Role of inter-layer dwell time on residual stress generation of a thin wall structure by directed energy deposition of ferritic steel

Abstract

The quality of thin-wall printed components by directed energy deposition (DED) technique rely on physical basis of interface characteristics where inter-layer dwell time $\left(D_{\text{TM}}\right)$ plays a significant role. The numerical computation of transient thermal and residual stress profiles is time-consuming. Moreover, in-situ monitoring and optimization of the deposition attributes following the process simulation is difficult-to-use directly. The use of data-driven deep learning (DL) model to understand the role of $D_{\text{TM}}$ on thermo-mechanical responses enhances the physical basis of DED process of a printed thin-wall by implying the augmented state of process parameters. The optimization route of dwell time is proposed using the DL models to find a way of mitigating the tensile residual stress. A well-tested finite element (FE) model is employed to generate an adequate thermal and highly nonlinear residual stress path at different levels of dwell times within the known experimental conditions. The rigorous validation of physics-informed DL (PIDL) model with experimental data enables the prediction of non-linear thermal and stress profiles quite a faster rate than only a physics-based numerical model. The PIDL model interprets the interpolated and extrapolated datasets as a function of dwell times with reasonable accuracy. Further, inverse prediction of dwell times indicates the versatility of the developed model. Overall, the study shows the development of computationally efficient models for large parameter-based thin-walled structure development by coupling the physical basis of the DED process and DL algorithms.