Physics-Informed Machine Learning for Accurate Prediction of Temperature and Melt Pool Dimension in Metal Additive Manufacturing.

The temperature distribution and melt pool size have a great influence on the microstructure and mechanical behavior of metal additive manufacturing process. The numerical method can give relatively accurate results but is time-consuming and, therefore, unsuitable for in-process prediction. Owing to its remarkable capabilities, machine learning methods have been applied to predict melt pool size and temperature distribution. However, the success of traditional data-driven machine learning methods is highly dependent on the amount and quality of the training data, which is not always convenient to access. This article proposes a physics-informed machine learning (PIML) method, which integrates data and physics laws in the training parts, overcoming the problems of low speed and data availability. An artificial neural network constrained by the heat transfer equation and a small amount of labeled data is developed to predict the melt pool size and temperature distribution. Besides, the locally adaptive activation function is utilized to improve the prediction performance. The result shows that the developed PIML model can accurately predict the temperature and melt pool dimension under different scanning speeds with a small amount of labeled data, which shows significant potential in practical application.

High-resolution lensless holographic microscopy using a physics-aware deep network.

Significance: Lensless digital inline holographic microscopy (LDIHM) is an emerging quantitative phase imaging modality that uses advanced computational methods for phase retrieval from the interference pattern. The existing end-to-end deep networks require a large training dataset with sufficient diversity to achieve high-fidelity hologram reconstruction. To mitigate this data requirement problem, physics-aware deep networks integrate the physics of holography in the loss function to reconstruct complex objects without needing prior training. However, the data fidelity term measures the data consistency with a single low-resolution hologram without any external regularization, which results in a low performance on complex biological data. Aim: We aim to mitigate the challenges with trained and physics-aware untrained deep networks separately and combine the benefits of both methods for high-resolution phase recovery from a single low-resolution hologram in LDIHM. Approach: We propose a hybrid deep framework (HDPhysNet) using a plug-and-play method that blends the benefits of trained and untrained deep models for phase recovery in LDIHM. The high-resolution phase is generated by a pre-trained high-definition generative adversarial network (HDGAN) from a single low-resolution hologram. The generated phase is then plugged into the loss function of a physics-aware untrained deep network to regulate the complex object reconstruction process. Results: Simulation results show that the SSIM of the proposed method is increased by 0.07 over the trained and 0.04 over the untrained deep networks. The average phase-SNR is elevated by 8.2 dB over trained deep models and 9.8 dB over untrained deep networks on the experimental biological cells (cervical cells and red blood cells). Conclusions: We showed improved performance of the HDPhysNet against the unknown perturbation in the imaging parameters such as the propagation distance, the wavelength of the illuminating source, and the imaging sample compared with the trained network (HDGAN). LDIHM, combined with HDPhysNet, is a portable and technology-driven microscopy best suited for point-of-care cytology applications.

Data-driven continuum damage mechanics with built-in physics.

Soft materials such as rubbers and soft tissues often undergo large deformations and experience damage degradation that impairs their function. This energy dissipation mechanism can be described in a thermodynamically consistent framework known as continuum damage mechanics. Recently, data-driven methods have been developed to capture complex material behaviors with unmatched accuracy due to the high flexibility of deep learning architectures. Initial efforts focused on hyperelastic materials, and recent advances now offer the ability to satisfy physics constraints such as polyconvexity of the strain energy density function by default. However, modeling inelastic behavior with deep learning architectures and built-in physics has remained challenging. Here we show that neural ordinary differential equations (NODEs), which we used previously to model arbitrary hyperelastic materials with automatic polyconvexity, can be extended to model energy dissipation in a thermodynamically consistent way by introducing an inelastic potential: a monotonic yield function. We demonstrate the inherent flexibility of our network architecture in terms of different damage models proposed in the literature. Our results suggest that our NODEs re-discover the true damage function from synthetic stress-deformation history data. In addition, they can accurately characterize experimental skin and subcutaneous tissue data.

Machine learning-guided reconstruction of cytoskeleton network from live-cell AFM images.

How actin filaments (F-actins) are dynamically reorganized in motile cells at the level of individual filaments is an open question. To find the answer, a high-speed atomic force microscopy (HS-AFM) system has been developed to live-imagine intracellular dynamics of the individual F-actins. However, noise and low resolution made it difficult to fully recognize individual F-actins in the HS-AFM images. To tackle this problem, we developed a new machine learning method that quantitatively recognizes individual F-actins. The method estimates F-actin orientation from the image while improving the resolution. We found that F-actins were oriented at ±35° toward the membrane in the lamellipodia, which is consistent with Arp2/3 complex-induced branching. Furthermore, in the cell cortex our results showed non-random orientation at four specific angles, suggesting a new mechanism for F-actin organization demonstrating the potential of our newly developed method to fundamentally improve our understanding of the structural dynamics of F-actin networks.

Regulating oxygen vacancies in ammonium vanadate electrode materials for advanced aqueous zinc ion batteries.

In the past decade, vanadates have attracted one's attention as the electrode materials for aqueous zinc ion batteries (AZIBs). Nevertheless, their structural instability and sluggish ion/electron dynamics lead to an inevitable decline in the electrochemical performance. To address these issues, we introduce oxygen vacancies into NH4V4O10 nanosheets to improve the ion transport rate during the electrochemical reaction. The prepared NHVO-40 samples provide many active sites compared to NH4V4O10 materials. The assembled cell delivers a capacity of 452.03 mAh g-1 at a current density of 0.2 A g-1. It also presents a retention rate of 94.6% at 10 A g-1 after 4000 times cycling. In addition, they still possess an energy density of 332.5 Wh kg-1 at a power density of 70 W kg-1.

Physics-Informed Machine Learning with Data-Driven Equations for Predicting Organic Solar Cell Performance.

Organic solar cells (OSCs) have emerged as a promising solution in pursuing sustainable energy. This study presents a comprehensive approach to advancing OSC development by integrating data-driven equations from quantum mechanical (QM) descriptors with physics-informed machine learning (PIML) models. We circumvent traditional experimental limitations through high-throughput QM calculations, prioritizing transparent and interpretable models. Using the SISSO++ method, we identified key descriptors that effectively map the relationships between input variables and photovoltaic performance metrics. Our innovative predictive models, derived from SISSO outputs, excel in forecasting critical OSC parameters such as short-circuit current (JSC), open-circuit voltage (VOC), fill factor (FF), and power conversion efficiency (PCEmax), achieving high accuracy even with limited data sets. To validate our models' practical utility, we applied the PIML framework to a newly compiled data set of OSC devices, demonstrating their versatility and capability in pinpointing high-performance materials. This research underscores the strong predictive power of our models, bridging the gap between experimental results and theoretical predictions and making significant contributions to the advancement of sustainable energy technologies.

Spectral integrated neural networks (SINNs) for solving forward and inverse dynamic problems.

This study introduces an innovative neural network framework named spectral integrated neural networks (SINNs) to address both forward and inverse dynamic problems in three-dimensional space. In the SINNs, the spectral integration technique is utilized for temporal discretization, followed by the application of a fully connected neural network to solve the resulting partial differential equations in the spatial domain. Furthermore, the polynomial basis functions are employed to expand the unknown function, with the goal of improving the performance of SINNs in tackling inverse problems. The performance of the developed framework is evaluated through several dynamic benchmark examples encompassing linear and nonlinear heat conduction problems, linear and nonlinear wave propagation problems, inverse problem of heat conduction, and long-time heat conduction problem. The numerical results demonstrate that the SINNs can effectively and accurately solve forward and inverse problems involving heat conduction and wave propagation. Additionally, the SINNs provide precise and stable solutions for dynamic problems with extended time durations. Compared to commonly used physics-informed neural networks, the SINNs exhibit superior performance with enhanced convergence speed, computational accuracy, and efficiency.

A Hamilton principle-based model for diffusion-driven biofilm growth.

Dense communities of bacteria, also known as biofilms, are ubiquitous in all of our everyday life. They are not only always surrounding us, but are also active inside our bodies, for example in the oral cavity. While some biofilms are beneficial or even necessary for human life, others can be harmful. Therefore, it is highly important to gain an in-depth understanding of biofilms which can be achieved by in vitro or in vivo experiments. Since these experiments are often time-consuming or expensive, in silico models have proven themselves to be a viable tool in assisting the description and analysis of these complicated processes. Current biofilm growth simulations are using mainly two approaches for describing the underlying models. The volumetric approach splits the deformation tensor into a growth and an elastic part. In this approach, the mass never changes, unless some additional constraints are enforced. The density-based approach, on the other hand, uses an evolution equation to update the growing tissue by adding mass. Here, the density stays constant, and no pressure is exerted. The in silico model presented in this work combines the two approaches. Thus, it is possible to capture stresses inside of the biofilm while adding mass. Since this approach is directly derived from Hamilton's principle, it fulfills the first and second law of thermodynamics automatically, which other models need to be checked for separately. In this work, we show the derivation of the model as well as some selected numerical experiments. The numerical experiments show a good phenomenological agreement with what is to be expected from a growing biofilm. The numerical behavior is stable, and we are thus capable of solving complicated boundary value problems. In addition, the model is very reactive to different input parameters, thereby different behavior of various biofilms can be captured without modifying the model.

Physics-Informed Graph Neural Networks to solve 1-D equations of blood flow.

BACKGROUND AND OBJECTIVE: Computational models of hemodynamics can contribute to optimizing surgical plans, and improve our understanding of cardiovascular diseases. Recently, machine learning methods have become essential to reduce the computational cost of these models. In this study, we propose a method that integrates 1-D blood flow equations with Physics-Informed Graph Neural Networks (PIGNNs) to estimate the propagation of blood flow velocity and lumen area pulse waves along arteries. METHODS: Our methodology involves the creation of a graph based on arterial topology, where each 1-D line represents edges and nodes in the blood flow analysis. The innovation lies in decoding the mathematical data connecting the nodes, where each node has velocity and lumen area pulse waveform outputs. The training protocol for PIGNNs involves measurement data, specifically velocity waves measured from inlet and outlet vessels and diastolic lumen area measurements from each vessel. To optimize the learning process, our approach incorporates fundamental physical principles directly into the loss function. This comprehensive training strategy not only harnesses the power of machine learning but also ensures that PIGNNs respect fundamental laws governing fluid dynamics. RESULTS: The accuracy was validated in silico with different arterial networks, where PIGNNs achieved a coefficient of determination (R2) consistently above 0.99, comparable to numerical methods like the discontinuous Galerkin scheme. Moreover, with in vivo data, the prediction reached R2 values greater than 0.80, demonstrating the method's effectiveness in predicting flow and lumen dynamics using minimal data. CONCLUSIONS: This study showcased the ability to calculate lumen area and blood flow rate in blood vessels within a given topology by seamlessly integrating 1-D blood flow with PIGNNs, using only blood flow velocity measurements. Moreover, this study is the first to compare the PIGNNs method with other classic Physics-Informed Neural Network (PINNs) approaches for blood flow simulation. Our findings highlight the potential to use this cost-effective and proficient tool to estimate real-time arterial pulse waves.

Ultrasound Physics.

With the growing use of point-of-care ultrasound (POCUS) in various clinical settings, it is essential for users of ultrasound to have a thorough understanding of the basics of ultrasound physics, including sound wave properties, its interaction with various tissues, common artifacts, and knobology. The authors introduce and discuss these concepts in this article, with a focus on clinical implications.

Molecular nano-I-beam class of materials: options based on configuration, first principles-based optimization and properties.

Nanotubes showed merits including high structural strength-to-weight ratio. However, tubes are less favored regarding stiffness and strength. Nano-I-beams are proposed for improved nano-mechanics. Computationally, the study proposes novel molecular designs of I-beam-like shaped structures. A conformation analysis, molecular dynamics and first principles-based optimization are presented. The study proposes options based on the configuration of the molecular nano-I-beam structure providing less number of planes of symmetry and hence more stability than nanotube-like structures. These designs feature a unique geometrical differentiator of having the walls of the out-of-plane hexagonal motif-based molecular nano-I-beam (C60H46) inclined with different inclination angles enabling promising properties. The stability of the proposed nano-I-beam is proved on par with the corresponding nanotube-like structure. First principles-based evidence is provided on the comparable polarizability and the comparable ability to store energy of the supercell of the crystalline slab nano-I-beam in comparison with the corresponding nanotube. A proposed hybrid octa-hexagonal-cubic molecular nano-I-beam (C24H12) remedies the nano-buckling observed in the alike square-octagonal nanostructure. The molecular nano-I-beam exhibits intrinsic switchability that enables the nano-I-beam to be a topological semiconductor/insulator. The results show promising electronic and elastic properties of the proposed nano-I-beams that suit several applications such as their use in capacitors, transistors, insulators, batteries, quantization-based nano-devices, solid lubricant additive to grease, toughening fibers of nanocomposites, hydrophobic films, emissions adsorbents, catalytic sensors, PAH materials for space, and sustainable energy. The molecular nano-I-beam provides the base of the corresponding 2-D crystalline slab nano-I-beams.

Electrical Quantum Coupling of Subsurface-Nanolayer Quasipolarons.

We perform dielectric and impedance spectrums on the compressively-strained ceramics of multiferroic bismuth ferrite. The subsurface-nanolayer quasipolarons manifest the step-like characteristic of pressure-dependent transient frequency and, furthermore, pressure-dependency fails in the transformation between complex permittivity and electrical impedance, which is well-known in classic dielectric physics, as well as the bulk dipole chain at the end of the dissipation peak.

Deep learning for rapid analysis of cell divisions in vivo during epithelial morphogenesis and repair.

Cell division is fundamental to all healthy tissue growth, as well as being rate-limiting in the tissue repair response to wounding and during cancer progression. However, the role that cell divisions play in tissue growth is a collective one, requiring the integration of many individual cell division events. It is particularly difficult to accurately detect and quantify multiple features of large numbers of cell divisions (including their spatio-temporal synchronicity and orientation) over extended periods of time. It would thus be advantageous to perform such analyses in an automated fashion, which can naturally be enabled using deep learning. Hence, we develop a pipeline of deep learning models that accurately identify dividing cells in time-lapse movies of epithelial tissues in vivo. Our pipeline also determines their axis of division orientation, as well as their shape changes before and after division. This strategy enables us to analyse the dynamic profile of cell divisions within the Drosophila pupal wing epithelium, both as it undergoes developmental morphogenesis and as it repairs following laser wounding. We show that the division axis is biased according to lines of tissue tension and that wounding triggers a synchronised (but not oriented) burst of cell divisions back from the leading edge.