A liquefied natural gas (LNG) facility often incorporates replicate liquefaction trains. The performance of equivalent units across trains, designed using common numerical models, might be expected to be similar. In this article, we discuss statistical analysis of real plant data to validate this assumption. Analysis of operational data for end flash vessels from a pair of replicate trains at an LNG facility indicates that one train produces 2.8%–6.4% more end flash gas than the other. We then develop statistical models for train operation, facilitating reduced flaring and hence a reduction of up to 45% in CO2 equivalent flaring emissions, noting that flaring emissions for a typical LNG facility account for ~4%–8% of the overall facility emissions. We recommend that operational data-driven models be considered generally to improve the performance of LNG facilities and reduce their CO2 footprint, particularly when replica units are present.

Surrogate models of turbulent diffusive flames could play a strategic role in the design of liquid rocket engine combustion chambers. The present article introduces a method to obtain data-driven surrogate models for coaxial injectors, by leveraging an inductive transfer learning strategy over a U-Net with available multifidelity Large Eddy Simulations (LES) data. The resulting models preserve reasonable accuracy while reducing the offline computational cost of data-generation. First, a database of about 100 low-fidelity LES simulations of shear-coaxial injectors, operating with gaseous oxygen and gaseous methane as propellants, has been created. The design of experiments explores three variables: the chamber radius, the recess-length of the oxidizer post, and the mixture ratio. Subsequently, U-Nets were trained upon this dataset to provide reasonable approximations of the temporal-averaged two-dimensional flow field. Despite the fact that neural networks are efficient non-linear data emulators, in purely data-driven approaches their quality is directly impacted by the precision of the data they are trained upon. Thus, a high-fidelity (HF) dataset has been created, made of about 10 simulations, to a much greater cost per sample. The amalgamation of low and HF data during the the transfer-learning process enables the improvement of the surrogate model’s fidelity without excessive additional cost.

This study introduces an advanced reinforcement learning (RL)-based control strategy for heating, ventilation, and air conditioning (HVAC) systems, employing a soft actor-critic agent with a customized reward mechanism. This strategy integrates time-varying outdoor temperature-dependent weighting factors to dynamically balance thermal comfort and energy efficiency. Our methodology has undergone rigorous evaluation across two distinct test cases within the building optimization testing (BOPTEST) framework, an open-source virtual simulator equipped with standardized key performance indicators (KPIs) for performance assessment. Each test case is strategically selected to represent distinct building typologies, climatic conditions, and HVAC system complexities, ensuring a thorough evaluation of our method across diverse settings. The first test case is a heating-focused scenario in a residential setting. Here, we directly compare our method against four advanced control strategies: an optimized rule-based controller inherently provided by BOPTEST, two sophisticated RL-based strategies leveraging BOPTEST’s KPIs as reward references, and a model predictive control (MPC)-based approach specifically tailored for the test case. Our results indicate that our approach outperforms the rule-based and other RL-based strategies and achieves outcomes comparable to the MPC-based controller. The second scenario, a cooling-dominated environment in an office setting, further validates the versatility of our strategy under varying conditions. The consistent performance of our strategy across both scenarios underscores its potential as a robust tool for smart building management, adaptable to both residential and office environments under different climatic challenges.

We introduce a comprehensive data-driven framework aimed at enhancing the modeling of physical systems, employing inference techniques and machine-learning enhancements. As a demonstrative application, we pursue the modeling of cathodic electrophoretic deposition, commonly known as e-coating. Our approach illustrates a systematic procedure for enhancing physical models by identifying their limitations through inference on experimental data and introducing adaptable model enhancements to address these shortcomings. We begin by tackling the issue of model parameter identifiability, which reveals aspects of the model that require improvement. To address generalizability, we introduce modifications, which also enhance identifiability. However, these modifications do not fully capture essential experimental behaviors. To overcome this limitation, we incorporate interpretable yet flexible augmentations into the baseline model. These augmentations are parameterized by simple fully-connected neural networks, and we leverage machine-learning tools, particularly neural ordinary differential equations, to learn these augmentations. Our simulations demonstrate that the machine-learning-augmented model more accurately captures observed behaviors and improves predictive accuracy. Nevertheless, we contend that while the model updates offer superior performance and capture the relevant physics, we can reduce off-line computational costs by eliminating certain dynamics without compromising accuracy or interpretability in downstream predictions of quantities of interest, particularly film thickness predictions. The entire process outlined here provides a structured approach to leverage data-driven methods by helping us comprehend the root causes of model inaccuracies and by offering a principled method for enhancing model performance.

Numerical solutions of partial differential equations require expensive simulations, limiting their application in design optimization, model-based control, and large-scale inverse problems. Surrogate modeling techniques aim to decrease computational expense while retaining dominant solution features and characteristics. Existing frameworks based on convolutional neural networks and snapshot-matrix decomposition often rely on lossy pixelization and data-preprocessing, limiting their effectiveness in realistic engineering scenarios. Recently, coordinate-based multilayer perceptron networks have been found to be effective at representing 3D objects and scenes by regressing volumetric implicit fields. These concepts are leveraged and adapted in the context of physical-field surrogate modeling. Two methods toward generalization are proposed and compared: design-variable multilayer perceptron (DV-MLP) and design-variable hypernetworks (DVH). Each method utilizes a main network which consumes pointwise spatial information to provide a continuous representation of the solution field, allowing discretization independence and a decoupling of solution and model size. DV-MLP achieves generalization through the use of a design-variable embedding vector, while DVH conditions the main network weights on the design variables using a hypernetwork. The methods are applied to predict steady-state solutions around complex, parametrically defined geometries on non-parametrically-defined meshes, with model predictions obtained in less than a second. The incorporation of random Fourier features greatly enhanced prediction and generalization accuracy for both approaches. DVH models have more trainable weights than a similar DV-MLP model, but an efficient batch-by-case training method allows DVH to be trained in a similar amount of time as DV-MLP. A vehicle aerodynamics test problem is chosen to assess the method’s feasibility. Both methods exhibit promising potential as viable options for surrogate modeling, being able to process snapshots of data that correspond to different mesh topologies.

Deep learning (DL) has become the most effective machine learning solution for addressing and accelerating complex problems in various fields, from computer vision and natural language processing to many more. Training well-generalized DL models requires large amounts of data which allows the model to learn the complexity of the task it is being trained to perform. Consequently, performance optimization of the deep-learning models is concentrated on complex architectures with a large number of tunable model parameters, in other words, model-centric techniques. To enable training such large models, significant effort has also gone into high-performance computing and big-data handling. However, adapting DL to tackle specialized domain-related data and problems in real-world settings presents unique challenges that model-centric techniques do not suffice to optimize. In this paper, we tackle the problem of developing DL models for seismic imaging using complex seismic data. We specifically address developing and deploying DL models for salt interpretation using seismic images. Most importantly, we discuss how looking beyond model-centric and leveraging data-centric strategies for optimization of DL model performance was crucial to significantly improve salt interpretation. This technique was also key in developing production quality, robust and generalized models.

Vibration-based structural health monitoring (SHM) of (large) infrastructure through operational modal analysis (OMA) is a commonly adopted strategy. This is typically a four-step process, comprising estimation, tracking, data normalization, and decision-making. These steps are essential to ensure structural modes are correctly identified, and results are normalized for environmental and operational variability (EOV). Other challenges, such as nonstructural modes in the OMA, for example, rotor harmonics in (offshore) wind turbines (OWTs), further complicate the process. Typically, these four steps are considered independently, making the method simple and robust, but rather limited in challenging applications, such as OWTs. Therefore, this study aims to combine tracking, data normalization, and decision-making through a single machine learning (ML) model. The presented SHM framework starts by identifying a “healthy” training dataset, representative of all relevant EOV, for all structural modes. Subsequently, operational and weather data are used for feature selection and a comparative analysis of ML models, leading to the selection of tree-based learners for natural frequency prediction. Uncertainty quantification (UQ) is introduced to identify out-of-distribution instances, crucial to guarantee low modeling error and ensure only high-fidelity structural modes are tracked. This study uses virtual ensembles for UQ through the variance between multiple truncated submodel predictions. Practical application to monopile-supported OWT data demonstrates the tracking abilities, separating structural modes from rotor dynamics. Control charts show improved decision-making compared to traditional reference-based methods. A synthetic dataset further confirms the approach’s robustness in identifying relevant natural frequency shifts. This study presents a comprehensive data-driven approach for vibration-based SHM.

Applications of deep learning to physical simulations such as Computational Fluid Dynamics have recently experienced a surge in interest, and their viability has been demonstrated in different domains. However, due to the highly complex, turbulent, and three-dimensional flows, they have not yet been proven usable for turbomachinery applications. Multistage axial compressors for gas turbine applications represent a remarkably challenging case, due to the high-dimensionality of the regression of the flow field from geometrical and operational variables. This paper demonstrates the development and application of a deep learning framework for predictions of the flow field and aerodynamic performance of multistage axial compressors. A physics-based dimensionality reduction approach unlocks the potential for flow-field predictions, as it re-formulates the regression problem from an unstructured to a structured one, as well as reducing the number of degrees of freedom. Compared to traditional “black-box” surrogate models, it provides explainability to the predictions of the overall performance by identifying the corresponding aerodynamic drivers. The model is applied to manufacturing and build variations, as the associated performance scatter is known to have a significant impact on \mathrm{C}{\mathrm{O}}_2 $\mathrm{C}{\mathrm{O}}_2$ emissions, which poses a challenge of great industrial and environmental relevance. The proposed architecture is proven to achieve an accuracy comparable to that of the CFD benchmark, in real-time, for an industrially relevant application. The deployed model is readily integrated within the manufacturing and build process of gas turbines, thus providing the opportunity to analytically assess the impact on performance with actionable and explainable data.

This article establishes a data-driven modeling framework for lean hydrogen ( {\mathrm{H}}_2 ${\mathrm{H}}_2$)-air reaction rates for the Large Eddy Simulation (LES) of turbulent reactive flows. This is particularly challenging since {\mathrm{H}}_2 ${\mathrm{H}}_2$ molecules diffuse much faster than heat, leading to large variations in burning rates, thermodiffusive instabilities at the subfilter scale, and complex turbulence-chemistry interactions. Our data-driven approach leverages a Convolutional Neural Network (CNN), trained to approximate filtered burning rates from emulated LES data. First, five different lean premixed turbulent {\mathrm{H}}_2 ${\mathrm{H}}_2$-air flame Direct Numerical Simulations (DNSs) are computed each with a unique global equivalence ratio. Second, DNS snapshots are filtered and downsampled to emulate LES data. Third, a CNN is trained to approximate the filtered burning rates as a function of LES scalar quantities: progress variable, local equivalence ratio, and flame thickening due to filtering. Finally, the performances of the CNN model are assessed on test solutions never seen during training. The model retrieves burning rates with very high accuracy. It is also tested on two filter and downsampling parameters and two global equivalence ratios between those used during training. For these interpolation cases, the model approximates burning rates with low error even though the cases were not included in the training dataset. This a priori study shows that the proposed data-driven machine learning framework is able to address the challenge of modeling lean premixed {\mathrm{H}}_2 ${\mathrm{H}}_2$-air burning rates. It paves the way for a new modeling paradigm for the simulation of carbon-free hydrogen combustion systems.

Digital twins are a new paradigm for our time, offering the possibility of interconnected virtual representations of the real world. The concept is very versatile and has been adopted by multiple communities of practice, policymakers, researchers, and innovators. A significant part of the digital twin paradigm is about interconnecting digital objects, many of which have previously not been combined. As a result, members of the newly forming digital twin community are often talking at cross-purposes, based on different starting points, assumptions, and cultural practices. These differences are due to the philosophical world-view adopted within specific communities. In this paper, we explore the philosophical context which underpins the digital twin concept. We offer the building blocks for a philosophical framework for digital twins, consisting of 21 principles that are intended to help facilitate their further development. Specifically, we argue that the philosophy of digital twins is fundamentally holistic and emergentist. We further argue that in order to enable emergent behaviors, digital twins should be designed to reconstruct the behavior of a physical twin by “dynamically assembling” multiple digital “components”. We also argue that digital twins naturally include aspects relating to the philosophy of artificial intelligence, including learning and exploitation of knowledge. We discuss the following four questions (i) What is the distinction between a model and a digital twin? (ii) What previously unseen results can we expect from a digital twin? (iii) How can emergent behaviours be predicted? (iv) How can we assess the existence and uniqueness of digital twin outputs?

The principal function of an open recirculating system (ORS) is to remove heat from power plant equipment. In particular, the presence of scale on the internal surfaces of ORS heat exchange equipment can reduce heat transfer efficiency, which leads to increased energy consumption and operating costs. The purpose of this article is to investigate the process of calcium carbonate (CaCO3) precipitation formation in terms of the components of the carbonate system and parameters affecting the shift of carbonate equilibrium in an ORS. An appraisal model was used to represent the processes occurring during the operation of an ORS. In this study, it is demonstrated that water heating in ORS condensers leads to the excretion of carbon dioxide (CO2) from the water, while cooling in the cooling towers results in CO2 uptake by the water. These processes significantly influence the state of carbonate equilibrium within the ORS. The study used the results of chemical control of the make-up and cooling water at the ORS Rivne Nuclear Power Plant (RNPP) for 2022. Furthermore, the dependencies of changes in the components of the carbonate system on the pH levels of the make-up (pH 7.51–9.52) and cooling (pH 8.21–9.53) water were revealed, and changes in the cycles of concentration (CоC), total hardness (TH), total dissolved solids (TSD), and total alkalinity (TA) were estimated. Taking into account the obtained correlation dependencies, in general, it was found that the lower the CoC levels, the lower the TA reduction value, and it is possible to increase or decrease the cooling water pH levels, which is determined by the initial state of carbonate equilibrium of make-up water. These findings enable the prediction and control of CaCO3 scale formation through continuous monitoring of water chemistry, making the process more efficient, reliable, and sustainable. The results emphasize the importance of data-driven modeling for optimizing water treatment and reducing operational costs in power plants by reducing CaCO3 scale formation.

Novel methods of data collection and analysis can enhance traditional risk management practices that rely on expert engineering judgment and established safety records, specifically when key conditions are met: Analysis is linked to the decisions it is intended to support, standards and competencies remain up to date, and assurance and verification activities are performed. This article elaborates on these conditions. The reason engineers are required to perform calculations is to support decision-making. Since humans are famously weak natural statisticians, rather than ask stakeholders to implicitly assimilate data, and arrive at a decision, we can instead rely on subject matter experts to explicitly define risk management decision problems. The results of engineering calculation can then also communicate which interventions (if any) are considered to be risk-optimal. It is also proposed that the next generation of engineering standards should learn from the success of open source software development in community building. Interacting with open datasets and code can promote engagement, identification (and resolution) of errors, training and ultimately competence. Finally, the profession’s tradition of independent verification should also be applied to the complex models that will increasingly contribute to the safety of the built environment. Model assurance will be required to keep pace with model development to identify suitable use cases as adequately safe. These are considered to be increasingly important components in ensuring that methods of data-centric engineering can be safely and appropriately adopted in industry.

Regression is a fundamental prediction task common in data-centric engineering applications that involves learning mappings between continuous variables. In many engineering applications (e.g., structural health monitoring), feature-label pairs used to learn such mappings are of limited availability, which hinders the effectiveness of traditional supervised machine learning approaches. This paper proposes a methodology for overcoming the issue of data scarcity by combining active learning (AL) for regression with hierarchical Bayesian modeling. AL is an approach for preferentially acquiring feature-label pairs in a resource-efficient manner. In particular, the current work adopts a risk-informed approach that leverages contextual information associated with regression-based engineering decision-making tasks (e.g., inspection and maintenance). Hierarchical Bayesian modeling allow multiple related regression tasks to be learned over a population, capturing local and global effects. The information sharing facilitated by this modeling approach means that information acquired for one engineering system can improve predictive performance across the population. The proposed methodology is demonstrated using an experimental case study. Specifically, multiple regressions are performed over a population of machining tools, where the quantity of interest is the surface roughness of the workpieces. An inspection and maintenance decision process is defined using these regression tasks, which is in turn used to construct the active-learning algorithm. The novel methodology proposed is benchmarked against an uninformed approach to label acquisition and independent modeling of the regression tasks. It is shown that the proposed approach has superior performance in terms of expected cost—maintaining predictive performance while reducing the number of inspections required.

Stochastic generators are essential to produce synthetic realizations that preserve target statistical properties. We propose GenFormer, a stochastic generator for spatio-temporal multivariate stochastic processes. It is constructed using a Transformer-based deep learning model that learns a mapping between a Markov state sequence and time series values. The synthetic data generated by the GenFormer model preserve the target marginal distributions and approximately capture other desired statistical properties even in challenging applications involving a large number of spatial locations and a long simulation horizon. The GenFormer model is applied to simulate synthetic wind speed data at various stations in Florida to calculate exceedance probabilities for risk management.

Robotic manipulation inherently involves contact with objects for task accomplishment. Traditional motion planning techniques, while having shown their success in collision-free scenarios, may not handle manipulation tasks effectively because they typically avoid contact. Although geometric constraints have been introduced into classical motion planners for tasks that involve interactions, they still lack the capability to fully incorporate contact. In addition, these planning methods generally do not operate on objects that cannot be directly controlled. In this work, building on a recently proposed framework for energy-based quasi-static manipulation, we propose an approach to manipulation planning by adapting a numerical continuation algorithm to compute the equilibrium manifold (EM), which is implicitly derived from physical laws. By defining a manipulation potential energy function that captures interaction and natural potentials, the numerical continuation approach is integrated with adaptive ordinary differential equations that converge to the EM. This allows discretizing the implicit manifold as a graph with a finite set of equilibria as nodes interconnected by weighted edges defined via a haptic metric. The proposed framework is evaluated with an inverted pendulum task, where the explored branch of the manifold demonstrates effectiveness.

Bayesian model updating (BMU) is frequently used in Structural Health Monitoring to investigate the structure’s dynamic behavior under various operational and environmental loadings for decision-making, e.g., to determine whether maintenance is required. Data collected by sensors are used to update the prior of some physics-based model’s latent parameters to yield the posterior. The choice of prior may significantly affect posterior predictions and subsequent decision-making, especially under the typical case in engineering applications of little informative data. Therefore, understanding how the choice of prior affects the posterior prediction is of great interest. In this article, a robust Bayesian inference technique evaluates the optimal and worst-case prior in the vicinity of a chosen nominal prior and their corresponding posteriors. This technique derives an interacting Wasserstein gradient flow that minimizes and maximizes/minimizes the KL divergence between the posterior and the approximation to the posterior, with respect to the approximation to the posterior and the prior. Two numerical case studies are used to showcase the proposed algorithm: a double-banana-posterior and a double-beam structure. Optimal and worst-case priors are modeled by specifying an ambiguity set containing any distribution at a statistical distance to the nominal prior, less than or equal to the radius. The resulting posteriors may be used to yield the lower and upper bounds on subsequent calculations of an engineering metric (e.g., failure probability) used for decision-making. If the metric used for decision-making is not sensitive to the resulting posteriors, it may be assumed that decisions taken are robust to prior uncertainty.

Machine learning’s integration into reliability analysis holds substantial potential to ensure infrastructure safety. Despite the merits of flexible tree structure and formulable expression, random forest (RF) and evolutionary polynomial regression (EPR) cannot contribute to reliability-based design due to absent uncertainty quantification (UQ), thus hampering broader applications. This study introduces quantile regression and variational inference (VI), tailored to RF and EPR for UQ, respectively, and explores their capability in identifying material indices. Specifically, quantile-based RF (QRF) quantifies uncertainty by weighting the distribution of observations in leaf nodes, while VI-based EPR (VIEPR) works by approximating the parametric posterior distribution of coefficients in polynomials. The compression index of clays is taken as an exemplar to develop models, which are compared in terms of accuracy and reliability, and also with deterministic counterparts. The results indicate that QRF outperforms VIEPR, exhibiting higher accuracy and confidence in UQ. In the regions of sparse data, predicted uncertainty becomes larger as errors increase, demonstrating the validity of UQ. The generalization ability of QRF is further verified on a new creep index database. The proposed uncertainty-incorporated modeling approaches are available under diverse preferences and possess significant prospects in broad scientific computing domains.

Processing and extracting actionable information, such as fault or anomaly indicators originating from vibration telemetry, is both challenging and critical for an accurate assessment of mechanical system health and subsequent predictive maintenance. In the setting of predictive maintenance for populations of similar assets, the knowledge gained from any single asset should be leveraged to provide improved predictions across the entire population. In this paper, a novel approach to population-level health monitoring is presented adopting a transfer learning approach. The new methodology is applied to monitor multiple rotating plant assets in a power generation scenario. The focus is on the detection of statistical anomalies as a means of identifying deviations from the typical operating regime from a time series of telemetry data. This is a challenging task because the machine is observed under different operating regimes. The proposed methodology can effectively transfer information across different assets, automatically identifying segments with common statistical characteristics and using them to enrich the training of the local supervised learning models. The proposed solution leads to a substantial reduction in mean square error relative to a baseline model.