Graphic Learning Enabled Intelligent Optimizations of the Non-equilibrium Solidified Microstructure and Properties of Welded Titanium Alloy

Author of the article:WANG Xinzhao 1 , DU Kun 2 , WANG William Yi 1,2 , CHAI Zaixian 1 , ZHANG Jia 3 , YANG Chao 3 , CHEN Liyang 4 , LIU Xilin 4 , ZHANG Zhiyuan 4 , SUN Feng 5 , TANG Bin 1,2 , KOU Hongchao 1,2 , LI Jinshan

Author's Workplace：1. StateKeyLaboratoryofSolidificationProcessing,NorthwesternPolytechnicalUniversity,Xi'an710072,China;2. Innovation Center NPU, Chongqing 401135, China; 3. Western Superconducting Technologies Co., Ltd., Xi'an 710018, China; 4. Luoyang Ship Material Research Institute, Luoyang 471000, China; 5. Ti-MAST High Performance Alloy Co., Ltd., Chongqing 401135, China

Key Words：non-equilibrium solidification; titanium alloy; welding; machine vision; machine learning

Abstract： Manufacturing is transitioning into the “ intelligent manufacturing ” paradigm accelerated by the development of
information technologies. Artificial intelligence has become one kind of key generic technologyfrom shape control to
performance control for a wide variety of manufacturing processes, including precision casting, welding, additive
manufacturing, and so on. Moreover, the controlling mechanisms based on the non-equilibrium solidified microstructure and
properties are the fundamental and scientific problems to be addressed. It is essential to develop the integrated technology
that controls the shape and the performance spontaneously and build a perfect process quality system via advanced artificial
intelligence technology to enhance casting, welding, additive manufacturing, and other manufacturing processes towards a  more advanced level. Compared with traditional physical metallurgy and solidification, the key concept of controlling the
micro-area metallurgical process reveals that processes, such as additive manufacturing, welding, laser cladding repair, and
single crystal growth, can reveal the microstructure always dominated mechanisms related and morphology evolution laws
of solid-liquid two-phase regions under non-equilibrium conditions. In particular, the segregation behavior of alloy elements,
heat transfer and mass transfer characteristics of the interface, nucleation, growth, columnar to equiaxed transition,
competitive growth of dendrites, etc, are always dominated by mechanisms related to several processing parameters, such as
temperature gradients, solidification rates, and molten pool scales. The integration of calculation and experimental methods
will provide common technical and theoretical support for the composition-process-structure-performance regulation of
advanced metal materials in intelligent manufacturing, additive intelligent manufacturing, and space intelligent
manufacturing. The present paper is concerned with the control of the non-equilibrium solidification structure and the
properties of typical welding. Recent results revealing the advances in welding control technology based on visual detection
are reviewed and discussed. Correspondingly, the state-of-the-art applications and technical challenges of intelligent welding
in complex industrial environments are summarized, highlighting the graphic learning enabled intelligent optimizations of
the non-equilibrium solidified microstructure and properties of welded titanium alloy as a case study of digital twin
workshop.