Search

Machine learned generative design is an extension of the generative design process, addressing its inherent limitations, particularly those of interoperability. The proposed approach uses machine learning-based surrogate models, trained on computational model data, to replicate design evaluations and integrate them into a common environment. In this way, design alternatives can be generated and tested that satisfy all design requirements and considerations. The effectiveness of this approach is demonstrated by the design and optimisation of the enclosure structure for the New Robotic Telescope. Its complexity is characterised by multiple operating states that the enclosure can assume, in particular the closed state and the opening/closing state, each of which has a different structural behaviour. Using our approach, the results from each state were replicated with machine learning models and combined into a single evaluation model. This resulted in finding multiple solutions that outperformed the benchmark design. The results demonstrate not only the success of our method over conventional strategies, but also highlight its potential to redefine future design optimisation processes. • Novel method for extending generative design methodology through machine learning • Addresses the limitation of interoperability by integrating trained machine learning models into a combined evaluation model. • Machine learning-based surrogate models enable efficient design exploration. • Demonstrated effectiveness in successfully designing and optimising the enclosure structure of the New Robotic Telescope. • Several solutions were found that outperform the benchmark design, demonstrating the potential of this method. [ABSTRACT FROM AUTHOR]

Physics informed neural networks (PINNs) have been explored extensively in the recent past for solving various forward and inverse problems for facilitating querying applications in fluid mechanics. However, investigations on PINNs for unsteady flows past moving bodies, such as flapping wings are scarce. Earlier studies mostly relied on transferring the problems to a body-attached frame of reference, which could be restrictive towards handling multiple moving bodies/deforming structures. The present study attempts to couple the benefits of PINNs with a fixed Eulerian frame of reference, and proposes an immersed boundary aware framework for developing surrogate models for unsteady flows past moving bodies. Specifically, high-resolution velocity reconstruction and pressure recovery as a hidden variable are the main goals. The framework has been developed by using downsampled velocity data obtained from prior simulations to train the PINNs model. The efficacy of the velocity reconstruction has been tested against high resolution IBM simulation data, whereas the efficacy of the pressure recovery has been tested against high resolution simulation data from an arbitrary Lagrange Eulerian (ALE) solver. Under the present framework, two PINN variants, (i) a moving-boundary-enabled standard Navier–Stokes based PINN (MB-PINN), and, (ii) a moving-boundary-enabled IBM based PINN (MB-IBM-PINN) have been formulated. Relaxation of physics constraints in PINNs models has been identified to be a useful strategy in improving the predictions. A fluid-solid partitioning of the physics losses in MB-IBM-PINN has been allowed, in order to investigate the effects of solid body points while training. This strategy enables MB-IBM-PINN to match with the performance of MB-PINN under certain loss-weighting conditions. Interestingly, MB-PINN is found to be superior to MB-IBM-PINN when a priori knowledge of the solid body position and velocity is available. To improve the data efficiency of MB-PINN, a physics based data sampling technique has also been investigated. It is observed that a suitable combination of physics constraint relaxation and physics based sampling can achieve a model performance comparable to the case of using all the data points, under a fixed training budget. [ABSTRACT FROM AUTHOR]

Uncertainty quantification has proven to be an indispensable study for enhancing reliability and robustness of engineering systems in the early design phase. Single and multi-fidelity surrogate modelling methods have been used to replace the expensive high fidelity analyses which must be repeated many times for uncertainty quantification. However, since the number of analyses required to build an accurate surrogate model increases exponentially with the number of random input variables, most surrogate modelling methods suffer from the curse of dimensionality. As an alternative approach, the Low-Rank Approximation method can be applied to high-dimensional uncertainty quantification studies with a low computational cost, where the number of coefficients for building the surrogate model increases only linearly with the number of random input variables. In this study, the Low-Rank Approximation method is implemented for multi-fidelity applications with additive and multiplicative correction approaches to make the high-dimensional uncertainty quantification analysis more efficient and accurate. The developed uncertainty quantification methodology is tested on supersonic aircraft design problems and its predictions are compared with the results of single- and multi-fidelity Polynomial Chaos Expansion and Monte Carlo methods. For the same computational cost, the Low-Rank Approximation method outperformed both in surrogate modeling and uncertainty quantification cases for all the benchmarks and real-world engineering problems addressed in the present study.

Fast machine learning-based surrogate models are trained to emulate slow, high-fidelity engineering simulation models to accelerate engineering design tasks. This introduces uncertainty as the surrogate is only an approximation of the original model.Bayesian methods can quantify that uncertainty, and deep learning models exist that follow the Bayesian paradigm. These models, namely Bayesian neural networks and Gaussian process models, enable us to give predictions together with an estimate of the model’s uncertainty. As a result we can derive uncertainty-aware surrogate models that can automatically identify unseen design samples that may cause large emulation errors. For these samples the high-fidelity model can be queried instead. This paper outlines how the Bayesian paradigm allows us to hybridize fast but approximate and slow but accurate models.In this paper, we train two types of Bayesian models, dropout neural networks and stochastic variational Gaussian Process models, to emulate a complex high dimensional building energy performance simulation problem. The surrogate model processes 35 building design parameters (inputs) to estimate 12 annual building energy performance metrics (outputs). We benchmark both approaches, prove their accuracy to be competitive, and show that errors can be reduced by up to 30% when the 10% of samples with the highest uncertainty are transferred to the high-fidelity model.

Embedding ergonomics in the design process is often delayed as it requires building physical prototypes and human subject data collection. This causes retrofitting or expensive late design changes which increase overall cost and time-to-market. This paper presents a computational design methodology that employs Digital Human Modelling (DHM) research and the surrogate modelling technique to integrate ergonomics early in design. The goal of this methodology is to provide a proactive ergonomics assessment toolset that allows designers to identify the effects of design variables on human performance, computationally. In this research, a design space exploration study was conducted to evaluate the ergonomics of a concept aircraft cockpit model built based on a set of flight deck layout variables. Variability in the pilot's anthropometry was added through accommodating the extremities in population-based on computational human manikins. A genetic algorithm with the surrogate model was used to search the concept variants that meet ergonomics objectives. The results show that the proposed proactive ergonomics approach allows designers to explore the design space and evaluate trade-offs between the competing design objectives early in design. Our computational ergonomics approach, based on the cockpit packaging study, demonstrates the advantages of the DHM based surrogate modelling during product conceptualisation. [ABSTRACT FROM AUTHOR]

Operating satellites at altitudes in Very Low Earth Orbit (VLEO) has many advantages. However, due to the higher atmospheric density of this region, satellites encounter significantly higher atmospheric drag. Depending on the mission, this may require a propulsive system to maintain the orbit which costs both fuel mass and volume. It is therefore desirable to reduce the drag in order to either reduce these costs or to extend the operational life. In this paper a series of viable aeroshell profiles are identified for satellites operating in VLEO using a Radial Basis Function-based surrogate model with data generated using both Panel Methods and Discrete Simulation Monte Carlo simulations. It was demonstrated that a maximum drag reduction of between 21% and 35% was achievable for the profiles when optimising a bi-conic profile for minimum drag based on Discreet Simulation Monte Carlo simulations with an energy accommodation coefficient of 0.95. Accounting for the loss of internal volume and assuming the reduction in fuel mass results in an equally proportioned reduction in fuel system volume it was observed that only a 13% to 27% reduction was achieved. • Bi-conic Aeroshell profiles can be used to reduces aerodynamic drag in VLEO. • A reduction in the aerodynamic drag can be achieved by adjusting the tail profile. • For equal reduction in volume and drag, blunter Aeroshell profiles are preferable. • An application of a Radial Basis Function-based surrogate model trained using DSMC. [ABSTRACT FROM AUTHOR]

Virtual assembly (VA) is a method to simulate the physical assembly (PA) of scanned parts. Small local part deviations can accumulate to large assembly deviations limiting the product quality. The propagation of geometrical deviations onto the assembly is a crucial step in tolerance management to assess the assembly quality. Current approaches for VA do not sufficiently consider the physical joining process. Therefore, the propagated assembly geometry may deviate strongly from the PA. In the state of the art, only specific and complex methods for particular joining processes are known. In this paper, the concept of Surrogate Models (SMs) is introduced, representing the connection between part and assembly geometries for particular joining processes. A Surrogate Modelling Toolbox (SMT) is developed that is intended to cover the variety of joining processes by the implementation of suitable SMs. A particular SM is created by the composition of suitable Surrogate Operations (SOs). An open list of SOs is presented. The composition of a SM is studied for a laser welding process of two polymer components. The resulting VA is compared to the PA in order to validate the developed model and is quantified by the exploitation ratio R. [ABSTRACT FROM AUTHOR]

A comparative study of four well established surrogate models used to predict the non-linear entrainment performance of a dual-phase fluid driving jet pump (JP) apparatus is performed. A JP design flow configuration comprising a dual-phase (air and water) flow driving a secondary gas-air flow, for which no one has ever provided a unique set of design solutions, is described. For the construction of the global approximations (GA), the response surface methodology (RSM), Kriging and the radial basis function artificial neural network (RBFANN), were primarily used. The stacked/ensemble models methodology was integrated in this study, to improve the predictive model results, thus providing accurate GA that facilitate the multi-variable non-linear response design optimisation. An error analysis of all four models along with a multiple model accuracy analysis of each case study were performed. The RSM, Kriging, RBFANN and stacked models formed part of the surrogate-based optimisation, having the entrainment ratio as the main objective function. Optimisation problems were solved by the interior-point algorithm and the genetic algorithm and incurred a hybrid formulation of both algorithms. A total of 60 optimisation problems were formulated and solved with all three approximation models. Results showed that the hybrid formulation having the level-2 ensemble Kriging model performed best, predicting the experimental performance results for all JP models within an error margin of less than 10 % in 90 % of the cases. [ABSTRACT FROM AUTHOR]

We consider Bayesian inference when only a limited number of noisy log-likelihood evaluations can be obtained. This occurs for example when complex simulator-based statistical models are fitted to data, and synthetic likelihood (SL) method is used to form the noisy log-likelihood estimates using computationally costly forward simulations. We frame the inference task as a sequential Bayesian experimental design problem, where the log-likelihood function is modelled with a hierarchical Gaussian process (GP) surrogate model, which is used to efficiently select additional log-likelihood evaluation locations. Motivated by recent progress in the related problem of batch Bayesian optimisation, we develop various batch-sequential design strategies which allow to run some of the potentially costly simulations in parallel. We analyse the properties of the resulting method theoretically and empirically. Experiments with several toy problems and simulation models suggest that our method is robust, highly parallelisable, and sample-efficient. [ABSTRACT FROM AUTHOR]

Excitable systems give rise to important phenomena such as heat waves, epidemics and cardiac arrhythmias. Understanding, forecasting and controlling such systems requires reliable mathematical representations. For cardiac tissue, computational models are commonly generated in a reaction-diffusion framework based on detailed measurements of ionic currents in dedicated single-cell experiments. Here, we show that recorded movies at the tissue-level of stochastic pacing in a single variable are sufficient to generate a mathematical model. Via exponentially weighed moving averages, we create additional state variables, and use simple polynomial regression in the augmented state space to quantify excitation wave dynamics. A spatial gradient-sensing term replaces the classical diffusion as it is more robust to noise. Our pipeline for model creation is demonstrated for an in-silico model and optical voltage mapping recordings of cultured human atrial myocytes and only takes a few minutes. Our findings have the potential for widespread generation, use and on-the-fly refinement of personalised computer models for non-linear phenomena in biology and medicine, such as predictive cardiac digital twins., Competing Interests: Declaration of competing interest None declared., (Copyright © 2024 The Authors. Published by Elsevier Ltd.. All rights reserved.)

Electromagnetic (EM) simulations have become an indispensable tool in the design of contemporary antennas. EM‐driven tasks, for example, parametric optimization, entail considerable computational efforts, which may be reduced by employing surrogate models. Yet, data‐driven modelling of antenna characteristics is largely hindered by the curse of dimensionality. This may be addressed using the recently reported domain‐confinement techniques, especially the nested‐kriging framework, which permits rendering of reliable surrogates over wide ranges of antenna parameters while greatly reducing the computational overhead of training data acquisition. Focused on modelling of multi‐band antennas, this paper attempts to reduce the cost of surrogate construction even further by incorporating variable‐fidelity simulations into the nested kriging. The principal challenge being design‐dependent frequency shifts between the models of various fidelities is handled through the development of a customized frequency scaling and output space mapping. Validation is carried out using a dual‐band dipole antenna modeled over broad ranges of operating conditions. A small training data set is sufficient to secure the predictive power comparable to that of the nested kriging model set up using solely high‐fidelity data, and by far exceeding the accuracy of conventional surrogates. Application examples for antenna optimization and experimental verification of the selected designs are also provided. [ABSTRACT FROM AUTHOR]

Sparse grid interpolation is a popular numerical discretization technique for the treatment of high dimensional, multivariate problems. We consider the case of using time-series data to calibrate epidemiological models from both phenomenological and mechanistic perspectives using this computational tool. By capturing the dynamics underlying both global and local spaces, our algorithm identifies potentially optimal regions of the parameter space and directs computational effort towards resolving the dynamics and resulting fits of these regions. We demonstrate how sparse grid interpolants can be effectively deployed to fit available data and discriminate between competing hypotheses to explain the current cholera epidemic in Yemen.

In this paper, a multi-objective probabilistic design optimisation approach is presented for reliability and robustness analysis of composite structures and demonstrated on a mono-omega-stringer stiffened panel. The proposed approach utilises a global surrogate model of the composite structure while accounting for uncertainties in material properties as well as geometry. Unlike the multi-level optimisation approach which freezes some parameters at each level, the proposed approach allows for all parameters to change at the same time and hence ensures global optimum solutions in the given parameter design space (for both probabilistic and deterministic optimisations) within a certain degree of accuracy. The proposed approach is used in this study to conduct extensive multi-objective probabilistic and deterministic optimisations (without considering safety factors) on a mono-stringer stiffened panel. In particular, a global surrogate model is developed utilising the computational power of a high-performance computing facility. The inputs of the surrogate model are the omega-stringer geometry and the mechanical properties of the composite material, while the outputs are the fundamental linear buckling load (LBL) and the nonlinear post-buckling strength (NPS). LBL and NPS are obtained via detailed parametric finite element models of the mono-stringer stiffened panel; in the nonlinear model, the interface between the skin and the omega-stringer is modelled via cohesive elements to allow for debonding in the post-buckled regime. Extensive multi-objective optimisations are conducted on the surrogate model using deterministic and probabilistic approaches to examine the omega-stringer geometric parameters mostly affecting the system robustness and reliability. The differences between deterministic and probabilistic designs are highlighted as well. [ABSTRACT FROM AUTHOR]

This article applies methods typically used in engineering applications to the optimization of human locomotion, more specifically a swimmer's underwater 'dolphin kick'. This is a dual-objective problem which seeks the optimal trade-off between thrust (simulated using Lighthill's fish propulsion method) and the force in the muscles to produce this thrust (simulated using musculoskeletal modelling). The expense of the analyses leads to the use of a surrogate modelling based optimization technique (multi-objective expected improvement using Kriging). The results indicate that optimal human motion does in many respects follow that of fish, with low frequency eel-like techniques suitable for endurance and a high frequency tuna-like kick for high thrust. The MATLAB® code, including thrust and muscle activity models, is made available in an online repository. [ABSTRACT FROM AUTHOR]

Full-wave electromagnetic (EM) analysis is a versatile tool for evaluating the performance of high-frequency components. Its potential drawback is its high computational cost, inhibiting the execution of EM-driven tasks requiring massive simulations. The applicability of equivalent network models is limited owing to the topological complexity of compact microstrip components because of EM cross-coupling effects. Development of alternative representations (surrogate models) is therefore necessary. This article proposes a two-level methodology for reliable modelling of compact microstrip components. The keystone is to define the surrogate domain using the first-level model approximating the set of pre-existing reference designs. This limits the volume of the parameter space region that needs to be sampled when constructing the second-level model. The presented approach provides far greater accuracy than conventional methods and is capable of establishing surrogates covering wide ranges of geometric parameters and operating conditions of a particular structure. Applications for parametric optimization are also provided. [ABSTRACT FROM AUTHOR]

This paper proposes a Kriging assisted strategy to expedite evolutionary computation for solving Optimal Power Flow (OPF) problems. First, two algorithms were developed–a Kriging Assisted Genetic Algorithm (KAGA) and a Kriging Assisted Particle Swarm Optimization (KAPSO) - and tested using unconstrained benchmark functions; it was found that both algorithms provided reliable and robust solutions. Accordingly, KAGA and KAPSO were selected and tested on the IEEE 30 and 118 bus systems for minimizing generation costs and active power losses. It is shown that the proposed KAPSO outperforms other algorithms, especially in terms of the computation time. In reference to the solution quality yielded by KAGA and KAPSO, the proposed Kriging assisted strategy offers a promising method to improve the performance of evolutionary based computation when solving OPF problems. [ABSTRACT FROM AUTHOR]

We consider the problem of constructing predictive models (surrogate models) to tackle challenges of industrial engineering design. The author analyzed the needs of industrial applications, formulated a number of new mathematical and algorithmic problems and developed appropriate methods of data modeling. [ABSTRACT FROM AUTHOR]

Reliability-based design optimization (RBDO) is an active field of research with an ever increasing number of contributions. Numerous methods have been proposed for the solution of RBDO, a complex problem that combines optimization and reliability analysis. Classical approaches are based on approximation methods and have been classified in review papers. In this paper, we first review classical approaches based on approximation methods such as FORM, and also more recent methods that rely upon surrogate modelling and Monte Carlo simulation. We then propose a generalization of the existing surrogate-assisted and simulation-based RBDO techniques using a unified framework that includes three independent blocks, namely adaptive surrogate modelling, reliability analysis, and optimization. These blocks are non-intrusive with respect to each other and can be plugged independently in the framework. After a discussion on numerical considerations that require attention for the framework to yield robust solutions to various types of problems, the latter is applied to three examples (using two analytical functions and a finite element model). Kriging and support vector machines regression together with their own active learning schemes are considered in the surrogate model block. In terms of reliability analysis, the proposed framework is illustrated using both crude Monte Carlo and subset simulation. Finally, the covariance matrix adaptation-evolution scheme (CMA-ES), a global search algorithm, or sequential quadratic programming (SQP), a local gradient-based method, is used in the optimization block. The comparison of the results to benchmark studies shows the effectiveness and efficiency of the proposed framework. [ABSTRACT FROM AUTHOR]

Parametric reduced-order modelling often serves as a surrogate method for hemodynamics simulations to improve the computational efficiency in many-query scenarios or to perform real-time simulations. However, the snapshots of the method require to be collected from the same discretisation, which is a straightforward process for physical parameters, but becomes challenging for geometrical problems, especially for those domains featuring unparameterised and unique shapes, e.g. patient-specific geometries. In this work, a data-driven surrogate model is proposed for the efficient prediction of blood flow simulations on similar but distinct domains. The proposed surrogate model leverages group surface registration to parameterise those shapes and formulates corresponding hemodynamics information into geometry-informed snapshots by the diffeomorphisms constructed between a reference domain and original domains. A non-intrusive reduced-order model for geometrical parameters is subsequently constructed using proper orthogonal decomposition, and a radial basis function interpolator is trained for predicting the reduced coefficients of the reduced-order model based on compressed geometrical parameters of the shape. Two examples of blood flowing through a stenosis and a bifurcation are presented and analysed. The proposed surrogate model demonstrates its accuracy and efficiency in hemodynamics prediction and shows its potential application toward real-time simulation or uncertainty quantification for complex patient-specific scenarios. • Propose a reduced-order modelling method for hemodynamics simulations within different shapes of domains. • Surface registration with currents is leveraged to parametrise domains and achieve the diffeomorphisms between domains. • A non-intrusive reduced-order model is constructed with geometry-inform snapshots. • Two synthetic examples of blood flow simulations in stenosis vessels and bifurcation vessels are demonstrated and discussed. [ABSTRACT FROM AUTHOR]

This article proposes an auxiliary function for imposing the boundary and initial conditions in physics-informed neural network models in a hard manner that accelerates the learning process. This auxiliary function consists of two pre-trained neural networks and a main deep neural network with trainable parameters. The novelty of this new auxiliary function is the input of main deep neural network, which takes the outputs of distance function and boundary function as inputs in addition to the spatiotemporal variables. We demonstrate the efficacy and general applicability of the proposed model by applying it to several benchmark-forward problems namely, advection, Helmholtz and Klein-Gordon equations. The accuracy of predictions is examined by comparison with exact solutions. Our findings imply the superiority of the proposed model because of the improvement of the loss convergence to lower values by one order of magnitude for the same number of epochs. In the case of the advection equation, the relative L2 error has been reduced from 0.025 to 0.0201, and from 0.016 to 0.0152. When applied to the Helmholtz equation, our novel model achieved an error of 8.67 × 10−3, surpassing the conventional model's performance, which yielded an error of 6.04 × 10−2. Furthermore, for the Klein-Gordon equation, our new model led to a remarkable reduction in the relative L2 error, from 0.18 to an impressive 4.2 × 10−2. [ABSTRACT FROM AUTHOR]

Stress prediction in porous materials and structures is challenging due to the high computational cost associated with direct numerical simulations. Convolutional Neural Network (CNN) based architectures have recently been proposed as surrogates to approximate and extrapolate the solution of such multiscale simulations. These methodologies are usually limited to 2D problems due to the high computational cost of 3D voxel based CNNs. We propose a novel geometric learning approach based on a Graph Neural Network (GNN) that efficiently deals with three-dimensional problems by performing convolutions over 2D surfaces only. Following our previous developments using pixel-based CNN, we train the GNN to automatically add local fine-scale stress corrections to an inexpensively computed coarse stress prediction in the porous structure of interest. Our method is Bayesian and generates densities of stress fields, from which credible intervals may be extracted. As a second scientific contribution, we propose to improve the extrapolation ability of our network by deploying a strategy of online physics-based corrections. Specifically, we condition the posterior predictions of our probabilistic predictions to satisfy partial equilibrium at the microscale, at the inference stage. This is done using an Ensemble Kalman algorithm, to ensure tractability of the Bayesian conditioning operation. We show that this innovative methodology allows us to alleviate the effect of undesirable biases observed in the outputs of the uncorrected GNN, and improves the accuracy of the predictions in general. • Graph Neural Networks are used to predict stress fields in porous structures • Coarse-scale fields and microstructures surface meshes given as inputs to the GNN • Bayesian framework yields uncertainty intervals for the GNN predictions • Physical constraints enforced online via an ensemble Kalman technique • Reasonably sized datasets required for stress predictions in random microstructures [ABSTRACT FROM AUTHOR]

• The physics guided wavelet convolutional neural network (PGwCNN) is proposed to predict the structural wind-induced response. • The law of dynamics, is embedded in the wCNN to inform the learning and constrain the training to a feasible space. • A framework of efficient reliability analysis using the proposed method is introduced. • The dynamic reliability analysis of a high and flexible wind turbine tower was conducted. Deep neural network (NN) has become one of the common choices of surrogate model for reliability analysis of structural dynamic response under complex wind loads. However, the cost of data acquisition is usually prohibitive, making it hard for NN training to accurately replace the complex wind-structure coupled system using the limited data. In this work, a physics guided wavelet convolutional NN (PGwCNN) is regarded as surrogate model of a wind turbine tower (WinTT) and utilized for wind effects prediction. The law of dynamics is regarded as the prior knowledge and encoded in the NN, which can enhance the learning capability of the trained model. Then, the performance of the PGwCNN is demonstrated through the dynamic reliability analysis of a high and flexible WinTT using probability density evolution method (PDEM). What's more, the L2 regularization and dropout layers are utilized to alleviate the overfitting issues. N-fold cross validation and grid search method are adopted to maximize the use of the limited training datasets. Results demonstrate that PGwCNN is capable to reach a compromise between efficiency and accuracy for predicting wind effects on the WinTT. In addition, the PGwCNN can effectively enhance the calculation efficiency of the reliability analysis. [ABSTRACT FROM AUTHOR]

This paper presents an innovative hierarchical meta-modelling approach for predicting the effective elastic properties of stone injected with composite cement grouts containing Carbon NanoTubes (CNTs) for structural rehabilitation purposes. The study first addresses the homogenization of the nano- and micro-scales through numerical representative volume elements. The first numerical model focuses on the cement matrix doped with randomly oriented CNTs, while the second model represents the porous stone with CNT/cement grout filling the pores. However, the computational burden associated with these models poses a significant limitation when analysing large-scale macrostructural elements. To overcome this challenge, a hierarchical meta-modelling approach based on two nested Kriging surrogate models is proposed. This approach offers a fast and accurate alternative to bypass the time-consuming numerical homogenization process. The first surrogate model establishes a connection between the microstructural characteristics of the CNTs and cement and the effective elastic properties of the composite grout. Subsequently, a second meta-model expands over the first one by introducing two additional variables: the elastic modulus of the stone and the porosity, allowing for estimation of the overall elastic properties of CNT/cement/stone composites. The accuracy and efficiency of the proposed hierarchical meta-model is demonstrated through detailed parametric analyses. To showcase its potential for optimizing rehabilitation interventions, the paper is closed with its application to a benchmark case study of a stone column injected with CNT/cement grout. • Composite CNT/cement grouts are investigated for the first in the literature. • A novel nested surrogate modelling approach is proposed for multi-scale composites. • Fast stiffness estimates of porous stone injected with CNT/cement grout are obtained. • The proposed approach is used to assess the performance of grout injected columns. [ABSTRACT FROM AUTHOR]

Generalisability and the consistency of the a posteriori results are the most critical points of view regarding data-driven turbulence models. This study presents a progressive improvement of turbulence models using simulation-driven Bayesian optimisation with Kriging surrogates where the optimisation of the models is achieved by a multi-objective approach based on duct flow quantities. We aim for the augmentation of secondary-flow prediction capability in the linear eddy-viscosity model k − ω SST without violating its original performance on canonical cases e.g. channel flow. Progressively data-augmented explicit algebraic Reynolds stress models (PDA-EARSMs) for k − ω SST are obtained enabling the prediction of secondary flows that the standard model fails to predict. The new models are tested on channel flow cases guaranteeing that they preserve the successful performance of the original k − ω SST model. Subsequently, numerical verification is performed for various test cases. Regarding the generalisability of the new models, results of unseen test cases demonstrate a significant improvement in the prediction of secondary flows and streamwise velocity. These results highlight the potential of the progressive approach to enhance the performance of data-driven turbulence models for fluid flow simulation while preserving the robustness and stability of the solver. • Two models have been developed to augment the k − ω SST turbulence model. • The new PDA-EARSMs are able to predict secondary flows. • CFD-driven surrogate optimisation has proven to be a robust method to enhance models. • The surrogate and Pareto-front solutions obtained are validated numerically. [ABSTRACT FROM AUTHOR]

Assessing return level RL (with return period typically ranging from 100 to 500 years) for extreme waves in cyclone-prone regions is often made problematic by the lack of sufficient representative samples to properly fit the extreme value probability distributions. Motivated by the Guadeloupe context (French West Indies), we address this problem when the wave numerical model is too expensive to be run a large number of times. A possible option is to use a machine learning (ML) model in place of the long-running numerical mode. By construction, ML predictions are however uncertain, because they are learned from a limited number of pre-calculated wave simulation results. We propose to account for this type of error along the inference of the parameters of the extreme value probability distribution within an Approximate Bayesian Computation (ABC) scheme. Using a subset of the database of numerically computed H S for Guadeloupe (for approximately 700 synthetic cyclones representative of 1000 years of cyclonic activity), we train a random forest (RF) regression model to relate the cyclone characteristics (radius, atmospheric pressure, distance to cyclone eye) to H S. The quantile RF model is then used to model the prediction error within the ABC scheme. By comparison to the 100-year and 500-year RL reference solutions (calculated using the entire database), we show that the ML-based approach achieves low bias and high reliability of the RL estimates at locations both along Guadeloupe coasts and in deep ocean environments for low sample size (down to ∼70). Provided that the prediction error is low-to-moderate, integrating this type of uncertainty improves the performance of the ML-based approach, and decreases the computational burden of the return level estimation. [ABSTRACT FROM AUTHOR]

Pollution source identification is vital in water safety management. An integrated simulation-optimization modelling framework comprising a process-based hydrodynamic water quality model, artificial neural network surrogate model and particle swarm optimization (PSO) was proposed to achieve rapid, accurate and reliable pollution source identification. In this study, the hydrodynamics and water quality processes in a straight lab-based flume were simulated to test pollution source identification under steady flow conditions. Additionally, the pollution source identification in the unsteady flow conditions was examined using a real-life estuary, specifically the Yangtze River estuary. First, we developed two process-based models to simulate hydrodynamics and water quality in the flume and estuary. Then, the data generated from the process-based models were used to develop surrogate models. Three typical artificial neural networks (ANNs) algorithms: backpropagation (BP), radial basis function (RBF) and general regression neural networks (GRNN) were selected to develop surrogates for process-based models (PBMs), and they were coupled with PSO algorithm to achieve the hybrid modelling framework for pollution source identification. Our results showed that hybrid PBM-ANNs-PSO models could be applied to identify the pollution source and quantify release intensity in spatial distribution when the discharge type was assumed as the point source with a continuous release. Multiple-performance criteria metrics, in terms of the coefficient of determination, root-mean-square error, mean absolute error, evaluated the model performance as "Excellent prediction". The BP-PSO models consistently appear to be the top-performing source identification model within the developed models, with most cases of relative error (RE) values lower than 5%. The new insights from the hybrid modelling framework would provide useful information for the local government agency to make reasonable decisions regarding pollution source identification issues. [Display omitted] • The modelling framework was developed to identify the multiple pollution sources in surface water. • The modelling framework was tested in both flume and estuary cases. • Backpropagation models outperform other models within the tested surrogate models. • Population size of 15–25 was found suitable for particle swarm optimization in tested cases. [ABSTRACT FROM AUTHOR]

Accurate lifetime prediction of transformers operated in power grids with renewable energy systems is a challenging task because it requires a large amount of data that is not usually available. In the case of wind energy, this complexity is intensified with the stochastic ageing process influenced by the intermittency of the wind and weather conditions. Existing models make use of detailed power topologies to evaluate transformer stress profiles and associated degradation. However, this modelling approach requires high computational resources and long simulation times. In this context, this paper presents a lifetime prediction model for transformers designed through probabilistic machine learning, thermal modelling and ageing analysis. The proposed model is compared with synthetic wind-to-power detailed simulations of a wind farm and validated with real data. The lifetime prediction is evaluated with different mission profile estimates and results show that the accuracy of the probabilistic machine learning model is very high, with an error of 0.47% for the median value and 80% prediction interval errors within 6%–7% with respect to observations. Moreover, there is a substantial reduction in the simulation time and memory requirements when compared to the synthetic model. A detailed sensitivity analysis demonstrates the influence on transformer ageing of different overloading strategies, thermal constants and the geographic location of the wind farm. • Probabilistic ML-aided lifetime prediction in transformers for wind energy. • Detailed wind-to-power models validated with real wind power data. • Very high model accuracy with substantial reduction in simulation time and memory. • Capture prediction uncertainty and connect with transformer lifetime estimation. • Detailed sensitivity analysis: overloading strategy, thermal constants and location. [ABSTRACT FROM AUTHOR]

Wind conditions in South Africa are suitable for small-scale wind turbines, with wind speeds below 7 m.s−1. This investigation is about a methodology to optimise a full wind turbine using a surrogate model. A previously optimised turbine was further optimised over a range of wind speeds in terms of a new parameterisation methodology for the aerodynamic profile of the turbine blades, using non-uniform rational B-splines to encompass a wide range of possible shapes. The optimisation process used a genetic algorithm to evaluate an input vector of 61 variables, which fully described the geometry, wind conditions and rotational speed of the turbine. The optimal performance was assessed according to a weighted coefficient of power, which rated the turbine blade’s ability to extract power from the available wind stream. This methodology was validated using XFOIL to assess the final solution. The results showed that the surrogate model was successful in providing an optimised solution and, with further refinement, could increase the coefficient of power obtained.

This work examines the reduced-cost design optimization of dual- and multi-band antennas. The primary challenge is independent yet simultaneous control of the antenna responses at two or more frequency bands. In order to handle this task, a feature-based optimization approach is adopted where the design objectives are formulated on the basis of the coordinates of so-called characteristic points (or response features) of the antenna response. Due to only slightly nonlinear dependence of the feature points on antenna geometry parameters, optimization can be attained at a low computational cost. Our approach is demonstrated using two antenna structures with the optimum designs obtained in just a few dozen of EM simulations of the respective structure.

This article presents a novel framework for the multi-objective optimization of offshore renewable energy mooring systems using a random forest based surrogate model coupled to a genetic algorithm. This framework is demonstrated for the optimization of the mooring system for a floating offshore wind turbine highlighting how this approach can aid in the strategic design decision making for real-world problems faced by the offshore renewable energy sector. This framework utilizes validated numerical models of the mooring system to train a surrogate model, which leads to a computationally efficient optimization routine, allowing the search space to be more thoroughly searched. Minimizing both the cost and cumulative fatigue damage of the mooring system, this framework presents a range of optimal solutions characterizing how design changes impact the trade-off between these two competing objectives. [ABSTRACT FROM AUTHOR]

Constructing approximation models with surrogate modelling is often carried out in engineering design to save computational cost. However, the problem of the 'curse of dimensionality' still exists, and high-dimensional model representation (HDMR) has been proven to be very efficient in solving high-dimensional, computationally expensive black-box problems. This article proposes a new HDMR by combining separate stand-alone metamodels to form an ensemble based on cut-HDMR. It can improve prediction accuracy and alleviate prediction uncertainty for different problems compared with previous HDMRs. In this article, 10 representative mathematical examples and two engineering examples are used to illustrate the proposed technique and previous HDMRs. Furthermore, a comprehensive comparison of four metrics between the ensemble HDMR and other single HDMRs is presented, with a wide scope of dimensionalities. The results show that the single HDMRs perform well on specified examples but the ensemble HDMR provides more accurate predictions for all the test problems. [ABSTRACT FROM AUTHOR]

The present paper discusses a new methodology to assess stress-cycle fatigue design using meta-modelling. Research on its application is presented for offshore wind turbine towers. Kriging models are used to surrogate the complex time-domain stress-cycle fatigue assessment that demands multiple evaluations in the design phase. The presented development highlights the importance of having a notion of improvement for the problem of meta-modelling. Literature shows that, when meta-modelling complex engineering problems, the idea of improvement is not always considered. To tackle the problem of assessing fatigue in the design phase, a learning criterion is introduced. It has the particularity of relating to the physical description of the stress-cycle fatigue. Results of the proposed criterion are compared with meta-modelling using a Latin hypercube sampling, and with the standard design approach that bins environmental conditions for fatigue design calculations. A full 1 year validation sample is used to study convergence. As surrogate models for stress-cycle fatigue, Kriging models significantly decrease the efforts of the design procedure. Results showed that computational efforts can be reduced consistently by a factor of 5–8 without compromising accuracy. This may correspond to a reduction of up to approximately 85% of the effort needed to assess stress-cycle fatigue in the design phase. To conclude, it is important to highlight that the methodology presented has a universal character. It can be implemented to reduce computational time or assess the probabilistic response with only one requirement, the definition of a single representative indicator. In the case of fatigue, the short-term stress-cycle damage rate was considered. [ABSTRACT FROM AUTHOR]

The use of high-fidelity computational simulations promises to enable high-throughput hypothesis testing and optimisation of cancer therapies. However, increasing realism comes at the cost of increasing computational requirements. This article explores the use of surrogate-assisted evolutionary algorithms to optimise the targeted delivery of a therapeutic compound to cancerous tumour cells with the multicellular simulator, PhysiCell. The use of both Gaussian process models and multi-layer perceptron neural network surrogate models are investigated. We find that evolutionary algorithms are able to effectively explore the parameter space of biophysical properties within the agent-based simulations, minimising the resulting number of cancerous cells after a period of simulated treatment. Both model-assisted algorithms are found to outperform a standard evolutionary algorithm, demonstrating their ability to perform a more effective search within the very small evaluation budget. This represents the first use of efficient evolutionary algorithms within a high-throughput multicellular computing approach to find therapeutic design optima that maximise tumour regression. [ABSTRACT FROM AUTHOR]

The first CFD-enabled multi-objective design optimisation methodology for continuous oscillatory baffled reactors (COBRs), used for flow chemistry-based process development, is described, where performance is quantified in terms of two metrics: a mixing efficiency index and the variance of the residence time distribution. The effect of cross-validation approaches on the surrogate modelling of these performance metrics is examined in detail and the resultant surrogate models used to demonstrate the influence of key design variables. Pareto fronts of non-dominated solutions are presented to illustrate the available design compromises for COBR performance and it is shown that these can give a narrow Residence Time Distribution and good mixing within the final design. The novel feature of offset baffles within a channel, explored here for the first time, is identified as a key parameter in improving the performance of COBRs. [ABSTRACT FROM AUTHOR]