Investigation of nanomachining-induced plastic behavior using machine learning-assisted high-throughput molecular dynamics simulations

The accurate relationship between machining-induced mechanical behavior and machining parameters is always expected to establish, for assisting to create the desired-quality machined components. However, this issue solved by the model, simulation, and experiment is extremely difficult due to the huge time-and-cost consumption. Here, a new approach combining machine learning and high-throughput molecular dynamic simulations is performed to investigate the evolution of dislocation behavior in a wide range of machining speeds. Using the trained machine learning surrogate model, a mass of "machining speed-dislocation behavior" data required can be obtained in a very short time compared to the traditional simulation. According to the plastic behavior of a workpiece, three machining speed ranges (low, medium, and high speed range) are divided. In a low speed range, the increasing machining speed accelerates the dislocation nucleation and multiplication, and thus causes the frequent dislocation interactions and the strong work hardening behavior in a workpiece. A medium/high machining speed leads to a shorter time for the destroyed crystal lattice to rearrange, which produces more noncrystal structures and simultaneously reduces the dislocation density. Furthermore, this finding reveals the deformation behavior dominated by a strong competition between the machining speed and the rate of dislocation nucleation. The present work provides a universal framework to accelerate the processing optimization for obtaining the high-quality products.