Gaussian Process (GP)-based Learning Control of Selective Laser Melting Process

TL;DR

This paper introduces a Gaussian process-based data-driven model and model predictive control for the complex selective laser melting process, validated through high-fidelity simulations to improve melt pool regulation.

Contribution

It presents a novel Gaussian process dynamic model and a model predictive control approach specifically tailored for SLM, considering physical constraints.

Findings

The GP model accurately predicts SLM dynamics.

The MPC effectively regulates melt pool size.

Validation shows superior performance over existing methods.

Abstract

Selective laser melting (SLM) is one of emerging processes for effective metal additive manufacturing. Due to complex heat exchange and material phase changes, it is challenging to accurately model the SLM dynamics and design robust control of SLM process. In this paper, we first present a data-driven Gaussian process based dynamic model for SLM process and then design a model predictive control to regulate the melt pool size. Physical and process constraints are considered in the controller design. The learning model and control design are tested and validated with high-fidelity finite element simulation. The comparison results with other control design demonstrate the efficacy of the control design.