Computational Effort Research Articles

Stochastic Black-Box Optimization using Multi-Fidelity Score Function Estimator

Abstract Optimizing parameters of physics-based simulators is crucial in the design pro-
cess of engineering and scientific systems. This becomes particularly challenging
when the simulator is stochastic, computationally expensive, black-box and when a
high-dimensional vector of parameters needs to be optimized, as e.g. is the case
in complex climate models that involve numerous interdependent variables and
uncertain parameters. Many traditional optimization methods rely on gradient
information, which is frequently unavailable in legacy black-box codes. To address
these challenges, we present SCOUT-Nd (Stochastic Constrained Optimization for
N dimensions), a gradient-based algorithm that can be used on non-differentiable
objectives. It can be combined with natural gradients in order to further enhance
convergence properties. and it also incorporates multi-fidelity schemes and an
adaptive selection of samples in order to minimize computational effort. We vali-
date our approach using standard, benchmark problems, demonstrating its superior
performance in parameter optimization compared to existing methods. Addition-
ally, we showcase the algorithm’s efficacy in a complex real-world application, i.e.
the optimization of a wind farm layout.

DC-YOLO: an improved field plant detection algorithm based on YOLOv7-tiny

Weeding is an important part of agricultural production. With the development of science and technology, automated weeding is regarded as the future development direction, and how to accurately and efficiently detect plants in the field is one of the key points. Corn seedlings and weeds are similar in color, shape and other characteristics, which brings serious challenges to plant detection. In this paper, we propose an improved model based on YOLOv7-tiny, called DC-YOLO. To improve the extraction of key features in the model, we propose Dual Coordinate Attention model (DCA). In addition, we introduce the Content-Aware ReAssembly of FEatures (CARAFE) operator to represent the up-sampling process as a learnable feature reorganization, which enriches the feature information of the sampled images. Finally, we decoupled the detection head to minimize conflicts between features from different tasks. The results show that applying the proposed method to corn and weed datasets, the detection accuracy of the model reaches 95.7% mean Average Precision (mAP@0.5), the computational effort of the model is 13.083 Giga Floating-point Operations (GFLOPs), and the parameter size is 5.223 Millon (M), which is better than the rest of the mainstream light-weight target detection model.

A Hybrid Modelling Approach Coupling Physics-based Simulation and Deep Learning for Battery Electrode Manufacturing Simulations

Lithium-ion battery (LIB) performance is significantly influenced by its manufacturing process. Manufacturing of an optimized electrode can incur high production costs such as high energy consumption, high scrap rates and emissions. This is due to the process that consists of a series of manufacturing steps presenting a complex interrelationship, thus limiting the understanding of performance as a function of manufacturing parameters. While several empirical and computational methods are employed for optimization, they are demanding in terms of resources such as materials or computational effort. By leveraging Deep Learning (DL), we can enhance our understanding of the complex manufacturing processes and accelerate its optimization. We propose a data-driven supervised DL methodology to complement physics-based LIB cathode manufacturing simulations. The trained DL-based predictive model integrates well into the manufacturing simulation framework to forecast cathode slurry microstructures. The DL model demonstrates robust predictive performance for LIB NMC-111 and LiFePO4–based slurries and slurries for a solid-state battery NMC-622/argyrodite composite electrode preparation. While the current work is focused on the cathode slurry process, the proposed methodology has potential for application to drying and calendering steps. This approach will be helpful in streamlining lab-scale electrode manufacturing, and reducing errors, waste and resource consumption.

Automated Finite Element Analysis of Stiffened Tanks Using Multifidelity Modeling

Accurate structural analysis of stiffened thin-walled pressure vessels enables critical design decisions to be made during the conceptual and baseline design phases. Furthermore, having multifidelity finite element models that couple shell and solid representations facilitates exploring many design iterations quickly with sufficient accuracy. However, multifidelity model creation and utilization can be cumbersome to perform interactively, such that automated model creation, mesh generation, analysis submission, and results postprocessing are highly desirable. Of particular difficulty is generating multifidelity meshes robustly. A framework for automatic multifidelity model development using parameterized scripting in commercial finite element packages is developed to address these critical modeling needs. Issues associated with automated meshing of highly complex assemblies are examined, and a volume decomposition scheme based upon face partitions is proposed. Application is made to a complex stiffened thin-walled pressure vessel with full shell and solid/shell multifidelity model accuracies assessed by comparison against all solid models. Results indicate the proposed technique’s potential for design iteration improvements with significant reductions in interactive model preparation and postprocessing costs, as well as overall computational effort. Potential novel applications and improvements to the presented framework that can enable its application to complex problems are discussed.

Capability of different multiaxial fatigue evaluation approaches on adhesively butt-bonded hollow cylinders under multiaxial loading with variable amplitudes

Currently, there is a lack of reliable approaches for the fatigue assessment of adhesively bonded joints that are subject to multiaxial stresses with variable amplitudes. To improve the current situation, investigations are made on the capability of three multiaxial evaluation approaches based on fatigue tests of adhesively butt-bonded hollow cylinders under multiaxial loading with constant phase shift and variable amplitudes.The investigated approaches are the Gough-Pollard criterion, the Findley criterion and a new method presented here, which is based on a multiaxial counting method and an invariant equivalent stress hypothesis (MCES-Method).While the Gough-Pollard criterion has a low computational effort, it does not reflect the fatigue life extending effect due to a phase shift without using a fitting parameter that is not physically based. The Findley criterion is less suitable for the investigated multiaxial stress states (with and without phase shift), although the fatigue life prediction under these load conditions is at least conservative. The MCES-Method has the highest accuracy of the methods considered. By combining research findings from published papers in the context of adhesively bonded joints, all investigated load scenarios are estimated with high prognosis quality.

Using calibrated minimum input models to predict retrofit performance on residential dwellings

Accurate energy models for existing buildings depend on the appropriate calibration of a large number of unknown parameters against detailed monitoring data. This process typically requires significant computational effort and often produces an underdetermined model. Simplifying the model by reducing unknown parameters can alleviate these challenges. This study assesses whether a previously introduced minimum input method, which uses a subset of significant parameters for simplified calibration, can predict post-retrofit performance. The paper compares the predictive power of this new method with conventional calibration, using post-retrofit monitored data from three social housing dwellings. Results indicate that minimum input models achieve significant improvements in retrofit prediction from uncalibrated models and can outperform conventionally calibrated models. On average, minimum input models reduced CVRMSE by 5.6% from baseline, slightly less than the 6.3% improvement from full calibration. Our results support simplified calibration as a viable alternative for predicting thermal performance from residential retrofits.

A block-based numerical strategy for modeling the dynamic out-of-plane behavior of unreinforced brick masonry walls

AbstractIn this paper, a numerical procedure is proposed to simulate the dynamic out-of-plane response of unreinforced masonry (URM) walls. A state-of-the-art damaging block-based model, originally developed for quasi-static simulations, is extended for the first time in a dynamic regime. The blocks are represented using solid 3D finite elements governed by a plastic-damage constitutive law for both tension and compression. A cohesive-frictional contact-based formulation is used to account for interactions between the blocks. A simplified mechanical characterization is formulated to improve efficiency in wall-level analyses. Dynamic simulation is performed using a generalized HHT-$$\alpha$$ α direct integration implicit solver and by implementing Rayleigh damping in the bulk. Such consideration allows the use of both mass and stiffness proportional terms of the Rayleigh damping without compromising efficiency. The strategy is applied to simulate incremental dynamic experiments performed on full-scale walls, showing good agreement between numerical and experimental results. The calibrated numerical model is then optimized to reduce computational effort while maintaining accuracy. The optimized model is used to investigate the effect of relative support motion on the one-way bending out-of-plane seismic response of URM walls, demonstrating the potential of the modeling strategy to explore the effect of boundary conditions that occur in real buildings but are often overlooked in laboratory experiments. This investigation also explores the adequacy of simplifications in capturing the effect of relative support motion, which can be adopted for simple modeling strategies commonly used in standard engineering practice.

Gas kick dynamic circulation in MPD operations with water based drilling fluid: Maximum casing pressure modeling and validation

The use of water-based drilling fluids enables the rapid detection of a gas influx by pit gain monitoring due to the low solubility of gases in the aqueous medium. Moreover, as the influx remains as a free gas, it undergoes a process of expansion the further it rises through the well, changing the pressure profiles and producing a pressure peak when it reaches the surface. Therefore, the gas influx must be controlled appropriately. Otherwise, the maximum pressure experienced during the kick circulation may exceed the safety limits of surface equipment, placing the operation and the rig crew in a hazardous situation. The Managed Pressure Drilling (MPD) technique allows for a more precise detection and control of the influx while retaining the ability to circulate dynamically small volumes of gas kick through the riser, minimizing the risks involved and the non-productive time (NPT). However, the decision of when to apply dynamic or conventional circulation techniques must be considered carefully, taking into account factors such as predictions of pressure peaks, as well as intensity and volume of the kick. In this way, it is paramount that the maximum pressure during the circulation be estimated before the operation begins to ensure that the decision is made correctly. This paper presents the results of a gas kick circulation simulator specifically developed to predict pressure peaks. This simulator uses a mathematical model based on algebraic equations whose solution requires low computational effort and, therefore, regarding gas kick incidents, is an interesting tool that can aid in guiding decision-making. The results of pressure peaks were successfully validated by employing literature data, experimental drilling setup runs and simulations performed in a commercial software largely consolidated in the petroleum industry: Drillbench.

GEEES: inferring cell-specific gene-enhancer interactions from multi-modal single-cell data.

Gene-enhancer interactions are central to transcriptional regulation. Current multi-modal single-cell datasets that profile transcriptome and chromatin accessibility simultaneously in a single cell are yielding opportunities to infer gene-enhancer associations in a cell type specific manner. Computational efforts for such multi-modal single-cell datasets thus far focused on methods for identification and refinement of cell types and trajectory construction. While initial attempts for inferring gene-enhancer interactions have emerged, these have not been evaluated against benchmark datasets that materialized from bulk genomic experiments. Furthermore, existing approaches are limited to inferring gene-enhancer associations at the level of grouped cells as opposed to individual cells, thereby ignoring regulatory heterogeneity among the cells. We present a new approach, GEEES for "Gene EnhancEr IntEractions from Multi-modal Single Cell Data," for inferring gene-enhancer associations at the single-cell level using multi-modal single-cell transcriptome and chromatin accessibility data. We evaluated GEEES alongside several multivariate regression-based alternatives we devised and state-of-the-art methods using a large number of benchmark datasets, providing a comprehensive assessment of current approaches. This analysis revealed significant discrepancies between gold-standard interactions and gene-enhancer associations derived from multi-modal single-cell data. Notably, incorporating gene-enhancer distance into the analysis markedly improved performance across all methods, positioning GEEES as a leading approach in this domain. While the overall improvement in performance metrics by GEEES is modest, it provides enhanced cell representation learning which can be leveraged for more effective downstream analysis. Furthermore, our review of existing experimentally driven benchmark datasets uncovers their limited concordance, underscoring the necessity for new high-throughput experiments to validate gene-enhancer interactions inferred from single-cell data. https://github.com/keleslab/GEEES.

Neural Network Architectures and Magnetic Hysteresis: Overview and Comparisons

The development of innovative materials, based on the modern technologies and processes, is the key factor to improve the energetic sustainability and reduce the environmental impact of electrical equipment. In particular, the modeling of magnetic hysteresis is crucial for the design and construction of electrical and electronic devices. In recent years, additive manufacturing techniques are playing a decisive role in the project and production of magnetic elements and circuits for applications in various engineering fields. To this aim, the use of the deep learning paradigm, integrated with the most common models of the magnetic hysteresis process, has become increasingly present in recent years. The intent of this paper is to provide the features of a wide range of deep learning tools to be applied to magnetic hysteresis context and beyond. The possibilities of building neural networks in hybrid form are innumerable, so it is not plausible to illustrate them in a single paper, but in the present context, several neural networks used in the scientific literature, integrated with various hysteretic mathematical models, including the well-known Preisach model, are compared. It is shown that this hybrid approach not only improves the modeling of hysteresis by significantly reducing computational time and efforts, but also offers new perspectives for the analysis and prediction of the behavior of magnetic materials, with significant implications for the production of advanced devices.

Simulation of Flood-Control Reservoirs: Comparing Fully 2D and 0D–1D Models

Flood-control reservoirs are often used as a structural measure to mitigate fluvial floods, and numerical models are a fundamental tool for assessing their effectiveness. This work aims to analyze the suitability of fully 2D shallow-water models to simulate these systems by adopting internal boundary conditions to describe hydraulic structures (i.e., dams) and by using a parallelized code to reduce the computational burden. The 2D results are also compared with the more established approach of coupling a 1D model for the river and a 0D model for the reservoir. Two test cases, including an in-stream reservoir and an off-stream basin, both located in Italy, are considered. Results show that the fully 2D model can effectively handle the simulation of a complex flood-control system. Moreover, compared with the 0D–1D model, it captures the velocity field and the filling/emptying process of the reservoir more realistically, especially for off-stream reservoirs. Conversely, when the basin is characterized by very limited flood dynamics, the two approaches provide similar results (maximum levels in the reservoir differ by less than 10 cm, and peak discharges by about 5%). Thanks to parallelization and the inclusion of internal boundary conditions, fully 2D models can be applied not only for local hydrodynamic analyses but also for river-scale studies, including flood-control reservoirs, with reasonable computational effort (i.e., ratios of physical to computational times on the order of 30–100).

Efficient computation of magnetic polarizability tensor spectral signatures for object characterisation in metal detection

PurposeMagnetic polarizability tensors (MPTs) provide an economical characterisation of conducting magnetic metallic objects and their spectral signature can aid in the solution of metal detection inverse problems, such as scrap metal sorting, searching for unexploded ordnance in areas of former conflict and security screening at event venues and transport hubs. In this work, the authors aim to discuss methods for efficiently building large dictionaries for classification approaches.Design/methodology/approachPrevious work has established explicit formulae for MPT coefficients, underpinned by a rigorous mathematical theory. To assist with the efficient computation of MPTs at differing parameters and objects of interest, this work applies new observations about the way the MPT coefficients can be computed. Furthermore, the authors discuss discretisation strategies for hp-finite elements on meshes of unstructured tetrahedra combined with prismatic boundary layer elements for resolving thin skin depths and using an adaptive proper orthogonal decomposition (POD) reduced-order modelling methodology to accelerate computations for varying parameters.FindingsThe success of the proposed methodologies is demonstrated using a series of examples. A significant reduction in computational effort is observed across all examples. The authors identify and recommend a simple discretisation strategy and improved accuracy is obtained using adaptive POD.Originality/valueThe authors present novel computations, timings and error certificates of MPT characterisations of realistic objects made of magnetic materials. A novel postprocessing implementation is introduced and an adaptive POD algorithm is demonstrated.

Helmet Wearing Detection Algorithm Based on YOLOv5s-FCW

An enhanced algorithm, YOLOv5s-FCW, is put forward in this study to tackle the problems that exist in the current helmet detection (HD) methods. These issues include having too many parameters, a complex network, and large computation requirements, making it unsuitable for deployment on embedded and other devices. Additionally, existing algorithms struggle with detecting small targets and do not achieve high enough recognition accuracy. Firstly, the YOLOv5s backbone network is replaced by FasterNet for feature extraction (FE), which reduces the number of parameters and computational effort in the network. Secondly, a convolutional block attention module (CBAM) is added to the YOLOv5 model to improve the detection model’s ability to detect small objects such as helmets by increasing its attention to them. Finally, to enhance model convergence, the WIoU_Loss loss function is adopted instead of the GIoU_Loss loss function. As reported by the experimental results, the YOLOv5s-FCW algorithm proposed in this study has improved accuracy by 4.6% compared to the baseline algorithm. The proposed approach not only enhances detection concerning small and obscured targets but also reduces computation for the YOLOv5s model by 20%, thereby decreasing the hardware cost while maintaining a higher average accuracy regarding detection.

An anisotropic eigenfracture approach accounting for mixed fracture modes in wooden structures by the Representative Crack Element framework

Finite Element analysis of anisotropic fracture phenomena in wood is a challenging task, particularly when dealing with intricate loading scenarios and mode-specific behavior. The appeal of energetically motivated approaches, such as the eigenfracture method, is that they enable simulation of fracture without prior knowledge of the crack path. The promising eigenfracture method has shown good numerical performance for isotropic materials, and this contribution showcases its application to anisotropic materials. Wood is one such anisotropic material and in this manuscript, the directional dependence of both elasticity and fracture evolution are incorporated into the eigenfracture approach. Further, the eigenfracture approach is used in conjunction with Representative Crack Elements (RCE), which permit accurate modeling of physical crack deformations. The governing equations are systematically derived and implemented into the Finite Element framework. By representative numerical examples, some advantages over the alternative phase-field method are demonstrated. Another highlight of this work is that it is possible to provide a realistic ratio of the energy release rates parallel to and perpendicular to the fiber direction in order to achieve physically accurate crack patterns. Additionally, the calculation effort is reduced, because the unknowns required to determine the crack kinematics can be solved analytically at the material level, a feature that also enables parallelization.

Optimal siting and sizing of battery energy storage systems in unbalanced distribution systems: A multi objective problem under uncertainty

In this paper the siting and sizing problem of battery energy storage systems in unbalanced active distribution systems is formulated as a mixed-integer, non-linear, constrained multi objective (MO) optimization problem under uncertainties. The problem is cumbersome from the computational point of view due to the presence of intertemporal constraints, a great number of state variables and the presence of uncertainties in the problem input data. A new approach based on the trade-off/risk analysis is proposed to obtain with acceptable computational efforts a solution that may not be the optimal solution but represents a reasonable and robust compromise. We use the trade-off/risk analysis, because it was specifically developed for power system planning problems in which we deal with a wide range of options, with possible conflicting objectives, and with uncertainty and risk. The proposed approach includes new procedures to select an adequate set of planning alternatives to be considered in the trade-off/risk analysis framework and to assist the planning engineer when difficulties arise in setting probabilities of the input data. Numerical applications to an IEEE unbalanced test system demonstrate the effectiveness of the proposed procedure and indicate the best alternatives of storage systems in the range from 450 kW to 600 kW globally installed in a reduced set of nodes.

Calibration-Based ALE Model Order Reduction for Hyperbolic Problems with Self-Similar Travelling Discontinuities

We propose a novel Model Order Reduction framework that is able to handle solutions of hyperbolic problems characterized by multiple travelling discontinuities. By means of an optimization based approach, we introduce suitable calibration maps that allow us to transform the original solution manifold into a lower dimensional one. The novelty of the methodology is represented by the fact that the optimization process does not require the knowledge of the discontinuities location. The optimization can be carried out simply by choosing some reference control points, thus avoiding the use of some implicit shock tracking techniques, which would translate into an increased computational effort during the offline phase. In the online phase, we rely on a non-intrusive approach, where the coefficients of the projection of the reduced order solution onto the reduced space are recovered by means of an Artificial Neural Network. To validate the methodology, we present numerical results for the 1D Sod shock tube problem, for the 2D double Mach reflection problem, also in the parametric case, and for the triple point problem.

Towards verifiable cancer digital twins: tissue level modeling protocol for precision medicine.

Cancer exhibits substantial heterogeneity, manifesting as distinct morphological and molecular variations across tumors, which frequently undermines the efficacy of conventional oncological treatments. Developments in multiomics and sequencing technologies have paved the way for unraveling this heterogeneity. Nevertheless, the complexity of the data gathered from these methods cannot be fully interpreted through multimodal data analysis alone. Mathematical modeling plays a crucial role in delineating the underlying mechanisms to explain sources of heterogeneity using patient-specific data. Intra-tumoral diversity necessitates the development of precision oncology therapies utilizing multiphysics, multiscale mathematical models for cancer. This review discusses recent advancements in computational methodologies for precision oncology, highlighting the potential of cancer digital twins to enhance patient-specific decision-making in clinical settings. We review computational efforts in building patient-informed cellular and tissue-level models for cancer and propose a computational framework that utilizes agent-based modeling as an effective conduit to integrate cancer systems models that encode signaling at the cellular scale with digital twin models that predict tissue-level response in a tumor microenvironment customized to patient information. Furthermore, we discuss machine learning approaches to building surrogates for these complex mathematical models. These surrogates can potentially be used to conduct sensitivity analysis, verification, validation, and uncertainty quantification, which is especially important for tumor studies due to their dynamic nature.

Analysing Flexural Response in RC Beams: A Closed-Form Solution Designer Perspective from Detailed to Simplified Modelling

This paper presents a detailed analytical approach for the bending analysis of reinforced concrete beams, integrating both structural mechanics principles and Eurocode 2 provisions. The general analytical expressions derived for the curvature were applied for the transverse displacement analysis of a simply supported reinforced concrete beam under four-point loading, focusing on key limit states: the initiation of cracking, the yielding of tensile reinforcement and the compressive failure of concrete. The displacement’s results were validated through experimental testing, showing a high degree of accuracy in the elastic and crack propagation phases. Deviations in the yielding phase were attributed to the conservative material assumptions within the Eurocode 2 framework, though the analytical model remained reliable overall. To streamline the computational process for more complex structures, a simplified model utilising a non-linear rotational spring was further developed. This model effectively captures the influence of cracking with significantly reduced computational effort, making it suitable for serviceability limit state analyses in complex loading scenarios, such as seismic impacts. The results demonstrate that combining detailed analytical methods with this simplified model provides an efficient and practical solution for the analysis of reinforced concrete beams, balancing precision with computational efficiency.

Deep Learning‐Driven Modeling of Dynamic Acoustic Sensing in Biomimetic Soft‐Robotic Pinnae

ABSTRACTBiological function often depends on complex mechanisms of a dynamic, time‐variant nature. An example is certain bat species (horseshoe bats—Rhinolophidae) that use intricate pinna musculatures to execute a variety of pinna deformations. While prior work has indicated the potential significance of these motions for sensory information encoding, it remains unclear how the complex time‐variant pinna geometries could be controlled to enhance sensory performance. To address this issue, this work has investigated deep neural network models as digital twins for biomimetic pinnae. The networks were trained to predict the acoustic impacts of the deformed pinna geometries. A total of three network architectures have been evaluated for this purpose using physical numerical simulations (boundary element method) as ground truth. The networks predicted the acoustic beampattern function from pinna shape or even directly from the states of actuators that were used to deform the pinna shapes in simulation. Inserting prior knowledge in the form of beam‐shaped basis functions did not improve network performance. The ability of the networks to produce beampattern predictions with low computational effort (in about three milliseconds each) should lend itself readily to supporting learning methods such as deep reinforcement learning that require many such functional evaluations.

Towards Self-Conscious AI Using Deep ImageNet Models: Application for Blood Cell Classification

The exceptional performance of ImageNet competition winners in image classification has led AI researchers to repurpose these models for a whole range of tasks using transfer learning (TL). TL has been hailed for boosting performance, shortening learning time and reducing computational effort. Despite these benefits, issues such as data sparsity and the misrepresentation of classes can diminish these gains, occasionally leading to misleading TL accuracy scores. This research explores the innovative concept of endowing ImageNet models with a self-awareness that enables them to recognize their own accumulated knowledge and experience. Such self-awareness is expected to improve their adaptability in various domains. We conduct a case study using two different datasets, PBC and BCCD, which focus on blood cell classification. The PBC dataset provides high-resolution images with abundant data, while the BCCD dataset is hindered by limited data and inferior image quality. To compensate for these discrepancies, we use data augmentation for BCCD and undersampling for both datasets to achieve balance. Subsequent pre-processing generates datasets of different size and quality, all geared towards blood cell classification. We extend conventional evaluation tools with novel metrics—“accuracy difference” and “loss difference”—to detect overfitting or underfitting and evaluate their utility as potential indicators for learning behavior and promoting the self-confidence of ImageNet models. Our results show that these metrics effectively track learning progress and improve the reliability and overall performance of ImageNet models in new applications. This study highlights the transformative potential of turning ImageNet models into self-aware entities that significantly improve their robustness and efficiency in various AI tasks. This groundbreaking approach opens new perspectives for increasing the effectiveness of transfer learning in real-world AI implementations.