In metal 3D printing (laser powder-bed fusion), thin layers of powder are spread in a chamber with a controlled atmosphere and are exposed to a laser beam. Following the specific scan vectors of the laser, the powder is consolidated into solid metal, ultimately forming the component. As a result, the process is extremely complex and involves high-energy light-matter interactions, rapid melting and solidification, non-equilibrium microstructure evolution, and the introduction of residual stress. 

Due to this complexity, finding the optimal printing conditions for a new component, a new alloy composition or powder batch often relies on trial-and-error. Several combinations of laser power, scan speed, and hatch spacing are put into tests until satisfactory values are found. However, this approach requires many samples, and it is material and labour-intensive, time-consuming, and costly.

An alternative to the trial-and-error approach is the use of advanced modelling and simulation of the 3D printing process. Performing the tests digitally is economically attractive, however, computer modelling demands deep scientific knowledge and understanding due to its complexity. Furthermore, the development of an actual software product requires extensive capabilities in scientific computing, high-performance parallel computations and software development. 

To address these challenges, A*STAR IHPC has developed the Additive Manufacturing Digital Twin (AM-DT), a software platform that allows users to generate virtual simulations of manufacturing to identify optimal printing conditions, requiring minimal physical prototyping or testing. 

At its foundation, the simulation platform is composed of two main components, namely, the server and the client. The two components are connected to each other through TCP/IP. The server component is where all numerical calculations are performed, intended to be compiled on a server-class computer and is scalable from workstations to supercomputers. In fact, the server takes full advantage of the message-passing interface (MPI) communication protocol to handle heavy workloads in parallel. While performing the calculations, the server also remains responsive because of the built-in asynchronous multithreading. This implementation has the further advantage of allowing multiple clients to connect to the same server. Because of its focus on numerical calculations, the server is compiled with processor-dependent flags to achieve maximum performance and tuned for Linux operating systems. Intentionally, the server does not have a graphical user interface (GUI) and is accessible only through TCP/IP. 

The user interacts with the client component, which is designed to be lightweight, portable, and focused on the GUI. For example, the client can be installed on both Windows and Linux computers and can run on a laptop. This architecture allows the server and the client to be run in different geographical locations, as well as on cloud infrastructures. The client is built on the Qt5 framework and its rendering is entirely developed using the VTK library.