A novel fatigue life prediction approach of a GH4169 superalloy-welded joint based on a physics-informed machine learning method

Advanced manufacturing processes are often based on complex multiphysics phenomena that are either poorly understood or are computationally too expensive to simulate in the context of process design, control, or planning. Traditionally, simplified physics models with prescribed heuristics or purely data-driven surrogate models are used as alternatives in such applications. The concept of physics-informed machine learning (PIML) has been shown to have unique advantages over both of these alternatives in various fields of complex system analysis. In this paper, a new PIML approach is presented to model the geometry of the cut produced by a magnetically assisted laser-induced plasma micro-machining (M-LIPMM) process. This PIML architecture uses a neural network to auto-adapt the parametric boundary condition and physical properties used in a simplified finite difference-based physics model (of 2D heat conduction), as a function of the inputs namely the laser settings. This network also estimates the scaling and shifting parameters used by a convolutional neural network that takes the temperature profile predicted by the simplified heat conduction model to predict the width and depth of the machined cut. Trained on physical experiment data, the PIML approach compares favorably to a pure data-driven neural network model in extrapolation tests, while also providing physical insights (that the latter cannot). The PIML approach also provides an 85% better accuracy overall compared to the simplified physics model with heuristic settings.

Defects such as inclusions and void can be the origin of fatigue failure particularly in welding and casting materials. The fatigue life is influenced by the size, direction, shape and location of the defects. Therefore, many fatigue tests are necessary to obtain the fatigue properties. On the other hand, the prediction of fatigue life by representing characteristic variations of defects with probability distribution functions has been investigated by using several physical models and empirical formulae. However, most of the prediction methods of fatigue life arising from defects have not included the crack initiation. In the present study, the prediction was conducted by dividing the process into crack initiation and crack propagation. Voids, hard inclusions (Al2O3) and soft inclusions (MnS) were supposed as defects and two prediction models were proposed. Only the life of crack propagation was predicted by Paris law in one model (model A) while the life of crack initiation as well as propagation was predicted by Tanaka and Mura model in the other model (model B). The stress intensity factor using √area (projected square root area of defects) proposed by Murakami et al. was applied to Paris law in both models. The stress concentration and Taylor factor were applied to Tanaka and Mura model in the model B. In case of casting materials including voids, the fatigue life predicted by both models was within the range of the experimental scattering. Although the fatigue life predicted by model A was not consistent with the experimental results under high and low stress in case of high strength steel including MnS, the fatigue life predicted by model B mostly showed a good agreement with experimental results. Therefore, the present result suggested that the fatigue life prediction considering crack initiation showed higher precision than the prediction without crack initiation.

The safe positioning of particles within an acoustofluidic device is critical in biomedical and biological applications. Relating the design of acoustofluidic device walls and the internal acoustic field is a complex, nonlinear problem. The field of Physics-Informed Machine Learning (PIML) offers a number of potential approaches to simplify the design of these devices. One such PIML approach is learning from synthetic data. With large scientific data sets with rich spatial-temporal data and high-performance computing providing large amounts of data to be inferred and interpreted, the task of PIML is to ensure that these predictions and inferences are enforced by, and conform to the limits imposed by physical laws. The tools employed in PIML can include large, deep neural networks, Bayesian modeling, and deep reinforcement learning with sophisticated simulations of the environment. In this work, we show a simplified version of PIML using a combination of a small fully connected neural network and a 2D meshfree simulator of acoustic devices to predict the boundary shape for an acoustically actuated device. We will discuss the real-world results and applications, as well as the current limitations of this approach and the path ahead to scale and include more complexity for more applications and designs.

Predictive modeling of material removal rate in chemical mechanical planarization with physics-informed machine learning

Despite their effectiveness in modeling complex phenomena, the adoption of machine learning (ML) methods in computational mechanics has been hindered by the lack of availability of training datasets, limitations on the accuracy of out-of-sample predictions, and computational cost. This work presents a physics-informed ML approach and network architecture that addresses these challenges in the context of modeling the behavior of materials with damage. The proposed methodology is a novel physics-informed general convolutional network (PIGCN) framework that features (1) the fusion of a dense edge network with a convolutional neural network (CNN) for specifying and enforcing boundary conditions and geometry information, (2) a data augmentation approach for learning more information from a static dataset that significantly reduces the necessary data for training, and (3) the use of a CNN for physics-informed ML applications, which is not as well explored as graph networks in the current literature. The PIGCN framework is demonstrated for a simple two-dimensional, rectangular plate with a hole or elliptical defect in a linear-elastic material, but the approach is extensible to three dimensions and more complex problems. The results presented in this article show that the PIGCN framework improves physics-based loss convergence and predictive capability compared to ML-only (physics-uninformed) architectures. A key outcome of this research is the significant reduction in training data requirements compared to ML-only models, which could reduce a considerable hurdle to using data-driven models in materials engineering where material experimental data are often limited.

Physics-informed machine learning for enhanced prediction of condensation heat transfer

Predicting performance of in-situ microbial enhanced oil recovery process and screening of suitable microbe-nutrient combination from limited experimental data using physics informed machine learning approach

The heavy-duty gas turbine is playing an increasingly significant role on power generation due to its lower-emission, higher flexibility and thermo-efficiency. Main subsystems of the gas turbine like compressor, combustor and turbine degrade over the operating time under the harsh environmental conditions, which largely impacts the efficiency and productivity of the system. Therefore, it is critical to develop effective approaches to monitor performance degradation of a heavy-duty gas turbine for system predictive maintenance thus improving the efficiency and productivity of the machine. This paper presents a new physics informed machine learning methodology to predict the degradation of gas turbine by seamlessly integrating thermodynamic heat balancing mechanism, component characteristics, multi-source data and artificial neural network model. The mechanism-based thermodynamic model is established for multiple subsystems considering the balance of flow, mass and energy, and then integrated to a system level for performance simulation of the gas turbine under different conditions. The system model is able to effectively simulate values for those parameters that are not measurable (e.g. GT exhaust flow) or inaccurately measured (e.g. fuel flow). Machine learning based data cleaning approach is employed to preprocess the multivariate raw data of the gas turbine. The difference between design performance data and corrected value obtained from the physics-informed model under ISO conditions is utilized to assess the performance degradation. A Long Short-Term Memory (LSTM) model is established from the fusion of the actual and simulation data to predict the performance degradation of the gas turbine. A comparison study with the classical Nonlinear Autoregressive Network with External Input (NARX) neural network is conducted to demonstrate the advantage of the proposed method. Key Word: Gas Turbine, Thermodynamic Balance, Performance Degradation Predict, Machine Learning, LSTM

Summary This paper describes an investigation of the very low cycle fatigue strength of HT570 steel welded joints containing weld defects to determine the acceptable size of weld defects in girth welds of underground gas pipelines subjected to cyclic ground displacements due to earthquakes. Butt welded joints containing incomplete penetration (IP), blowholes (BH), lack of fusion in theintermedi‐ate pass (LF), and cracks in the penetration bead (CR) were prepared and tested under strain‐controlled conditions. All specimens tested in the present research study satisfy the fatigue design curve for girth welds of underground gas pipelines. The fatigue strength of specimens containing weld defects generally decreases with an increasing equivalent defect size. The shape of a crack initiated from a defect is affected by the reinforcement, with the surface crack propagating rapidly along the weld toe of the penetration weld over the width of the specimen. The relationship between the J‐integral range and the crack propagation rate under a very low cycle fatigue load is virtually the same as the extension of the relationship in the low J‐integral range. A crack propagation analysis based on the defect being regarded as a crack is performed to determine the relationship between the defect size and the number of cycles over the plate thickness. The critical size of defects in welds of gas pipelines under cyclic ground displacements is proposed by the analysis. The critical crack size of surface defects is applicable to buried defects, because surface defects give a more conservative evaluation than buried defects. The analytical results provide more conservative estimations than the experimental ones.

This study models the temperature evolution during additive friction stir deposition (AFSD) using machine learning. AFSD is a solid-state additive manufacturing technology that deposits metal using plastic flow without melting. However, the ability to predict its performance using the underlying physics is in the early stage. A physics-informed machine learning approach, AFSD-Nets, is presented here to predict temperature profiles based on the combined effects of heat generation and heat transfer. The proposed AFSD-Nets includes a set of customized neural network approximators, which are used to model the coupled temperature evolution for the tool and build during multi-layer material deposition. Experiments are designed and performed using 7075 aluminum feedstock deposited on a substrate of the same material for 30 layers. A comparison of predictions and measurements shows that the proposed AFSD-Nets approach can accurately describe and predict the temperature evolution during the AFSD process.

This research focuses on optimizing the airfoil shape tailored for cardboard drones, a significant asset in the Russia-Ukraine war due to their cost-effectiveness and operational impact. Despite their advantages in storage and transport, these drones require manual assembly, with considerable time spent on connecting wing ribs and spars. To enhance both ease of assembly and aerodynamic performance, we formulated an airfoil shape optimization problem incorporating manufacturability constraints specific to cardboard drone construction. We utilized NeuralFoil, a physics-informed machine learning approach, to overcome the limitations of traditional tools. By integrating NeuralFoil with a gradient-based optimization method, the proposed approach resulted in a new airfoil that not only improves aerodynamic efficiency over existing shapes but also simplifies the assembly process.

Predicting downhole temperature (DHT) in real time is crucial for safe and efficient operations in geothermal and high-pressure/high-temperature (HPHT) drilling. Although numerical models accurately capture DHT transients, their high computational cost limits their real-time use. Conversely, machine learning (ML) models offer speed but often lack embedded physical constraints, impairing generalization and interpretation. To address these limitations, this study proposes a hybrid, physics-informed machine learning approach that integrates mechanistic accuracy with ML efficiency for practical, interpretable DHT monitoring and forecasting. Physics-informed neural networks (PINNs) are used to integrate embedded equations derived from the one-dimensional conservation equations for mass, momentum, and energy, directly into the network's learning frame through an augmented loss function that supplements the standard data-driven loss function. This physics-guided approach leverages physical principles to enhance learning from limited data while simultaneously preserving model interpretability. The hybrid model is evaluated by comparing its outputs to numerical simulation results from a thermo-hydraulic model previously validated using field data from the Utah FORGE geothermal site, as reported in prior literature. Additionally, a parametric analysis was conducted to assess the model's predictive capability under various temperature management strategies reported in the literature. Modeling results show exceptional precision in capturing the transient behavior of DHT compared to the reference numerical models. The developed PINNs model leverages differentiable, physics-based constraints to achieve rapid convergence during the training and validation process, significantly reducing computational time relative to conventional numerical methods without compromising accuracy. In all scenarios, the model captured the complete profile of temperature over depth and time as well as the DHT with a mean absolute error (MAE) ~ 1℃ in training and testing, compared to the thermo-hydraulic results. In comparisons across multiple drilling case scenarios, PINNs achieved near-instantaneous simulation times and superior computational efficiency relative to the numerical model. Additionally, the model demonstrated reliable mud temperature profile predictions under varying conditions, including different pump rates, mud inlet temperatures, demonstrating strong generalization and potential for real-time geothermal well simulation. The developed hybrid model uniquely integrates mechanistic thermo-hydraulic modeling with a rapid ML approach to achieve real-time, precise downhole temperature estimations and forecasting. Consequently, it enables proactive temperature control by optimizing cooling strategies and mitigating downhole tool failures caused by elevated temperatures in geothermal and HPHT wells.

Solving a class of Thomas–Fermi equations: A new solution concept based on physics-informed machine learning

Critical physics-informed fatigue life prediction of laser 3D printed AlSi10Mg alloys with mass internal defects

The simulation of thin film semiconductor devices is challenging, partly due to the unknown material and device parameters. In this contribution, we present two different approaches to determine the missing material and device parameters from measurements. They both have in common that they are based on machine learning (ML) and numerical models. First, a numerical model describing the experiment is used to generate synthetic data to train a machine learning model the underlying material parameters. After successful training, a measurement is presented to the ML model to predict the parameters. In a more recent physics-informed ML approach, we integrate the model into the ML method and thus reduce the training data set.