Enhance Wire Arc Additive Manufacturing Performance for Fabricating Complex Parts Using Artificial Intelligence: Learning, Prediction and Control

Direct Energy Deposition (DED) is an advanced additive manufacturing technique that uses focused thermal energy to fuse materials by melting them as they are deposited. This process allows for precise control over material placement, enabling the creation of complex geometries and the repair of existing components. A specialized form of DED, Wire Arc Additive Manufacturing (WAAM), utilizes an electric arc as the heat source and wire as the feedstock material. WAAM offers several advantages, including low capital investment and higher manufacturing efficiency, making it an attractive option for producing medium to large-sized parts, particularly in the maritime and aerospace industries. Over the past decade, advancements in WAAM technology have enabled more reliable and cost-effective manufacturing solutions, solidifying its role in the production of critical components for these sectors.

However, with the increasing demand for high-quality, complex metal parts in the industry, the traditional WAAM process which is primarily used for manufacturing large parts with simple and regular geometrical features, faces significant challenges. The main challenge in fabricating complex parts lies in optimizing deposition parameters to ensure the desired dimensions and mechanical properties of the final part. Traditional WAAM processes often struggle with adaptive control at fine scales, especially when fabricating custom parts with irregular or intricate geometries. Achieving precise control of weld bead geometry is essential, as deviations can lead to structural defects and compromised part quality. Furthermore, complex deposition paths increase the risk of defects such as porosity and incomplete fusion, necessitating advanced monitoring techniques to maintain production standards and prevent defects from propagating.

To address these challenges, this thesis proposes an intelligent WAAM (IWAAM) system that integrates cutting-edge AI techniques to improve the fabrication process for high-quality, complex parts. The key contributions of this work are: (i) the development of an automatic bead modelling system for solid parts, powered by machine learning algorithms capable of predicting and optimizing deposition parameters to ensure consistency and precision across varying geometries; (ii) a novel parameter optimization method for thinwall parts that harnesses and quantifies the expertise of skilled human welders, translating their decision-making processes into a systematic framework for adaptive control; (iii) an in-depth investigation into the flexibility enhancement provided by the surface tension transfer (STT) process, highlighting its benefits over the commonly used cold metal transfer (CMT) process in terms of more nuanced parameter adjustments and adaptability for complex geometries; and (iv) the design and implementation of a real-time monitoring system that employs recurrent neural networks to process sequential molten pool images, enabling timely detection and correction of potential defects during the fabrication process.

Experimental validation of the proposed IWAAM system involved fabricating multiple complex parts, demonstrating the system’s effectiveness in achieving high-quality results with enhanced control over bead geometry and defect mitigation. The research significantly advances the development of intelligent WAAM processes, paving the way for more sophisticated and widely adoptable manufacturing solutions within the industry.