Terahertz Nondestructive Testing Signal Recognition Based on PSO-BP Neural Network

Meihui JIA; Lijuan LI; Jiaojiao REN; Jian GU; Dandan ZHANG; Jiyang ZHANG; Weihua XIONG

doi:10.3788/gzxb20215009.0930004

Acta Photonica Sinica, Volume. 50, Issue 9, 0930004(2021)

Terahertz Nondestructive Testing Signal Recognition Based on PSO-BP Neural Network

Meihui JIA, Lijuan LI, Jiaojiao REN, Jian GU, Dandan ZHANG, Jiyang ZHANG, and Weihua XIONG

Key Laboratory of Optoelectronic Measurement and Control and Optical Information Transmission Technology of Ministry of Education， College of Optoelectronic Engineering， Changchun University of Science and Technology， Changchun130022， China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. Defect sample drawing of high temperature resistant composite material with multi-adhesive structure

Download full size

View in Article

Fig. 2. Terahertz time domain waveforms of different defect areas

Download full size

View in Article

Fig. 3. Operating principle diagram of BP neural network

Download full size

View in Article

Fig. 4. Flow chart of optimization process

Download full size

View in Article

Fig. 5. Schematic diagram of high temperature resistant composite bonding sample

Download full size

View in Article

Fig. 6. Schematic diagram of reflective THZ-TDS

Download full size

View in Article

Fig. 7. Comparison of mean square errors of training results between BP neural network and PSO-BP neural network.

Download full size

View in Article

Fig. 8. Defect recognition results of BP neural network.

Download full size

View in Article

Fig. 9. Defect recognition results of PSO-BP neural network

Download full size

View in Article

Fig. 10. Recognition results of different degree of debonding defects by PSO-BP neural network

Download full size

View in Article

Fig. 11. Error comparison of PSO-BP neural network in recognizing defects of different degrees

Download full size

View in Article

Table 1. Time domain characteristics and their expressions

View table

View in Article

Table 1. Time domain characteristics and their expressions

Serial number	Characteristics of the name	Characteristic expression
1	Kurtosis	$K_{m i n} = \frac{E [(X_{i} {- \sum_{i = 1}^{n} X_{i} / n)}^{4}]}{{E [(X_{i} {- \sum_{i = 1}^{n} X_{i} / n)}^{2} {]}}^{2}}$
2	Skewness	$S_{m i n} = \frac{E [(X_{i} {- \sum_{i = 1}^{n} X_{i} / n)}^{3}]}{{E [(X_{i} {- \sum_{i = 1}^{n} X_{i} / n)}^{2} {]}}^{3 / 2}}$
3	Waveform factor	$X_{p p - u p} = \frac{\sqrt[]{\sum_{i = 1}^{n} {X_{i}}^{2} / n}}{\sum_{i = 1}^{n} \|X_{i}\| / n}$
4	Absolute mean of amplitude	${\|X\|}_{m e a n} = \sum_{i = 1}^{n} \|X_{i}\| / n$
5	In peak value	$X_{p v} = X_{m a} - X_{m i}$
6	Minimum amplitude value	$X_{i m i n}$

Table 2. Defects of the sample
View table
View in Article
Table 2. Defects of the sample
The serial number Upper defect level Lower defect level
1 150 μm 350 μm
2 100 μm No defect
3 No defect 250 μm
4 No defect 100 μm

Tools

Get Citation

Copy Citation Text

Meihui JIA, Lijuan LI, Jiaojiao REN, Jian GU, Dandan ZHANG, Jiyang ZHANG, Weihua XIONG. Terahertz Nondestructive Testing Signal Recognition Based on PSO-BP Neural Network[J]. Acta Photonica Sinica, 2021, 50(9): 0930004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites