Digital twins involve the transfer of a physical entity into a virtual model, enabling the simulation, analysis, and optimization of physical products, processes, operations, or systems. This paper aims to explore the current state of research and development in digital twins and their applications within complex industrial processes. It examines concepts, government development strategies, and ongoing applications. The primary approach of this study involves analyzing research findings related to the status of multi-agent based digital twinning, as well as the design, manufacturing, and operation and maintenance of digital twins in intelligent manufacturing. Furthermore, this paper presents a digital twin scheme for continuous production in a blast furnace, along with a multi-agent discrete manufacturing scheme for large aircrafts. The digital twin of the blast furnace incorporates a comprehensive component field model and a small incremental learning model. It leverages computer vision to implement an incremental compensation mechanism within the complex process industry. In the realm of complex industrial manufacturing, digital twins and multi-agent technologies have the potential to enhance production efficiency and quality, reduce energy consumption and waste generation, while also mitigating complexity, safety risks, and costs.