Search for: All records

As industries transition into the Industry 4.0 paradigm, the relevance and interest in concepts like Digital Twin (DT) are at an all-time high. DTs offer direct avenues for industries to make more accurate predictions, rational decisions, and informed plans, ultimately reducing costs, increasing performance and productivity. Adequate operation of DTs in the context of smart manufacturing relies on an evolving data-set relating to the real-life object or process, and a means of dynamically updating the computational model to better conform to the data. This reliance on data is made more explicit when physics-based computational models are not available or difficult to obtain in practice, as it's the case in most modern manufacturing scenarios. For data-based model surrogates to adequately represent the underlying physics, the number of training data points must keep pace with the number of degrees of freedom in the model, which can be on the order of thousands. However, in niche industrial scenarios like the one in manufacturing applications, the availability of data is limited (on the order of a few hundred data points, at best), mainly because a manual measuring process typically must take place for a few of the relevant quantities, e.g., level of wear of a tool. In other words, notwithstanding the popular notion of big-data, there is still a stark shortage of ground-truth data when examining, for instance, a complex system's path to failure. In this work we present a framework to alleviate this problem via modern machine learning tools, where we show a robust, efficient and reliable pathway to augment the available data to train the data-based computational models. Small sample size data is a key limitation in performance in machine learning, in particular with very high dimensional data. Current efforts for synthetic data generation typically involve either Generative Adversarial Networks (GANs) or Variational AutoEncoders (VAEs). These, however, are are tightly related to image processing and synthesis, and are generally not suited for sensor data generation, which is the type of data that manufacturing applications produce. Additionally, GAN models are susceptible to mode collapse, training instability, and high computational costs when used for high dimensional data creation. Alternatively, the encoding of VAEs greatly reduces dimensional complexity of data and can effectively regularize the latent space, but often produces poor representational synthetic samples. Our proposed method thus incorporates the learned latent space from an AutoEncoder (AE) architecture into the training of the generation network in a GAN. The advantages of such scheme are twofold: \textbf{(\textit{i})} the latent space representation created by the AE reduces the complexity of the distribution the generator must learn, allowing for quicker discriminator convergence, and \textbf{(\textit{ii})} the structure in the sensor data is better captured in the transition from the original space to the latent space. Through time statistics (up to the fifth moment), ARIMA coefficients and Fourier series coefficients, we compare the synthetic data from our proposed AE+GAN model with the original sensor data. We also show that the performance of our proposed method is at least comparable with that of the Riemannian Hamiltonian VAE, which is a recently published data augmentation framework specifically designed to handle very small high dimensional data sets. 

The elusive PcFe(DABCO)₂(Pc = phthalocyaninato(2-) ligand; DABCO = 1,4-diazabicyclo[2.2.2]octane) complex was prepared and characterized by UV-Vis, MCD,¹H NMR, and Mössbauer spectroscopies. The X-ray crystal structure of this complex indicates the longest Fe-N(DABCO) bond distance among all known PcFeL₂complexes with nitrogen donors as the axial ligands. The target compound is only stable in the presence of large access of the axial ligand and rapidly converts into the (PcFe)₂O [Formula: see text]-oxo dimer even at a modest temperature. The electronic structure of the PcFe(DABCO)₂complex was elucidated by DFT and TDDFT methods. The DFT calculations predicted a very small singlet-triplet gap in this compound. The femtosecond transient absorption spectroscopy is indicative of extremely fast ([Formula: see text]200 fs) deactivation of the first excited state in PcFe(DABCO)₂with a lack of formation of the long-lived low-energy triplet state. 

Abstract Earthquakes result from fast slip that occurs along a fault surface. Interestingly, numerous dense geodetic observations over the last two decades indicate that such dynamic slip may start by a gradual unlocking of the fault surface and related progressive slip acceleration. This first slow stage is of great interest, because it could define an early indicator of a devastating earthquake. However, not all slow slip turns into fast slip, and sometimes it may simply stop. In this study, we use a numerical model based on the discrete element method to simulate crustal strike‐slip faults of 50 km length that generate a wide variety of slip‐modes, from stable‐slip, to slow earthquakes, to fast earthquakes, all of which show similar characteristics to natural cases. The main goal of this work is to understand the conditions that allow slow events to turn into earthquakes, in contrast to those that cause slow earthquakes to stop. Our results suggest that fault surface geometry and related dilation/contraction patterns along strike play a key role. Slow earthquakes that initiate in large dilated regions bounded by neutral or low contracted domains, might turn into earthquakes. Slow events occurring in regions dominated by closely spaced, alternating, small magnitude dilational and contractional zones tend not to accelerate and may simply stop as isolated slow earthquakes. 

Abstract We propose a binary classification model rooted in state‐of‐the‐art deep learning techniques to predict whether or not complete‐interface rupture is imminent along a numerical megathrust fault. The models are trained on labeled 2D space‐time input features taken from the synthetic fault system. We contrast the performance of two neural networks trained on three types of data, to determine the relative predictive power of each. The neural networks are able to discriminate imminent complete rupture precursors from everything else, thus providing a relative size and time forecast. Vertical displacements along the fault demonstrate relatively good predictive power. The results confirm previous qualitative observations that precursory deformation scales with upcoming event size, consistent with the preslip model for earthquake nucleation. The methods we propose are adaptable and can be modified to use 3D data in the future. 

The electronic communication between two ferrocene groups in the electron-deficient expanded aza-BODIPY analogue of zinc manitoba-dipyrromethene (MB-DIPY) was probed by spectroscopic, electrochemical, spectroelectrochemical, and theoretical methods. The excited-state dynamics involved sub- ps formation of the charge-separated state in the organometallic zinc MB-DIPYs, followed by recovery of the ground state via charge recombination in 12 ps. The excited-state behavior was contrasted with that observed in the parent complex that lacked the ferrocene electron donors and has a much longer excited-state lifetime (670 ps for the singlet state). Much longer decay times observed for the parent complex without ferrocene confirm that the main quenching mechanism in the ferrocene-containing 4 is reflective of the ultrafast ferrocene-to-MB-DIPY core charge transfer (CT