News & Analysis
/
Article

Predicting cracks during laser-directed energy deposition of ceramics

NOV 26, 2025
Machine learning models trained on a framework that links thermal characteristics to crack rate can accurately predict crack formation.
Predicting cracks during laser-directed energy deposition of ceramics internal name

Predicting cracks during laser-directed energy deposition of ceramics lead image

In laser-directed energy deposition (LDED), an additive manufacturing process, a laser melts and deposits a material layer by layer. This efficient and cost-effective process could fabricate ceramic components for defense, aerospace, and biomedical applications. However, LDED often leads to performance-degrading cracks in brittle ceramics.

Researchers lack practical methods to predict and prevent the formation of these cracks. Li et al. developed an approach that combines thermal feature analysis and machine learning to accurately predict crack formation during LDED of ceramics. This method eliminates the need for measuring internal residual stress, a key contributor to crack formation that is challenging to measure directly and in real time.

The rapid heating and cooling in LDED drive the buildup of residual stress. Traditional methods, which try to predict crack formation using fixed-point temperature sampling, fail to take into account the dynamic behavior of the melt pool, where the material is molten. In contrast, the authors introduced a strategy that extracts stable, frequency-domain thermal features directly from dynamic contour points, which outline the evolving melt pool.

They found that this framework enables a more robust and mechanistic correlation between melt-pool thermal dynamics and resultant crack formation, improving the accuracy and interpretability of crack-rate prediction.

“This work establishes a data-driven and physically interpretable framework for predicting crack-rate in ceramic additive manufacturing,” said author Haiying Wei. “The proposed method enhances the reliability and accuracy of defect prediction and supports the future development of process monitoring and control strategies in laser-directed energy deposition.”

Next, the authors will extend this approach to monitor other features of a material during LDED.

Source: “Physically interpretable crack prediction using dynamic contour point-based frequency-domain gray-level features in laser-directed energy deposition of alumina ceramics,” by Bin Li, Haiying Wei, Yi Zhang, Yuchao Lei, Xuni Yin, and Shiyu Cao, Journal of Laser Applications (2025). The article can be accessed at https://doi.org/10.2351/7.0001969 .

More Science
/
Article
Approach provides way for designers to assess losses in the driving bearing in compressors and optimize their efficiency
/
Article
Fluid dynamics study reveals how air flow spreads cooling created with urban greening projects.
AAS
/
Article
Though less destructive than coronal mass ejections, moderate space weather events like stream interaction regions can still cause geomagnetic storms. Researchers have used machine learning to investigate the drivers behind these events.
/
Article
Researchers created a programmable polymer film that can change its surface texture and color on demand.