Prediction of Cr, Mn, and Ni in Medium and Low Alloy Steels by GA-BP Neural Network Combined with EDXRF Technology

Haisheng Song; Zhao Chen; Dacheng Xu; Rongwang Xu

doi:10.3788/LOP202259.1234001

Journals >Laser & Optoelectronics Progress >Volume 59 >Issue 12 >Page 1234001 > Article

Laser & Optoelectronics Progress
Vol. 59, Issue 12, 1234001 (2022)

Prediction of Cr, Mn, and Ni in Medium and Low Alloy Steels by GA-BP Neural Network Combined with EDXRF Technology

Haisheng Song¹, Zhao Chen^1、*, Dacheng Xu², and Rongwang Xu³

Author Affiliations

¹School of Physics and Electronic Engineering, Northwest Normal University, Lanzhou 730070, Gansu , China

²School of Electronic Information, Soochow University, Suzhou 215031, Jiangsu , China

³Kunshan Soohow Instrument Technology Co., Ltd., Suzhou 215300, Jiangsu , China

show less

DOI: 10.3788/LOP202259.1234001 Cite this Article Set citation alerts

Haisheng Song, Zhao Chen, Dacheng Xu, Rongwang Xu. Prediction of Cr, Mn, and Ni in Medium and Low Alloy Steels by GA-BP Neural Network Combined with EDXRF Technology[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1234001 Copy Citation Text

show less

Fig. 1. Structure diagram of BP neural network

Download full size

Fig. 2. Schematic of background subtraction by two-point method

Download full size

Fig. 3. Mean square error curve of training process

Download full size

Fig. 4. Linear regression results of training process

Download full size

Fig. 5. Error distributions of analysis results of three elements by GA-BP and FP methods. (a) Cr element; (b) Mn element; (c) Ni element

Download full size

Sample name	Cr	Mn	Ni
18CrNiW	0.340-1.830	0.120-0.710	2.680-4.580
30CrMnSiNiA	0.587-1.592	0.744-1.607	0.971-2.064
38CrMoAl	0.995-2.028	0.161-0.691	0.069-0.443
CrWMn	0.319-1.354	0.172-1.280	0.037-0.245
GSBH40067-93	0.581-1.950	0.216-1.350	0.229-2.290
YSBS15301-94	0.313-1.520	0.292-1.310	0.521-3.180

Table 1. Element content range of experimental samples of various alloy steels

Parameter	Numeric value	Parameter	Numeric value
Training sample	108×3	Output layer Activation function	purelin
Test sample	36×3	Population size	45
Input neuron	3	Iteration ordinal Number	20
Hidden layer neuron	5	Learning rate	0.1
Output layer neuron	3	Training steps	1000
Training function	trainlm	Crossover rate	0.2
Hidden layer Activation function	tansig	Mutation rate	0.0025

Table 2. Important parameters of GA-BP neural network

Sample	Analysis method	Cr		Mn		Ni
Sample	Analysis method	Predicted value	Relative standard deviation	Predicted value	Relative standard deviation	Predicted value	Relative standard deviation
CrWMn1	Standard	0.319		1.280		0.037
	FP	0.350	9.718	1.270	0.781	0.020	45.946
	GA-BP	0.330	3.448	1.285	0.391	0.057	54.054
CrWMn2	Standard	0.444		0.959		0.088
	FP	0.440	0.901	0.920	4.067	0.030	65.909
	GA-BP	0.447	0.676	1.017	6.048	0.151	71.591
CrWMn3	Standard	0.763		0.883		0.131
	FP	0.740	3.014	0.960	8.720	0.070	46.565
	GA-BP	0.749	1.835	0.899	1.812	0.101	22.901
CrWMn4	Standard	1.060		0.347		0.191
	FP	1.090	2.830	0.370	6.628	0.140	26.702
	GA-BP	1.073	1.226	0.379	9.222	0.151	20.942
CrWMn5	Standard	1.354		0.172		0.245
	FP	1.450	7.090	0.200	16.279	0.200	18.367
	GA-BP	1.391	2.733	0.144	16.279	0.261	6.531

Table 3. Prediction results of CrWMn series samples

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information