Search for: All records

Abstract In recent research, additions of solute to Ti and some Ti-based alloys have been employed to produce equiaxed microstructures when processing these materials using additive manufacturing. The present study develops a computational scheme for guiding the selection of such alloying additions, and the minimum amounts required, to effect the columnar to equiaxed microstructural transition. We put forward two physical mechanisms that may produce this transition; the first and more commonly discussed is based on growth restriction factors, and the second on the increased freezing range effected by the alloying addition coupled with the imposed rapid cooling rates associated with AM techniques. We show in the research described here, involving a number of model binary as well as complex multi-component Ti alloys, and the use of two different AM approaches, that the latter mechanism is more reliable regarding prediction of the grain morphology resulting from given solute additions. 

A key challenge in complex visuomotor control is learning abstract representations that are ef- fective for specifying goals, planning, and gen- eralization. To this end, we introduce universal planning networks (UPN). UPNs embed differen- tiable planning within a goal-directed policy. This planning computation unrolls a forward model in a latent space and infers an optimal action plan through gradient descent trajectory optimization. The plan-by-gradient-descent process and its un- derlying representations are learned end-to-end to directly optimize a supervised imitation learning objective. We find that the representations learned are not only effective for goal-directed visual imi- tation via gradient-based trajectory optimization, but can also provide a metric for specifying goals using images. The learned representations can be leveraged to specify distance-based rewards to reach new target states for model-free reinforce- ment learning, resulting in substantially more ef- fective learning when solving new tasks described via image-based goals. We were able to achieve successful transfer of visuomotor planning strate- gies across robots with significantly different mor- phologies and actuation capabilities.