Skip to main content
U.S. flag

An official website of the United States government

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

Deep Representation Learning for Process Control in Laser Additive Manufacturing



Yan Lu, Sepehr Fathizadan, Feng Ju


Laser Powder Bed Fusion (LPBF) is a technology of additive manufacturing that involves with applying laser to fabricate parts by layer upon layer fusion of powder. Print quality has long been a major barrier for its widespread implementation. Retroactive/offline actions such as post-manufacturing inspections to detect the defects in finished products are not only expensive and time consuming but are also unable to issue corrective action signals in real time during the build span. In-situ monitoring methods by relying on the in-process sensor data, on the other hand, can provide viable alternatives to aid with online detection of anomalies during the process. Given that the characteristics of melt pool are of crucial importance to the quality of final products, this paper provides a framework to process the melt pool images by a configuration of Convolutional Auto-Encoder (CAE) neural networks. The corresponding bottleneck layer in the network learns a deep yet low- dimensional representation from melt pools while preserving the spatial correlation and complex features intrinsic in the images. As opposed to manual annotation of data by X-ray imaging or destructive tests, an agglomerative clustering algorithm is applied to these representations to automatically extract the anomalies and annotate the data accordingly. In an effort to provide an accurate visualization tool for practitioners, a Hotteling's T2 and S2 control charting scheme is developed to monitor the stability of process by keep tracking of the learned deep representation and the obtained residuals from the reconstruction of original images, respectively. Testing the proposed methodology on the collected data from an experimental build demonstrates that the method can extract a set of complex features that are inextricable otherwise by hand-crafted feature engineering methods. Furthermore, the integration of a statistical process monitoring scheme is shown to be capable of detecting the an
Additive Manufacturing


Laser powder bed fusion, additive manufacturing melt pool image deep learning anomalydetection


Lu, Y. , Fathizadan, S. and Ju, F. (2021), Deep Representation Learning for Process Control in Laser Additive Manufacturing, Additive Manufacturing, [online], (Accessed September 27, 2023)
Created April 26, 2021