论文信息 - BP Neural Network based on Dropout Applied to the EDXRF Quantitative Analysis of Heavy Metal Elements

BP Neural Network based on Dropout Applied to the EDXRF Quantitative Analysis of Heavy Metal Elements

BP neural network has been used in EDXRF quantitative analysis, and the improvement of calculating model is one of the vital research fields. Dropout is considered as an optimization method to solve the over-fitting problem of BP. Therefore, the new experiment of applying the optimized dropout on small sample BP neural network in the quantitative analysis of XRF is proposed. 100 groups of experimental samples are used for establishing model and comparative analysis. The results verify that the optimized BP algorithm can effectively reduce over-fitting problem, and ensure the quality of data, with high confidence and fast operation time.

[1] Liqiang Luo,et al. Chemometrics and its applications to x-ray spectrometry , 2006 .

[2] E. Costa,et al. Correlation methods for the analysis of X-ray polarimetric signals , 2018, 1801.01391.

[3] Fei Li,et al. Application of artificial neural networks to X‐ray fluorescence spectrum analysis , 2019, X-Ray Spectrometry.

[4] Nitish Srivastava,et al. Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[5] Eugene P. Bertin,et al. Principles and Practice of X-Ray Spectrometric Analysis , 1970 .

[6] L. Ge,et al. Recent developments on XRF spectra evaluation , 2020, Applied Spectroscopy Reviews.

[7] B. Kanngießer,et al. Handbook of practical X-ray fluorescence analysis , 2006 .

[8] J. Lowenstern. Evidence for a copper-bearing fluid in magma erupted at the Valley of ten thousand smokes, Alaska , 1993 .

[9] G. Weltje,et al. Calibration of XRF core scanners for quantitative geochemical logging of sediment cores: Theory and application , 2008 .