基于激光诱导击穿光谱与GA-BP神经网络的塑料分类识别 
下载： 1152次
Plastic Classification and Recognition by Laser-Induced Breakdown Spectroscopy and GA-BP Neural Network
西北师范大学物理与电子工程学院, 甘肃, 兰州 730070
图 & 表
图 1. BP神经网络的组成
Fig. 1. Composition of BP neural network
下载图片 查看原文
图 2. GA-BP神经网络的流程图
Fig. 2. Flow chart of GA-BP neural network
下载图片 查看原文
图 3. 实验装置的原理图
Fig. 3. Schematic diagram of the experimental device
下载图片 查看原文
图 4. 激光灼烧样品表面的离焦状态。(a)正焦点;(b)焦点;(c)负焦点
Fig. 4. Defocused state of laser burning sample surface. (a) Positive focus; (b) focus; (c) negative focus
下载图片 查看原文
图 5. 特征谱线与离焦量的关系
Fig. 5. Relationship between the characteristic spectral line and the amount of defocusing
下载图片 查看原文
图 6. ABS的发射光谱。(a)原始光谱;(b)处理后的光谱
Fig. 6. Emission spectrum of ABS. (a) Original spectrum; (b) spectrum after treatment
下载图片 查看原文
图 7. 不同塑料样品的前三个主成分散点图
Fig. 7. First three main component dispersion points of different plastic samples
下载图片 查看原文
图 8. PCA-GA-BP神经网络的预测结果
Fig. 8. Prediction results of PCA-GA-BP neural network
下载图片 查看原文
图 9. 不同算法的性能对比。(a) GA;(b) PCA-BP神经网络;(c) PCA-GA-BP神经网络
Fig. 9. Performance comparison of different algorithms. (a) GA; (b) PCA-BP neural network; (c) PCA-GA-BP neural network
下载图片 查看原文
图 10. 三种神经网络的分类误差数
Fig. 10. Classification errors of three neural networks
下载图片 查看原文
表 19种塑料样品的分子式和颜色
Table1. Molecular formula and color of 9 plastic samples
| Sample | Molecular formula | Color |
|---|
| ABS | (C58H64N2)n | yellow opaque | | PC | (C16H14O3)n | light yellow-transparent | | PP | [CH2CH(CH3)]n | white opaque | | PE | (C2H4)n | white opaque | | POM | (CH2O)n | white opaque | | PU | (CHNO2)n | yellow black-transparent | | PS | (C8H8)n | colorless and transparent | | PA-6 | (C6H11ON)n | light yellow-transparent | | PMMA | (C5H8O2)n | colorless and transparent |
|
查看原文
表 2特征谱线与波长
Table2. Characteristic spectral line and wavelength
| Characteristic spectral line | Wavelength /nm |
|---|
| C(I) | 247.86 | | Mg (II) | 279.55 | | Al (I) | 309.27 | | Ti (II) | 334.90 | | C-N | 388.30 | | Ca (II) | 393.34 | | F(II) | 429.91 | | C2 | 516.50 | | Na (I) | 589.06 | | H(I) | 656.30 | | CL(I) | 725.70 | | F(I) | 739.90 | | N (I) | 746.90 | | K(I) | 766.50 | | O (I) | 777.30 |
|
查看原文
表 3三种神经网络的训练结果
Table3. Training results of three neural networks
| Neural networks | Total-error /piece | Mean-error /piece | Total training-time /s | Average-time /s | Average recognition-accuracy /% |
|---|
| PCA-BP | 304 | 6.08 | 46.24 | 0.93 | 98.31 | | GA-BP | 155 | 3.10 | 113.60 | 2.27 | 99.14 | | PCA-GA-BP | 33 | 0.66 | 82.10 | 1.64 | 99.82 |
|
查看原文
宋海声, 麻林召, 朱恩功, 王一帆, 刘宇平, 孙文健, 彭鹏, 李承飞. 基于激光诱导击穿光谱与GA-BP神经网络的塑料分类识别[J]. 激光与光电子学进展, 2020, 57(15): 153002. Haisheng Song, Linzhao Ma, Engong Zhu, Yifan Wang, Yuping Liu, Wenjian Sun, Peng Peng, Chengfei Li. Plastic Classification and Recognition by Laser-Induced Breakdown Spectroscopy and GA-BP Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(15): 153002.
PDF全文
