Recurrence network analysis of design-quality interactions in additive manufacturing

Area of Science:

• Materials Science
• Manufacturing Engineering
• Data Science

Background:

• Powder bed fusion (PBF) additive manufacturing (AM) offers design flexibility but faces quality control challenges, hindering widespread adoption.
• Advanced imaging and data processing are crucial for understanding design-quality interactions in complex AM builds.
• Thin-wall structures in AM are particularly sensitive to geometric parameters and build orientation.

Purpose of the Study:

• To investigate the interaction between design parameters and quality characteristics in PBF-AM thin-wall builds.
• To develop a novel method for analyzing spatial image data from AM processes.
• To establish a predictive model for design complexity's impact on build quality.

Main Methods:

• A generalized recurrence network (GRN) was developed to represent AM spatial image data.
• Network quantifiers (centralities) were extracted to characterize layerwise build quality.
• A regression model was used to predict the impact of design complexity on GRN behavior.

Main Results:

• Network features proved sensitive to build orientation, width, height, and contour space (p < 0.05).
• Thin-walls (>0.1 mm width) printed at 0° orientation exhibited superior quality compared to 60° and 90°.
• Build orientation 60° showed increased sensitivity to contour space variations, suggesting it should be avoided.

Conclusions:

• The proposed GRN-based design-quality analysis effectively characterizes PBF-AM builds.
• Specific build orientations and design parameters significantly influence thin-wall quality.
• This approach holds potential for optimizing engineering designs and improving PBF-AM quality control.