Truncated pulse height estimator based on an improved UNet model

Lin TANG; Shuang ZHOU; Xianli LIAO; Ze LIU; Bo LI

doi:10.11889/j.0253-3219.2023.hjs.46.110505

Journals >NUCLEAR TECHNIQUES >Volume 46 >Issue 11 >Page 110505 > Article

NUCLEAR TECHNIQUES
Vol. 46, Issue 11, 110505 (2023)

Truncated pulse height estimator based on an improved UNet model

Lin TANG^1、3, Shuang ZHOU¹, Xianli LIAO^1、2, Ze LIU¹, and Bo LI^1、*

Author Affiliations

¹College of Electronic Information and Electrical Engineering, Chengdu University, Chengdu 610106, China

²School of Nuclear Technology and Automation Engineering, Chengdu University of Technology, Chengdu 610059, China

³School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, Singapore

show less

DOI: 10.11889/j.0253-3219.2023.hjs.46.110505 Cite this Article

Lin TANG, Shuang ZHOU, Xianli LIAO, Ze LIU, Bo LI. Truncated pulse height estimator based on an improved UNet model[J]. NUCLEAR TECHNIQUES, 2023, 46(11): 110505 Copy Citation Text

show less

Fig. 1. Output signal of a preamplifier

Download full size | View in the Article

Fig. 2. Generation process of datasets

Download full size | View in the Article

Fig. 3. Structural diagram of UNet-LSTM model

Download full size | View in the Article

Fig. 4. Changes in loss values during the training of the UNet-LSTM model (color online)

Download full size | View in the Article

Fig. 5. Simulated pulse sequence diagram

Download full size | View in the Article

Fig. 6. Trapezoidal shaping result (color online)

Download full size | View in the Article

Fig. 7. Photographs of experimental setup (a) Detector, (b) X-ray tube

Download full size | View in the Article

序列 Sequence	脉冲 Pulse	是否截断 Truncated or not	脉冲高度Pulse amplitude $A_{r e a l}$	梯形成形值Trapezoidal shaping value $A_{T r a p e}$	相对误差Relative error $δ_{T r a p e}$ / %	模型估计值Model estimate $A_{U N e t}$	相对误差Relative error $δ_{U N e t}$ / %
S1 （ $R_{t}$ =25%）	P1	No	0.96	1.01	4.98	0.93	3.01
	P2	No	0.32	0.30	6.10	0.31	1.08
	P3	Yes	0.59	0.46	21.33	0.57	3.73
	P4	No	0.43	0.45	3.72	0.40	5.88
S2 （ $R_{t}$ =50%）	P5	No	0.71	0.73	2.38	0.71	0.38
	P6	Yes	0.50	0.49	1.82	0.52	3.67
	P7	Yes	0.46	0.47	1.88	0.44	3.27
	P8	No	0.73	0.76	4.21	0.71	3.15
S3 （ $R_{t}$ =50%）	P9	Yes	0.66	0.67	1.73	0.66	0.76
	P10	No	0.40	0.41	4.45	0.40	1.01
	P11	Yes	0.71	0.75	5.30	0.69	2.34
	P12	No	0.52	0.54	4.36	0.51	1.16
S4 （ $R_{t}$ =25%）	P13	No	0.71	0.72	1.95	0.72	1.27
	P14	Yes	0.51	0.52	1.54	0.52	1.97
	P15	No	0.60	0.61	2.13	0.62	3.35
	P16	No	0.79	0.82	4.25	0.80	1.39
S5 （ $R_{t}$ =50%）	P17	No	0.46	0.45	2.63	0.45	2.18
	P18	Yes	0.72	0.75	4.42	0.70	2.09
	P19	Yes	0.94	0.95	0.31	0.93	1.38
	P20	No	0.46	0.44	4.94	0.48	3.12

Table 1. Comparison between the estimated and true values of the pulse height using neural network models

序号Number	样品类型Sample type	截断率Truncation rate / %	平均相对误差Average relative error / %
1	粉末铁矿样品Powder iron ore sample	25	2.21
2		50	2.46
3		75	2.42
4		100	2.75
5	粉末岩石样品Powder rock sample	25	1.19
6		50	2.41
7		75	2.15
8		100	3.28

Table 2. Comparison of the pulse height estimation effects of neural network models on different samples

Lin TANG, Shuang ZHOU, Xianli LIAO, Ze LIU, Bo LI. Truncated pulse height estimator based on an improved UNet model[J]. NUCLEAR TECHNIQUES, 2023, 46(11): 110505

Download Citation

Tools

Save the article for my favorites

Paper Information