The digital twin concept has developed as a method for extracting value from data, and is being developed as a new technique for the design and asset management of high-value engineering systems such as aircraft, energy generating plant, and wind turbines. In terms of implementation, many proprietary digital twin software solutions have been marketed in this domain. In contrast, this paper describes a recently released open-source software framework for digital twins, which provides a browser-based operational platform using Python and Flask. The new platform is intended to maximize connectivity between users and data obtained from the physical twin. This paper describes how this type of digital twin operational platform (DTOP) can be used to connect the physical twin and other Internet-of-Things devices to both users and cloud computing services. The current release of the software—DTOP-Cristallo—uses the example of a three-storey structure as the engineering asset to be managed. Within DTOP-Cristallo, specific engineering software tools have been developed for use in the digital twin, and these are used to demonstrate the concept. At this stage, the framework presented is a prototype. However, the potential for open-source digital twin software using network connectivity is a very large area for future research and development.

The resolution of the Poisson equation is usually one of the most computationally intensive steps for incompressible fluid solvers. Lately, DeepLearning, and especially convolutional neural networks (CNNs), has been introduced to solve this equation, leading to significant inference time reduction at the cost of a lack of guarantee on the accuracy of the solution.This drawback might lead to inaccuracies, potentially unstable simulations and prevent performing fair assessments of the CNN speedup for different network architectures. To circumvent this issue, a hybrid strategy is developed, which couples a CNN with a traditional iterative solver to ensure a user-defined accuracy level. The CNN hybrid method is tested on two flow cases: (a) the flow around a 2D cylinder and (b) the variable-density plumes with and without obstacles (both 2D and 3D), demonstrating remarkable generalization capabilities, ensuring both the accuracy and stability of the simulations. The error distribution of the predictions using several network architectures is further investigated in the plume test case. The introduced hybrid strategy allows a systematic evaluation of the CNN performance at the same accuracy level for various network architectures. In particular, the importance of incorporating multiple scales in the network architecture is demonstrated, since improving both the accuracy and the inference performance compared with feedforward CNN architectures. Thus, in addition to the pure networks’ performance evaluation, this study has also led to numerous guidelines and results on how to build neural networks and computational strategies to predict unsteady flows with both accuracy and stability requirements.

This article develops an ontological description of land use and applies it to incorporate geospatial information describing land coverage into a knowledge-graph-based Universal Digital Twin. Sources of data relating to land use in the UK have been surveyed. The Crop Map of England (CROME) is produced annually by the UK Government and was identified as a valuable source of open data. Formal ontologies to represent land use and the geospatial data arising from such surveys have been developed. The ontologies have been deployed using a high-performance graph database. A customized vocabulary was developed to extend the geospatial capabilities of the graph database to support the CROME data. The integration of the CROME data into the Universal Digital Twin is demonstrated in two use cases that show the potential of the Universal Digital Twin to share data across sectors. The first use case combines data about land use with a geospatial analysis of scenarios for energy provision. The second illustrates how the Universal Digital Twin could use the land use data to support the cross-domain analysis of flood risk. Opportunities for the extension and enrichment of the ontologies, and further development of the Universal Digital Twin are discussed.

As a Bayesian approach to fitting motorway traffic flow models remains rare in the literature, we empirically explore the sampling challenges this approach offers which have to do with the strong correlations and multimodality of the posterior distribution. In particular, we provide a unified statistical model to estimate using motorway data both boundary conditions and fundamental diagram parameters in a motorway traffic flow model due to Lighthill, Whitham, and Richards known as LWR. This allows us to provide a traffic flow density estimation method that is shown to be superior to two methods found in the traffic flow literature. To sample from this challenging posterior distribution, we use a state-of-the-art gradient-free function space sampler augmented with parallel tempering.

Bayesian optimization (BO) has been a successful approach to optimize expensive functions whose prior knowledge can be specified by means of a probabilistic model. Due to their expressiveness and tractable closed-form predictive distributions, Gaussian process (GP) surrogate models have been the default go-to choice when deriving BO frameworks. However, as nonparametric models, GPs offer very little in terms of interpretability and informative power when applied to model complex physical phenomena in scientific applications. In addition, the Gaussian assumption also limits the applicability of GPs to problems where the variables of interest may highly deviate from Gaussianity. In this article, we investigate an alternative modeling framework for BO which makes use of sequential Monte Carlo (SMC) to perform Bayesian inference with parametric models. We propose a BO algorithm to take advantage of SMC’s flexible posterior representations and provide methods to compensate for bias in the approximations and reduce particle degeneracy. Experimental results on simulated engineering applications in detecting water leaks and contaminant source localization are presented showing performance improvements over GP-based BO approaches.

Many real-world problems require to optimize trajectories under constraints. Classical approaches are often based on optimal control methods but require an exact knowledge of the underlying dynamics and constraints, which could be challenging or even out of reach. In view of this, we leverage data-driven approaches to design a new end-to-end framework which is dynamics-free for optimized and realistic trajectories. Trajectories are here decomposed on function basis, trading the initial infinite dimension problem on a multivariate functional space for a parameter optimization problem. Then a maximum a posteriori approach which incorporates information from data is used to obtain a new penalized optimization problem. The penalized term narrows the search on a region centered on data and includes estimated features of the problem. We apply our data-driven approach to two settings in aeronautics and sailing routes optimization. The developed approach is implemented in the Python library PyRotor.

Neural networks-based learning of the distribution of non-dispatchable renewable electricity generation from sources, such as photovoltaics (PV) and wind as well as load demands, has recently gained attention. Normalizing flow density models are particularly well suited for this task due to the training through direct log-likelihood maximization. However, research from the field of image generation has shown that standard normalizing flows can only learn smeared-out versions of manifold distributions. Previous works on normalizing flow-based scenario generation do not address this issue, and the smeared-out distributions result in the sampling of noisy time series. In this paper, we exploit the isometry of the principal component analysis (PCA), which sets up the normalizing flow in a lower-dimensional space while maintaining the direct and computationally efficient likelihood maximization. We train the resulting principal component flow (PCF) on data of PV and wind power generation as well as load demand in Germany in the years 2013–2015. The results of this investigation show that the PCF preserves critical features of the original distributions, such as the probability density and frequency behavior of the time series. The application of the PCF is, however, not limited to renewable power generation but rather extends to any dataset, time series, or otherwise, which can be efficiently reduced using PCA.

The ability to wirelessly stream data from sensors on heavy mobile equipment provides opportunities to proactively assess asset condition. However, data analysis methods are challenging to apply due to the size and structure of the data, which contain inconsistent and asynchronous entries, and large periods of missing data. Current methods usually require expertise from site engineers to inform variable selection. In this work, we develop a data preparation method to clean and arrange this streaming data for analysis, including a data-driven variable selection. Data are drawn from a mining industry case study, with sensor data from a primary production excavator over a period of 9 months. Variables include 58 numerical sensors and 40 binary indicators captured in 45-million rows of data describing the conditions and status of different subsystems of the machine. A total of 57% of time stamps contain missing values for at least one sensor. The response variable is drawn from fault codes selected by the operator and stored in the fleet management system. Application to the hydraulic system, for 21 failure events identified by the operator, shows that the data-driven selection contains variables consistent with subject matter expert expectations, as well as some sensors on other systems on the excavator that are less easy to explain from an engineering perspective. Our contribution is to demonstrate a compressed data representation using open-high-low-close and variable selection to visualize data and support identification of potential indicators of failure events from multivariate streamed data.

In this article, a novel drive mode, “intelligent vehicle drive mode” (IVDM), was proposed, which augments the vehicle engine performance in real-time. This drive mode predicts the driver behavior vector (DBV), which optimizes the vehicle engine performance, and the metric of optimal vehicle engine performance was defined using the elements of engine operating point (EOP) and heating ventilation and air conditioning system (HVAC). Deep learning (DL) models were developed by mapping the vehicle level vectors (VLV) with EOP and HVAC parameters, and the trained functions were utilized to predict the future states of DBV reflecting augmented vehicle engine performance. The iterative analysis was performed by empirically estimating the future states of VLV in the allowable range of DBV and was fed into the DL model to predict the performance vectors. The defined vehicle engine performance metric was applied to the predicted vectors, and thus optimal DBV is the instantaneous output of the IVDM. The analytical and validation techniques were developed using field data obtained from General Motors Inc., Warren, Michigan. Finally, the proposed concept was quantified by analyzing the instantaneous engine efficiency (IEE) and smoothness measure of the instantaneous engine map (IEM).

Mathematical models are essential to analyze and understand the dynamics of complex systems. Recently, data-driven methodologies have gotten a lot of attention which is leveraged by advancements in sensor technology. However, the quality of obtained data plays a vital role in learning a good and reliable model. Therefore, in this paper, we propose an efficient heuristic methodology to collect data both in the frequency domain and the time domain, aiming at having more information gained from limited experimental data than equidistant points. In the frequency domain, the interpolation points are restricted to the imaginary axis as the transfer function can be estimated easily on the imaginary axis. The efficiency of the proposed methodology is illustrated by means of several examples, and its robustness in the presence of noisy data is shown.

Predicting the occurrence of thermoacoustic instabilities is of major interest in a variety of engineering applications such as aircraft propulsion, power generation, and industrial heating. Predictive methodologies based on a physical approach have been developed in the past decades, but have a moderate-to-high computational cost when exploring a large number of designs. In this study, the stability prediction capabilities and computational cost of four well-established classification algorithms—the K-Nearest Neighbors, Decision Tree (DT), Random Forest (RF), and Multilayer Perceptron (MLP) algorithms—are investigated. These algorithms are trained using an in-house physics-based low-order network model tool called OSCILOS. All four algorithms are able to predict which configurations are thermoacoustically unstable with a very high accuracy and a very low runtime. Furthermore, the frequency intervals containing unstable modes for a given configuration are also accurately predicted using multilabel classification. The RF algorithm correctly predicts the overall stability and finds all frequency intervals containing unstable modes for 99.6 and 98.3% of all configurations, respectively, which makes it the most accurate algorithm when a large number of training examples is available. For smaller training sets, the MLP algorithm becomes the most accurate algorithm. The DT algorithm is found to be slightly less accurate, but can be trained extremely quickly and runs about a million times faster than a traditional physics-based low-order network model tool. These findings could be used to devise a new generation of combustor optimization tools that would run much faster than existing codes while retaining a similar accuracy.

Data mining and knowledge discovery (DMKD) focuses on extracting useful information from data. In the chemical process industry, tasks such as process monitoring, fault detection, process control, optimization, etc., can be achieved using DMKD. However, the selection of the appropriate method for each step in the DMKD process, namely data cleaning, sampling, scaling, dimensionality reduction (DR), clustering, clustering analysis and data visualization to obtain meaningful insights is far from trivial. In this contribution, a computational environment (FastMan) is introduced and used to illustrate how method selection affects DMKD in chemical process data. Two case studies, using data from a simulated natural gas liquid plant and real data from an industrial pyrolysis unit, were conducted to demonstrate the applicability of these methodologies in real-life scenarios. Sampling and normalization methods were found to have a great impact on the quality of the DMKD results. Also, a neighbor graphs method for DR, t-distributed stochastic neighbor embedding, outperformed principal component analysis, a matrix factorization method frequently used in the chemical process industry for identifying both local and global changes.

Reproducibility of a deep-learning fully convolutional neural network is evaluated by training several times the same network on identical conditions (database, hyperparameters, and hardware) with nondeterministic graphics processing unit operations. The network is trained to model three typical time–space-evolving physical systems in two dimensions: heat, Burgers’, and wave equations. The behavior of the networks is evaluated on both recursive and nonrecursive tasks. Significant changes in models’ properties (weights and feature fields) are observed. When tested on various benchmarks, these models systematically return estimations with a high level of deviation, especially for the recurrent analysis which strongly amplifies variability due to the nondeterminism. Trainings performed with double floating-point precision provide slightly better estimations and a significant reduction of the variability of both the network parameters and its testing error range.

Reduced-order models (ROMs) are computationally inexpensive simplifications of high-fidelity complex ones. Such models can be found in computational fluid dynamics where they can be used to predict the characteristics of multiphase flows. In previous work, we presented a ROM analysis framework that coupled compression techniques, such as autoencoders, with Gaussian process regression in the latent space. This pairing has significant advantages over the standard encoding–decoding routine, such as the ability to interpolate or extrapolate in the initial conditions’ space, which can provide predictions even when simulation data are not available. In this work, we focus on this major advantage and show its effectiveness by performing the pipeline on three multiphase flow applications. We also extend the methodology by using deep Gaussian processes as the interpolation algorithm and compare the performance of our two variations, as well as another variation from the literature that uses long short-term memory networks, for the interpolation.