A highly efficient computational approach for part-scale microstructure predictions in Ti-6Al-4V additive manufacturing

TL;DR

This work develops a highly efficient, scan-resolved coupled thermo-microstructure framework for Ti-6Al-4V in additive manufacturing, enabling part-scale predictions with resolved laser tracks. The thermal field drives a phenomenological microstructure evolution consisting of $\alpha_s$, $\alpha_m$, and $\beta$ phases, modeled by diffusion-dominated and instantaneous transformations and solved per spatial point. The numerical strategy combines explicit/implicit time integration, adaptive meshing, and a vectorized implementation with fast polynomial approximations for transcendental functions, achieving strong scalability and modest overhead (often <10%) over the thermal solve. Demonstrations on cube and NIST AMBench cantilever geometries show the approach can capture scan-strategy–induced microstructural variations and deliver predictions within hours to days on modern HPC hardware, highlighting its practical potential for process optimization and microstructure-informed design.

Abstract

Fast and efficient simulations of metal additive manufacturing (AM) processes are highly relevant to exploring the full potential of this promising manufacturing technique. The microstructure composition plays an important role in characterizing the part quality and deriving mechanical properties. When complete parts are simulated, one often needs to resort to strong simplifications such as layer-wise heating due to the large number of simulated time steps compared to the small time step sizes. This article proposes a scan-resolved approach to the coupled thermo-microstructural problem. Building on a highly efficient thermal model, we discuss the implementation of a phenomenological microstructure model for the evolution of the three main constituents of Ti-6Al-4V: stable $α_s$-phase, martensite $α_m$-phase and $β$-phase. The implementation is tailored to modern hardware features using vectorization and fast approximations of transcendental functions. A performance model and numerical examples verify the high degree of optimization. We demonstrate the applicability and predictive power of the approach and the influence of scan strategy and geometry. Depending on the specific example, results can be obtained with moderate computational resources in a few hours to days. The numerical examples include a prediction of the microstructure on the full NIST AM Benchmark cantilever specimen.

Figures (14)

• Figure 1: Possible transformation paths between different microstructure phases.
• Figure 2: Illustration of vectorized microstructure algorithm. The algorithm works simultaneously on a local working copy of $n_\text{lanes}$ entries of the five global data vectors. Time integration, fixed-point iteration, and (optional) subcycling are all performed locally. All conditional computations are performed by blending the results of different conditional branches based on the condition mask.
• Figure 3: Following the arrows from top to bottom illustrates how to synthesize an IEEE754 double representation from a separate exponent $p$ (blue) and mantissa $m$ (red). Following the arrows from bottom to top shows the inverse operation, namely a decomposition of a double representation into exponent and mantissa.
• Figure 4: Impact of vectorization width on throughput of the microstructure model solved with an explicit or implicit scheme on Intel and AMD hardware.
• Figure 5: Roofline plot of a single time step performed in the microstructure model with an explicit or implicit scheme at different data block sizes. Data points move upwards along the dashed lines for increasing vectorization widths.