Chinese Journal of Lasers, Volume. 49, Issue 9, 0911002(2022)

Hardness Characterization of GCr15 Steel Based on Laser-Induced Breakdown Spectroscopy and Random Forest

Zhu Li1,2, Qingyong Zhang1, Linghua Kong1,2、*, Guofu Lian1, and Peng Li1,2
Author Affiliations
  • 1School of Mechanical and Automotive Engineering, Fujian University of Technology, Fuzhou 350118, Fujian, China
  • 2Digital Fujian Industrial Manufacturing IoT Lab, Fuzhou 350118, Fujian, China
  • show less
    Figures & Tables(15)
    Schematic of LIBS experimental system
    Schematic of RF algorithm model
    LIBS images of different samples. (a) S1; (b) S2; (c) S3; (d) S4; (e) S5; (f) S6
    Variation trend of spectral line intensity of elements in sample S1 with pulse number. (a) Fe; (b) Cr
    Variation trend of spectral line intensity of elements in sample S6 with pulse number. (a) Fe; (b) Cr
    Correlation between ratio of spectral line intensity of ion to that of atom and hardness. (a) IFe Ⅱ 316.786/IFe Ⅰ 375.745; (b) ICr Ⅱ 482.413/ICr Ⅰ 302.067
    Gravel plots of first 15 principal components
    Predicted result of RF
    Effect of number of decision trees on model performance
    Effect of maximum number of features on model performance
    • Table 1. Chemical compositions of GCr15 steel

      View table

      Table 1. Chemical compositions of GCr15 steel

      ElementFeCrCSiMnPMo
      Mass fraction /%Bal.1.470.970.240.310.16≤0.10
    • Table 2. Heat-treatment process parameters and hardness of GCr15 steel

      View table

      Table 2. Heat-treatment process parameters and hardness of GCr15 steel

      Sample No.Heat-treatment process parameterHardness/ HV
      Oil quenchingTempering
      S1860 ℃+1 h150 ℃+2 h730.82
      S2250 ℃+2 h633.56
      S3350 ℃+2 h547.94
      S4450 ℃+2 h479.78
      S5550 ℃+2 h419.66
      S6650 ℃+2 h292.26
    • Table 3. Prediction results of RF model based on full-spectrum data

      View table

      Table 3. Prediction results of RF model based on full-spectrum data

      Sample No.Classification accuracy /%
      S10.833(40/48)
      S21(29/29)
      S30.870(20/23)
      S40.771(27/35)
      S51(31/31)
      S61(32/32)
      Average0.904
    • Table 4. Characteristic spectral lines used for RF model training

      View table

      Table 4. Characteristic spectral lines used for RF model training

      Line typeWavelength /nm
      Fe Ⅰ300.303 304.760 310.030 315.331 315.704
      358.119 364.742 367.687 375.745 511.034
      Fe Ⅱ296.994 302.547 316.786 325.877 327.349
      450.828 492.392 510.061 510.745 515.991
      Cr Ⅰ302.067 303.704 451.543 454.070 455.617
      458.390 460.010 465.472 469.894 483.165
      519.199 558.805
      Cr Ⅱ482.413
      Mn Ⅰ382.351,476.586,482.352
      Mn Ⅱ299.361
    • Table 5. Prediction results of RF model based on characteristic spectral lines

      View table

      Table 5. Prediction results of RF model based on characteristic spectral lines

      Sample No.Classification accuracy /%
      S10.979(47/48)
      S20.931(27/29)
      S30.957(22/23)
      S40.971(34/35)
      S50.968(30/31)
      S60.938(30/32)
      Average0.960
    Tools

    Get Citation

    Copy Citation Text

    Zhu Li, Qingyong Zhang, Linghua Kong, Guofu Lian, Peng Li. Hardness Characterization of GCr15 Steel Based on Laser-Induced Breakdown Spectroscopy and Random Forest[J]. Chinese Journal of Lasers, 2022, 49(9): 0911002

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: spectroscopy

    Received: Sep. 13, 2021

    Accepted: Oct. 8, 2021

    Published Online: Mar. 3, 2022

    The Author Email: Kong Linghua (15392030898@163.com)

    DOI:10.3788/CJL202249.0911002

    Topics