In this paper, we implement a vision graph neural network (ViG) architecture to segment microstructures in X-ray computed tomography 3D data. Our ViG architecture is first trained on a synthetic augmented dataset, and then fine-tuned on experimental data to obtain an improved segmentation. Successively, we assess the accuracy of the segmentation on manually-labeled experimental slices. We exemplarily use the approach on a complex microstructure: a metal matrix composite, reinforced with two ceramic phases, intermetallic inclusions and a silicon network, in order to show the generality of our method. ViG model proves to be more efficient than U-Nets in adapting to new data when fine-tuned on a small portion of the experimental data. The fine-tuned ViG shows comparable performance to U-Nets, while largely reducing the number of trainable parameters, with the potential of greater adaptability and efficiency.
This study addresses biochar as a potential carbon-sequestering filler in cement and examines its effect on mechanical properties using X-ray computed tomography (XCT) and digital volume correlation (DVC). DVC was reliably used to measure global displacement and has proven to be an effective method for correcting displacement data obtained from mechanical tests conducted without traditional instrumentation, such as extensometer. This made it possible to measure strain and Young's modulus accurately. The results demonstrate that while 5 vol% biochar replacement had minimal effect on mechanical properties, a 25 vol% biochar replacement caused a 35 % reduction in Young's modulus and 40 % reduction in the ultimate compressive strength. Additionally, DVC detected strain concentrations and predicted material failure locations even when cracks could not be quantified using XCT alone. Moreover, the study reveals that biochar particles, due to their sharp geometry, increase internal shear strain during uniaxial compression, unlike round phases such as pores.
In the present work, we employ the concept of neural network temperature to prune unimportant features in input to a Graph Neural Network (GNN) architecture. In benchmark datasets for node and graph property prediction, each node comes equipped with a vector of numerous features. It is paramount to understand which information is actually necessary and which can be discarded, both for efficiency and explainability. The temperature is linked to the gradient activity due to the loss function minimization and leads to pruning of weight structures associated with small gradients. This study is done on different GNN architectures, one for node classification and another one for link prediction, and several benchmark datasets are employed. We compare the results with similar experiments previously conducted on the filters of Convolutional Neural Networks. Although still at the proof-of-concept stage, our temperature-based pruning technique stands as a promising alternative to state-of-the-art magnitude-based pruning techniques.
Symmetric organic flow batteries (SOFBs) can potentially address membrane crossover problems by employing bipolar redox-active organic molecules (BROMs). Herein, a triarylamine (TAA) skeleton was chosen as a posolyte moiety for a new class of bipolar molecules for pH-neutral aqueous flow batteries (FBs). Pyridinium and viologen derivatives were tethered to the posolyte moiety, and the new compounds were characterized. Cyclic voltammetry revealed that only viologen with a highly hydrophilic substituent, connected to the TAA moiety via a Zincke reaction, could be reversibly reduced. Varying the supporting electrolyte concentration on the selected derivative revealed water solubility as a challenge for further development. The selected derivative, MeO-TPA-Vi-DMAE, was subjected to hydrodynamic voltammetry, and a modified Koutecký-Levich analysis was developed to investigate the observed potential-dependent currents at the hydrodynamically dominated region, which are often seen with redox-active organic molecules. This model discarded a purely Ohmic effect, showing a useful Levich slope at a certain overpotential before the onset of a secondary reaction. TAA-based BROMs hold promise for pH-neutral aqueous SOFBs, and the results will guide the design of new derivatives. The three-term Koutecký-Levich relation here introduced will be useful not only to develop BROM-based FBs but will most likely appeal to a much broader audience.
Extended X-ray absorption fine structure (EXAFS) conducted on an equiatomic MoNbTaW bcc medium-entropy alloy that was annealed at 2273 K reveals unexpectedly small 1st and 2nd shell element-specific lattice distortions. An experimental size-mismatch parameter, delta(exp), is determined to be ca. 50% lower than the corresponding calculated value. Around W, short-range order (SRO) preferring 4d elements in the 1(st) and 2(nd) shells persists. A Nb-W ordering is found, which is reminiscent of ordering emerging at lower temperatures in the B2(Mo,W;Ta,Nb)- and B32(Nb,W)-phases. With high-temperature ordering preferences in fcc also foreshadowing low-temperature phase, these findings suggest a general feature of high-temperature SRO.
