO	20	62	12	28
	Ref. [23]	20	63	12	28
	Ref. [24]	19	51	10	21
Recognition rate	Improved YOLO/%	100	95.4	100	90.3
	Ref. [23]/ %	100	96.9	100	90.3
	Ref. [24]/ %	95	78.5	83.3	67.7

Table 2. Test results of different methods

Inspection method	P₁/%	R₁/%	P₂/%	R₂/%	Inspection time/s
Improved YOLO	100	100	97.1	94.4	0.97
Ref. [23]	100	100	97.2	95.4	5.13
Ref. [24]	100	95.0	85.4	75.9	0.36

Table 3. Evaluation of detection effect

Tong Wu, Jincheng Yang, Ruiying Liao, Linghui Yang. Weld Defect Inspection of Battery Pack Based on Deep Learning of Linear Array Image[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221502

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information