Implementation of Stochastic Gradient Descent in an Automated Glow Peak Identification Software for Multiple...

Area of Science:

• Physics
• Materials Science
• Computational Science

Background:

• Automated analysis of thermoluminescent dosimeter (TLD) glow curves is crucial for accurate radiation dosimetry.
• Previous C++ code utilized derivative-based peak detection and Levenberg-Marquardt fitting.
• Enhancements were needed to improve the robustness and accuracy of TLD glow curve analysis.

Purpose of the Study:

• To develop and evaluate an improved C++ code for automated thermoluminescent dosimeter glow curve analysis.
• To enhance automated peak detection and curve fitting algorithms for TLD data.
• To compare the performance of the new software version against the previous one using figures of merit.

Main Methods:

• Implemented Stochastic Gradient Descent Algorithm for automated peak detection, replacing derivative methods.
• Incorporated early stopping mechanisms to optimize the Levenberg-Marquardt Algorithm for curve fitting.
• Performed comparative glow curve analysis on various thermoluminescent dosimeter materials.

Main Results:

• The updated code demonstrated improved automated peak detection capabilities.
• A significant reduction in the mean figure of merit (approximately 46%) was observed.
• The number of detected peaks increased compared to the previous software version.

Conclusions:

• The enhanced glow curve analysis code offers superior performance for TLD data processing.
• The new algorithms lead to more accurate and reliable dosimetry results.
• This advancement contributes to improved radiation measurement accuracy in various applications.