You are currently viewing a new version of our website. To view the old version click .
Materials
  • This is an early access version, the complete PDF, HTML, and XML versions will be available soon.
  • Review
  • Open Access

25 December 2025

Monitoring and Control of the Direct Energy Deposition (DED) Additive Manufacturing Process Using Deep Learning Techniques: A Review

,
,
,
,
,
,
and
1
College of Engineering, Ocean University of China, Qingdao 266400, China
2
Shandong Key Laboratory of Additive Manufacturing Technology & Equipment, Ocean University of China, Qingdao 266400, China
3
Shandong Liwei Laser Technology Co., Ltd., Taian 271208, China
4
College of Mechanical and Electronic Engineering, China University of Petroleum, Qingdao 266580, China
This article belongs to the Special Issue Advancements in Friction Stir-Based Solid-State Additive Manufacturing: Mechanisms, Microstructures, and Properties

Abstract

Directed Energy Deposition (DED), as a core branch of additive manufacturing, encompasses two typical processes: laser directed energy deposition (LDED) and wire and arc additive manufacturing (WAAM), which are widely used in manufacturing aerospace engine blades and core components of high-end equipment. In recent years, with the increasing adoption of deep learning (DL) technologies, the research focus in DED has gradually shifted from traditional “process parameter optimization” to “AI-driven process optimization” and “online real-time monitoring”. Given the complex and distinct influence mechanisms of key parameters (such as laser power/arc current, scanning/travel speed) on melt pool behavior and forming quality in the two processes, the introduction of artificial intelligence to address both common and specific issues has become particularly necessary. This review systematically summarizes the application of DL techniques in both types of DED processes. It begins by outlining DL frameworks, such as artificial neural networks (ANNs), recurrent neural networks (RNNs), convolutional neural networks (CNNs), and reinforcement learning (RL), and their compatibility with DED data. Subsequently, it compares the application scenarios, monitoring accuracy, and applicability of AI in DED process monitoring across multiple dimensions, including process parameters, optical, thermal fields, acoustic signals, and multi-sensor fusion. The review further explores the potential and value of DL in closed-loop parameter adjustment and reinforcement learning control. Finally, it addresses current bottlenecks such as data quality and model interpretability, and outlines future research directions, aiming to provide theoretical and engineering references for the intelligent upgrade and quality improvement of both DED processes.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.