Research

From Limited Sensor Data to the Deformation Field in Metals: A Coupled Machine Learning, Physics Based Modeling and In Situ Experimental Approach

In the realm of plastic deformation in materials, the far-field boundary conditions, encompassing parameters such as strength and plastic strain, are essentially an amalgamation of contributions from myriad local sources dispersed throughout the material volume. When these local sources are activated, they induce abrupt localized changes in inelastic strain, thereby giving rise to the generation of high-frequency stress waves within the material. These stress waves, referred to as acoustic emissions (AE), can be effectively detected using surface sensors.

In the context of metals, AE measurements during deformation offer a unique, straightforward, and cost-effective in situ method for quantifying the statistical aspects of defect dynamics. However, the intricate interplay of AE between overlapping deformation mechanisms in metals, such as those between different dislocation mechanisms or the interactions between dislocation plasticity and fracture, remains a challenging frontier in AE studies. This is primarily due to the complexities associated with identifying and decoding the distinct AE signatures corresponding to each mechanism.

Consequently, our project is focused on the development of a coupled experimental setup that integrates in situ scanning electron microscopy micro-compression experiments with AE measurements. The objective is to precisely quantify the AE linked to dislocation slip and fracture at the microscale. In parallel, we are actively engaged in constructing a “digital twin” of AE by employing large-scale 3D discrete dislocation elastodynamic simulations. These simulations are designed to forecast the AE surface waves associated with various elementary dislocation mechanisms, as well as the AE signals arising from complex slip avalanches that occur during the deformation of single crystal Ni micropillars.

To enhance the depth of our insights, we have harnessed Machine Learning techniques, enabling us to predict the underlying deformation mechanisms from the AE surface measurements. This multifaceted approach will significantly advance our understanding of plastic deformation processes and contribute to the development of predictive models for materials behavior in response to external forces.

The interplay of local chemistry and plasticity in controlling microstructure formation and the mechanical properties of additively manufactured alloys. 

Additive manufacturing (AM) of metallic components holds significant promise over traditional metallurgical manufacturing methods due to its exceptional design flexibility spanning multiple length scales and precision, combined with a remarkable array of mechanical properties. However, a central challenge for this groundbreaking technology is to establish a clear link between microstructure and mechanical attributes.

A crucial preliminary step is to discern the influence of the processing path on the evolving microstructure. To address this, our project employs a comprehensive approach, melding experimental investigations, finite element analyses of the AM process, and large-scale three-dimensional discrete dislocation dynamics simulations. Our aim is to unravel the intricate mechanisms that underlie the creation of distinctive microstructures in additively manufactured metallic alloys. Concurrently, we explore how these microstructures impact the subsequent mechanical properties of these alloys.

Through these endeavors, we seek to advance the fundamental understanding of AM technology and its implications on material properties, paving the way for innovation and advancement in this field.