Measurement of transverse phase space based on machine learning

Yutao Han; Renkai Li; Weishi Wan

doi:10.11884/HPLPB202335.230074

Journals >High Power Laser and Particle Beams >Volume 35 >Issue 11 >Page 114005 > Article

High Power Laser and Particle Beams
Vol. 35, Issue 11, 114005 (2023)

Measurement of transverse phase space based on machine learning

Yutao Han¹, Renkai Li², and Weishi Wan¹

Author Affiliations

¹School of Physical Science and Technology, ShanghaiTech University, Shanghai 201210, China

²Department of Engineering Physics, Tsinghua University, Beijing 100084, China

show less

DOI: 10.11884/HPLPB202335.230074 Cite this Article

Yutao Han, Renkai Li, Weishi Wan. Measurement of transverse phase space based on machine learning[J]. High Power Laser and Particle Beams, 2023, 35(11): 114005 Copy Citation Text

show less

Fig. 1. Flow chart of hybrid domain processing

Download full size | View in the Article

Fig. 2. Diagram of Residual U-Net architecture

Download full size | View in the Article

Fig. 3. Layout of tomography section

Download full size | View in the Article

Fig. 4. Function diagram of focusing parameterk and rotation angle

Download full size | View in the Article

Fig. 5. A TPS distribution and its normalized TPS distribution

Download full size | View in the Article

Fig. 6. Rotation angles corresponding to different sampling methods

Download full size | View in the Article

Fig. 7. Sinograms using different sampling methods

Download full size | View in the Article

Fig. 8. Two laser spot with details

Download full size | View in the Article

Fig. 9. An example of the interpolation network results in the form of sinogram

Download full size | View in the Article

Fig. 10. An example of the interpolation network results in the form of tomography

Download full size | View in the Article

Fig. 11. Uniform K value sampling and its interpolation

Download full size | View in the Article

Fig. 12. Examples of the results from removing artifacts network

Download full size | View in the Article

name	parameters	output
Conv_block_1	1 $ \times $1 conv, 64	200 $ \times $57, 64
Conv_block_1	3 $ \times $3 conv, 64	200 $ \times $57, 64
Conv_block_2	2 $ \times $3 conv, s=2, p=0, 64	100 $ \times $28, 64
Conv_block_2	[3 $ \times $3 conv, 64] $ \times $2	100 $ \times $28, 64
Conv_block_3	2 $ \times $2 conv, s=2, p=0, 64	50 $ \times $14, 64
Conv_block_3	[3 $ \times $3 conv, 64] $ \times $2	50 $ \times $14, 64
Conv_block_4	2 $ \times $2 conv, s=2, p=0, 64	25 $ \times $7, 64
Conv_block_4	[3 $ \times $3 conv, 64] $ \times $2	25 $ \times $7, 64
ConvT_block_1	$2 \times 2$ convT, s=2, p=0, 64	50 $ \times $14, 64
ConvT_block_2	Conv_block_3, concatenation	100 $ \times $28, 64
	[ $3 \times 3$ conv,64] $ \times $2
	$2 \times 2$ convT, s=2, p=0, 64
ConvT_block_3	Conv_block_2, concatenation	200 $ \times $57, 64
	[ $3 \times 3$ conv, 64] $ \times $2
	$2 \times 3$ convT, s=2, p=0, 64
Conv_block_5	Conv_block_1, concatenation	200 $ \times $57, 1
	$3 \times 3$ conv,16
	$3 \times 3$ conv, 1
	shortcut connection

Table 1. Residual U-Net network parameter settings

name	parameters	output
Conv_1	[5, Conv, 16] $ \times $2	192 $ \times $192, 16
Conv_2	[5, Conv, 16] $ \times $2	184 $ \times $184, 16
Conv_3	[5, Conv, 16] $ \times $2	176 $ \times $176, 16
ConvT_1	[5, ConvT, 16] $ \times $2	184 $ \times $184, 16
ConvT_2	Conv_2, addition	192 $ \times $192, 16
ConvT_2	[5, convT, 16] $ \times $2	192 $ \times $192, 16
ConvT_3	Conv_1, addition	200 $ \times $200, 1
	[5, convT1, 1] $ \times $2
	shortcut connection

Table 2. RED-CNN network parameter settings

Yutao Han, Renkai Li, Weishi Wan. Measurement of transverse phase space based on machine learning[J]. High Power Laser and Particle Beams, 2023, 35(11): 114005

Download Citation

Tools

Save the article for my favorites

Paper Information