Quantitative Analysis of NaCl Aerosols Based on Convolutional Neural Network and Filament‐Induced Fluorescence Spectroscopy

Mingming Liu; Desheng Kong; Yuyan Xiang; Fengyuan Zhao; Jing Zhang; Ruipeng Zhang; Yamin Gao; Chenhao Zhi; Yue Liu; Maoqiang Xie; Zhi Zhang; Lu Sun; Xing Zhao; Nan Zhang; Weiwei Liu

doi:10.3788/CJL221489

Chinese Journal of Lasers, Volume. 50, Issue 7, 0708007(2023)

Quantitative Analysis of NaCl Aerosols Based on Convolutional Neural Network and Filament‐Induced Fluorescence Spectroscopy

Mingming Liu¹, Desheng Kong¹, Yuyan Xiang¹, Fengyuan Zhao¹, Jing Zhang¹, Ruipeng Zhang¹, Yamin Gao¹, Chenhao Zhi¹, Yue Liu¹, Maoqiang Xie^1、*, Zhi Zhang², Lu Sun², Xing Zhao², Nan Zhang², and Weiwei Liu²

Author Affiliations

¹College of Software, Nankai University, Tianjin 300350, China

²Institute of Modern Optics, Nankai University, Tianjin 300350, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Schematic of filament-induced fluorescence spectrum (FIFS) collection device

Download full size

Fig. 2. Filament induced by femtosecond laser in NaCl aerosol

Download full size

Fig. 3. Filament-induced fluorescence spectra of NaCl aerosol with five mass concentrations near 589 nm

Download full size

Fig. 4. Calibration regression curve of NaCl aerosol’s filament-induced fluorescence spectra

Download full size

Fig. 5. Architecture of one-dimensional CNN for predicting mass concentration of NaCl aerosol

Download full size

Fig. 6. Prediction comparison of each model in full and characteristic spectra. (a)-(d) Full spectrum; (e)-(h) characteristic spectrum

Download full size

Table 1. Mass concentration prediction results for full spectral data
View table
Table 1. Mass concentration prediction results for full spectral data
Mass concentration prediction model R² RMSE RPD MAE ACC
MLR 0.991 0.185 10.581 0.144 0.93
PLSR 0.968 0.326 6.018 0.259 0.82
BPNN 0.985 0.230 8.572 0.167 0.91
1D-CNN 0.997 0.110 18.478 0.073 0.99

Table 2. Mass concentration prediction results for characteristic spectral data
View table
Table 2. Mass concentration prediction results for characteristic spectral data
Mass concentration prediction model R² RMSE RPD MAE ACC
CR 0.818 0.768 2.546 0.682 0.40
PLSR 0.990 0.198 10.058 0.154 0.90
BPNN 0.993 0.159 12.823 0.104 1
1D-CNN 0.997 0.110 18.702 0.071 1

Table 3. Generalized prediction experimental results for full spectral data
View table
Table 3. Generalized prediction experimental results for full spectral data
Mass concentration prediction model RMSE MAE ACC
MLR 1.175 1.175 0.43
PLSR 1.102 1.089 0.18
BPNN 0.614 0.606 0.64
1D-CNN 0.687 0.676 0.42

Table 4. Generalized prediction experimental results for characteristic spectral data
View table
Table 4. Generalized prediction experimental results for characteristic spectral data
Mass concentration prediction model RMSE MAE ACC
CR 0.776 0.878 0.32
PLSR 3.100 3.084 0.42
BPNN 0.669 0.601 0.69
1D-CNN 0.329 0.311 0.87

Table 5. Generalized prediction experimental results of 1D-CNN in characteristic spectral data

View table

Table 5. Generalized prediction experimental results of 1D-CNN in characteristic spectral data

Actual mass concentration /（mg·m^-3）	RMSE	MAE	ACC
Average	0.34	0.31	0.87
0.33	0.31	0.31	0
0.66	0.10	0.10	0.97
1.32	0.24	0.24	0.97
1.98	0.37	0.34	0.83
2.64	0.24	0.21	0.97
3.30	0.31	0.29	1
3.96	0.50	0.44	0.93
4.62	0.29	0.26	1
5.28	0.34	0.31	1
6.61	0.60	0.57	1

Tools

Get Citation

Copy Citation Text

Mingming Liu, Desheng Kong, Yuyan Xiang, Fengyuan Zhao, Jing Zhang, Ruipeng Zhang, Yamin Gao, Chenhao Zhi, Yue Liu, Maoqiang Xie, Zhi Zhang, Lu Sun, Xing Zhao, Nan Zhang, Weiwei Liu. Quantitative Analysis of NaCl Aerosols Based on Convolutional Neural Network and Filament‐Induced Fluorescence Spectroscopy[J]. Chinese Journal of Lasers, 2023, 50(7): 0708007

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites