Parameter Optimization in Cluster Identification Algorithms for Characterizing Nanoclusters in Al-Mg-Si-Cu Alloys

Area of Science:

• Materials Science
• Computational Materials Science
• Metallurgy

Background:

• Accurate characterization of nanoclusters in aluminum alloys is essential for understanding material properties.
• The Density-based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is a common tool for this analysis.
• User-defined parameters in DBSCAN can significantly impact the accuracy of nanocluster identification.

Purpose of the Study:

• To optimize the user-defined parameters (Dmax, Nmin, order (K)) of the DBSCAN algorithm for nanocluster characterization.
• To minimize the formation of unphysical clusters in Al-0.9% Mg-1.0% Si-0.3% Cu alloy samples.
• To establish a reliable method for determining DBSCAN parameters for nanocluster analysis in Al-Mg-Si(-Cu) alloys.

Main Methods:

• Systematic optimization of DBSCAN parameters (Dmax, Nmin, K) for naturally aged (NA) and preaged (PA) samples.
• Analysis of cluster composition, size, atomic density, and atomic arrangement to identify and eliminate unphysical clusters.
• Validation of optimized parameters using volume rendering and isosurfacing to confirm high solute concentration regions.

Main Results:

• Identified and quantified four types of unphysical clusters.
• Determined optimal DBSCAN parameter combinations that minimize artificial clusters for both NA and PA samples.
• Confirmed that optimized parameters align with regions of high solute concentration, validating the method.

Conclusions:

• Tailoring DBSCAN parameters independently for each dataset is superior to using widely adopted fixed parameters.
• The proposed parameter determination strategy enhances the reliability of nanocluster characterization in aluminum alloys.
• This work introduces a robust approach for algorithm parameter selection in nanocluster analysis.