Search for: All records

Elemental partitioning during thermal processing can significantly affect the corrosion resistance of bulk alloys operating in aggressive electrochemical environments, for which, despite decades of experimental and theoretical studies, the thermodynamic and electrochemical mechanisms still lack accurate quantitative descriptions. Here, we formulate an ab initio thermodynamic model to obtain the composition- and temperature-dependent free energies of formation (Δ_fG) for Ni–Cr alloys, a prototypical group of corrosion-resistant metals, and discover two equilibrium states that produce the driving forces for the elemental partitioning in Ni–Cr. The results are in quantitative agreement with the experimental studies on the thermodynamic stability of Ni–Cr. We further construct electrochemical (potential–pH) diagrams by obtaining the required Δ_fGvalues of native oxides and (oxy)hydroxides using high-fidelity ab-initio calculations that include exact electronic exchange and phononic contributions. We then analyze the passivation and electrochemical trends of Ni–Cr alloys, which closely explain various oxide-film growth and corrosion behaviors observed on alloy surfaces. We finally determine the optimal Cr content range of 14–34 at%, which provides the Ni–Cr alloys with both the preferred heat-treatment stability and superior corrosion resistance. We conclude by discussing the consequences of these findings on other Ni–Cr alloys with more complex additives, which can guide the further optimization of industrial Ni–Cr-based alloys.

 

Transfer learning on edge is challenging due to on-device limited resources. Existing work addresses this issue by training a subset of parameters or adding model patches. Developed with inference in mind, Inverted Residual Blocks (IRBs) split a convolutional layer into depthwise and pointwise convolutions, leading to more stacking layers, e.g., convolution, normalization, and activation layers. Though they are efficient for inference, IRBs require that additional activation maps are stored in memory for training weights for convolution layers and scales for normalization layers. As a result, their high memory cost prohibits training IRBs on resource-limited edge devices, and making them unsuitable in the context of transfer learning. To address this issue, we present MobileTL, a memory and computationally efficient on-device transfer learning method for models built with IRBs. MobileTL trains the shifts for internal normalization layers to avoid storing activation maps for the backward pass. Also, MobileTL approximates the backward computation of the activation layer (e.g., Hard-Swish and ReLU6) as a signed function which enables storing a binary mask instead of activation maps for the backward pass. MobileTL fine-tunes a few top blocks (close to output) rather than propagating the gradient through the whole network to reduce the computation cost. Our method reduces memory usage by 46% and 53% for MobileNetV2 and V3 IRBs, respectively. For MobileNetV3, we observe a 36% reduction in floating-point operations (FLOPs) when fine-tuning 5 blocks, while only incurring a 0.6% accuracy reduction on CIFAR10. Extensive experiments on multiple datasets demonstrate that our method is Pareto-optimal (best accuracy under given hardware constraints) compared to prior work in transfer learning for edge devices.