Active Control of Material Microstructure Evolution

Abstract

Material microstructure directly affects the physical properties. While metallurgists have long studied microstructure control through thermal processing, most existing methods are largely experience-based, essentially open-loop and only consider bulk properties. To fill these gaps, this thesis focuses on using closed-loop control and distributed thermal processing to actively regulate the evolution and spatial distribution of microstructure.

To link this work with practical applications, we conduct case studies under two typical thermal processing scenarios: 1) Multiple stationary heat source. We consider a problem of controlling grain growth in a copper thin film with a multi-zone micro-heater array, which can be integrated into a Scanning Electron Microscope for in-situ observation of microstructure. The control problem is to achieve grain growth consensus by adjusting the heater inputs. The process is cascade of a thermal evolution system and a grain growth system. Temperature control method for the heater array is first developed and verified in experiments. For grain growth control, we present the development and comparison of three control methods and verify their performance with high-fidelity simulation combining an Finite Element Method thermal evolution model and a biased Monte Carlo grain growth model. At last, image processing techniques are applied to characterize microstructure images and extract average grain size as real-time measurements. 2) Moving heat source. We study the distributed temperature and microstructure control problem during Laser Additive Manufacturing. The objective is to adjust the input variables, laser speed and power, to achieve the desired values of key process parameters, cooling rate and melt pool size that directly determines the resulted microstructure.

Based on a partial differential equation (PDE) model,we pose the control problem as a regulation problem in the (moving) laser frame, where the target temperature distribution is designed based on desired cooling rate and melt pool width. The control law is developed combining adaptive feedforward, passive error temperature field feedback and model parameter estimation, with performance demonstrated in simulation using a high-order approximation of the PDE model.