研究目的
Investigating the use of artificial neural networks coupled with Monte Carlo simulations for the analysis of gamma spectra obtained from neutron activation analysis, specifically for quantifying trace elements in kidney stones.
研究成果
The study successfully demonstrated the use of artificial neural networks coupled with Monte Carlo simulations for analyzing gamma spectra from neutron activation analysis. The Levenberg–Marquardt algorithm with a 5-23-5 structure ANN provided optimal results for quantifying trace elements in kidney stones, with the largest difference observed for Zn. The method offers a time-efficient alternative to traditional spectrum analysis, with potential for further improvement through better correction methods and more precise measurements.
研究不足
The study faced limitations in the accuracy of Zn concentration estimation due to its long decay half-life. The correction between experimental and simulated spectra was imperfect, leading to larger differences in experimental data analysis. The method requires precise mass measurement and more sensitive neutron activation analysis for improved accuracy.
1：Experimental Design and Method Selection:
The study employed Monte Carlo simulations to generate training data sets for an artificial neural network (ANN) to analyze gamma spectra from neutron activation analysis (NAA). A Levenberg–Marquardt algorithm with a 5-23-5 structure ANN was used for quantitative analysis.
2：Sample Selection and Data Sources:
Four different types of kidney stones (Struvite, Apatite, Calcium Oxalate Monohydrate, and Uric Acid) were irradiated in the Missouri University of Science and Technology Reactor (MSTR). A 5 mg NaCl sample was also irradiated for calibration.
3：List of Experimental Equipment and Materials:
MCNP6 and CINDER’90 codes were used for simulations. MATLAB neural network toolbox was employed to construct the ANN. ROOT software was used for peak analysis.
4：Experimental Procedures and Operational Workflow:
Kidney stones were irradiated for 3 h at 100 kW power level. The samples were prepared in powder form. Gamma spectra were analyzed using ROOT to extract peak features.
5：Data Analysis Methods:
The ANN was trained with simulated data sets. Peak counts from spectra were used to train the ANN for element identification and quantification.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
