基于PCA-BP神经网络对甲醛和甲醇的识别研究 
下载： 1162次
Recognition of Formaldehyde, Methanol Based on PCA-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. 实验系统框图
Fig. 1. Experimental system block diagram
下载图片 查看原文
图 2. 气体识别过程
Fig. 2. Gas identification process
下载图片 查看原文
图 3. 采集数据波形。(a)原始数据波形;(b)滤波后的数据波形
Fig. 3. Acquisition of data waveform. (a) Raw data waveform; (b) filtered data waveform
下载图片 查看原文
图 4. PCA-BP神经网络模型
Fig. 4. PCA-BP neural network model
下载图片 查看原文
图 5. PCA得分图。(a)前2个主成分;(b)前3个主成分
Fig. 5. PCA score chart. (a) First two main components; (b) first three main components
下载图片 查看原文
图 6. BP神经网络分类结果。(a) A-PCA-BP分类;(b) W-PCA-BP分类
Fig. 6. Classification results of BP neural network. (a) A-PCA-BP classification; (b) W-PCA-BP classification
下载图片 查看原文
表 1气体传感器响应值的相对标准差
Table1. Relative standard deviation of gas sensor response
| Senor | Formaldehyde /μL | Methanol /μL |
|---|
| 2 | 4 | 6 | 2 | 4 | 6 |
|---|
| 1 | σμ | 0.0666 | 0.0621 | 0.0441 | 0.0031 | 0.0027 | 0.0021 | | 3.2691 | 3.6191 | 4.9390 | 3.9663 | 4.3863 | 5.9063 | | 2 | σ | 0.0116 | 0.0102 | 0.0074 | 0.0048 | 0.0036 | 0.0027 | | μ | 3.9011 | 4.4411 | 6.1211 | 3.5445 | 4.7445 | 6.3245 | | 3 | σ | 0.0071 | 0.0066 | 0.0066 | 0.0039 | 0.0025 | 0.0018 | | μ | 4.5214 | 4.8714 | 6.0893 | 2.8882 | 4.5382 | 6.3982 | | 4 | σ | 0.0115 | 0.0096 | 0.0096 | 0.0056 | 0.0041 | 0.0031 | | μ | 3.8311 | 4.5811 | 6.6411 | 3.1325 | 4.3825 | 5.8625 | | 5 | σ | 0.0112 | 0.0087 | 0.0074 | 0.0099 | 0.0086 | 0.0064 | | μ | 3.5972 | 4.5972 | 5.4572 | 3.7638 | 4.3438 | 5.8338 | | 6 | σ | 0.0161 | 0.0131 | 0.0107 | 0.1048 | 0.0669 | 0.0669 | | μ | 3.4628 | 4.3128 | 5.2388 | 0.9534 | 1.4934 | 2.0161 | | 7 | σ | 0.0261 | 0.0221 | 0.0165 | 0.3765 | 0.3369 | 0.3369 | | μ | 3.4741 | 4.1141 | 5.4641 | 3.0595 | 3.1495 | 5.1293 |
|
查看原文
表 2部分W矩阵数据相似度标征
Table2. Data similarity of partial W matrix
| τi | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|
| 1 | 1 | 0.2455 | -0.3589 | -0.4109 | 0.1243 | 0.1406 | 0.3054 | | 2 | 0.2455 | 1 | -0.0841 | 0.0651 | -0.4414 | 0.1488 | 0.5628 | | 3 | -0.3589 | -0.0841 | 1 | 0.2956 | -0.2973 | 0.0211 | -0.3623 | | 4 | -0.4109 | 0.0651 | 0.2956 | 1 | 0.0373 | -0.1307 | 0.0462 | | 5 | 0.1243 | -0.4414 | -0.2973 | 0.0373 | 1 | -0.0602 | 0.0502 | | 6 | 0.1406 | 0.1488 | 0.0211 | -0.1307 | -0.0602 | 1 | 0.0583 | | 7 | 0.3054 | 0.5628 | -0.3623 | -0.0462 | 0.0502 | 0.0583 | 1 |
|
查看原文
表 3W-PCA-BP网络对矩阵A识别结果
Table3. Recognition results of matrix A on W-PCA-BP network
| Sampletype | Studysamples /piece | Recognitionresult /piece | Identificationerror /% |
|---|
| Formaldehyde | 30 | 31 | 3.3 | | Methanol | 30 | 29 | 3.3 |
|
查看原文
表 4矩阵A训练的网络对甲醛和甲醇的识别
Table4. Identification of formaldehyde and methanol bymatrix A trained network
| Method | Formaldehydeidentification samples | Methanol identificationsamples | Total numberof samples | Number oferror /piece | Processingtime /s |
|---|
| A-BP | 38 | 22 | 60 | 8 | 3.2 | | W-BP | 100 | 100 | 200 | 8 | 3.5 | | A-PCA-BP | 35 | 25 | 60 | 5 | 3.0 | | W-PCA-BP | 94 | 106 | 200 | 6 | 3.3 |
|
查看原文
表 5矩阵W训练的网络对甲醛和甲醇的识别
Table5. Identification of formaldehyde and methanol bymatrix W trained network
| Method | Formaldehydeidentification samples | Methanol identificationsamples | Total numberof samples | Number oferror /piece | Processingtime /s |
|---|
| A-BP | 27 | 33 | 60 | 3 | 4.5 | | W-BP | 102 | 98 | 100 | 2 | 5.0 | | A-PCA-BP | 31 | 29 | 60 | 1 | 2.8 | | W-PCA-BP | 95 | 105 | 200 | 5 | 3.1 |
|
查看原文
宋海声, 麻林召, 王一帆, 朱恩功, 李承飞. 基于PCA-BP神经网络对甲醛和甲醇的识别研究[J]. 激光与光电子学进展, 2020, 57(7): 071201. Haisheng Song, Linzhao Ma, Yifan Wang, Engong Zhu, Chengfei Li. Recognition of Formaldehyde, Methanol Based on PCA-BP Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(7): 071201.
PDF全文
